decoder: remove mention of UnmarshalText in errors (#751)

Fixes #737

.gitattributes

@@ -0,0 +1,3 @@
+* text=auto
+
+benchmark/benchmark.toml text eol=lf

.github/PULL_REQUEST_TEMPLATE.md

@@ -1,5 +1,19 @@
-**Issue:** add link to pelletier/go-toml issue here
+<!--
+
+Thank you for your pull request!
+
+Please read the Code changes section of the CONTRIBUTING.md file,
+and make sure you have followed the instructions.
+
+https://github.com/pelletier/go-toml/blob/v2/CONTRIBUTING.md#code-changes
+
+-->
 
 Explanation of what this pull request does.
 
-More detailed description of the decisions being made and the reasons why (if the patch is non-trivial).
+More detailed description of the decisions being made and the reasons why (if
+the patch is non-trivial).
+
+---
+
+Paste `benchstat` results here

.github/dependabot.yml

@@ -1,6 +1,17 @@
 version: 2
 updates:
-  - package-ecosystem: "gomod"
-    directory: "/" # Location of package manifests
-    schedule:
-      interval: "daily"
+- package-ecosystem: gomod
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10
+- package-ecosystem: github-actions
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10
+- package-ecosystem: docker
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10

.github/release.yml

@@ -0,0 +1,21 @@
+changelog:
+  exclude:
+    labels:
+      - build
+      - testing
+  categories:
+    - title: What's new
+      labels:
+        - feature
+    - title: Performance
+      labels:
+        - performance
+    - title: Fixed bugs
+      labels:
+        - bug
+    - title: Documentation
+      labels:
+        - doc
+    - title: Other changes
+      labels:
+        - "*"

.github/workflows/coverage.yml

@@ -0,0 +1,20 @@
+name: coverage
+on:
+  pull_request:
+    branches:
+      - v2
+
+jobs:
+  report:
+    runs-on: "ubuntu-latest"
+    name: report
+    steps:
+      - uses: actions/checkout@master
+        with:
+          fetch-depth: 0
+      - name: Setup go
+        uses: actions/setup-go@master
+        with:
+          go-version: 1.18
+      - name: Run tests with coverage
+        run: ./ci.sh coverage -d "${GITHUB_BASE_REF-HEAD}"

.github/workflows/release.yml

@@ -0,0 +1,39 @@
+name: release
+on:
+  push:
+    tags:
+      - "v2.*"
+  workflow_call:
+    inputs:
+      args:
+        description: "Extra arguments to pass goreleaser"
+        default: ""
+        required: false
+        type: string
+
+jobs:
+  release:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 0
+      - name: Set up Go
+        uses: actions/setup-go@v2
+        with:
+          go-version: 1.18
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v1
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Run GoReleaser
+        uses: goreleaser/goreleaser-action@v2
+        with:
+          distribution: goreleaser
+          version: latest
+          args: release ${{ inputs.args }} --rm-dist
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/workflow.yml

@@ -12,17 +12,21 @@ jobs:
     strategy:
       matrix:
         os: [ 'ubuntu-latest', 'windows-latest', 'macos-latest']
-        go: [ '1.15', '1.16' ]
+        go: [ '1.17', '1.18' ]
     runs-on: ${{ matrix.os }}
     name: ${{ matrix.go }}/${{ matrix.os }}
     steps:
       - uses: actions/checkout@master
+        with:
+          fetch-depth: 0
       - name: Setup go ${{ matrix.go }}
         uses: actions/setup-go@master
         with:
           go-version: ${{ matrix.go }}
       - name: Run unit tests
         run: go test -race ./...
-      - name: Run benchmark tests
-        run: go test -race ./...
-        working-directory: benchmark
+  release-check:
+    if: ${{ github.ref != 'refs/heads/v2' }}
+    uses: pelletier/go-toml/.github/workflows/release.yml@v2
+    with:
+      args: --snapshot

.gitignore

@@ -3,3 +3,4 @@ fuzz/
 cmd/tomll/tomll
 cmd/tomljson/tomljson
 cmd/tomltestgen/tomltestgen
+dist

.golangci.toml

@@ -4,6 +4,9 @@ golangci-lint-version =  "1.39.0"
 [linters-settings.wsl]
 allow-assign-and-anything = true
 
+[linters-settings.exhaustive]
+default-signifies-exhaustive = true
+
 [linters]
 disable-all = true
 enable = [
@@ -18,13 +21,13 @@ enable = [
     "errcheck",
     "errorlint",
     "exhaustive",
-    "exhaustivestruct",
+    # "exhaustivestruct",
     "exportloopref",
     "forbidigo",
-    "forcetypeassert",
+    # "forcetypeassert",
     "funlen",
     "gci",
-    "gochecknoglobals",
+    # "gochecknoglobals",
     "gochecknoinits",
     "gocognit",
     "goconst",
@@ -32,7 +35,7 @@ enable = [
     "gocyclo",
     "godot",
     "godox",
-    "goerr113",
+    # "goerr113",
     "gofmt",
     "gofumpt",
     "goheader",
@@ -45,7 +48,7 @@ enable = [
     "gosec",
     "gosimple",
     "govet",
-    "ifshort",
+    # "ifshort",
     "importas",
     "ineffassign",
     "lll",
@@ -54,10 +57,10 @@ enable = [
     "nakedret",
     "nestif",
     "nilerr",
-    "nlreturn",
+    # "nlreturn",
     "noctx",
     "nolintlint",
-    "paralleltest",
+    #"paralleltest",
     "prealloc",
     "predeclared",
     "revive",
@@ -76,6 +79,6 @@ enable = [
     "varcheck",
     "wastedassign",
     "whitespace",
-    "wrapcheck",
-    "wsl"
+    # "wrapcheck",
+    # "wsl"
 ]

.goreleaser.yaml

@@ -0,0 +1,111 @@
+before:
+  hooks:
+    - go mod tidy
+    - go fmt ./...
+    - go test ./...
+builds:
+  - id: tomll
+    main: ./cmd/tomll
+    binary: tomll
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - windows_amd64
+      - darwin_amd64
+      - darwin_arm64
+  - id: tomljson
+    main: ./cmd/tomljson
+    binary: tomljson
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - windows_amd64
+      - darwin_amd64
+      - darwin_arm64
+  - id: jsontoml
+    main: ./cmd/jsontoml
+    binary: jsontoml
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - windows_amd64
+      - darwin_amd64
+      - darwin_arm64
+universal_binaries:
+  - id: tomll
+    replace: true
+    name_template: tomll
+  - id: tomljson
+    replace: true
+    name_template: tomljson
+  - id: jsontoml
+    replace: true
+    name_template: jsontoml
+archives:
+- id: jsontoml
+  format: tar.xz
+  builds:
+    - jsontoml
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+- id: tomljson
+  format: tar.xz
+  builds:
+    - tomljson
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+- id: tomll
+  format: tar.xz
+  builds:
+    - tomll
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+dockers:
+  - id: tools
+    goos: linux
+    goarch: amd64
+    ids:
+      - jsontoml
+      - tomljson
+      - tomll
+    image_templates:
+      - "ghcr.io/pelletier/go-toml:latest"
+      - "ghcr.io/pelletier/go-toml:{{ .Tag }}"
+      - "ghcr.io/pelletier/go-toml:v{{ .Major }}"
+    skip_push: false
+checksum:
+  name_template: 'sha256sums.txt'
+snapshot:
+  name_template: "{{ incpatch .Version }}-next"
+release:
+  github:
+    owner: pelletier
+    name: go-toml
+  draft: true
+  prerelease: auto
+  mode: replace
+changelog:
+  use: github-native
+announce:
+  skip: true

CONTRIBUTING.md

@@ -1,74 +1,74 @@
-## Contributing
+# Contributing
 
 Thank you for your interest in go-toml! We appreciate you considering
 contributing to go-toml!
 
-The main goal is the project is to provide an easy-to-use TOML
-implementation for Go that gets the job done and gets out of your way –
-dealing with TOML is probably not the central piece of your project.
+The main goal is the project is to provide an easy-to-use and efficient TOML
+implementation for Go that gets the job done and gets out of your way – dealing
+with TOML is probably not the central piece of your project.
 
-As the single maintainer of go-toml, time is scarce. All help, big or
-small, is more than welcomed!
+As the single maintainer of go-toml, time is scarce. All help, big or small, is
+more than welcomed!
 
-### Ask questions
+## Ask questions
 
-Any question you may have, somebody else might have it too. Always feel
-free to ask them on the [issues tracker][issues-tracker].  We will try to
-answer them as clearly and quickly as possible, time permitting.
+Any question you may have, somebody else might have it too. Always feel free to
+ask them on the [discussion board][discussions]. We will try to answer them as
+clearly and quickly as possible, time permitting.
 
 Asking questions also helps us identify areas where the documentation needs
 improvement, or new features that weren't envisioned before. Sometimes, a
-seemingly innocent question leads to the fix of a bug. Don't hesitate and
-ask away!
+seemingly innocent question leads to the fix of a bug. Don't hesitate and ask
+away!
 
-### Improve the documentation
+[discussions]: https://github.com/pelletier/go-toml/discussions
 
-The best way to share your knowledge and experience with go-toml is to
-improve the documentation. Fix a typo, clarify an interface, add an
-example, anything goes!
+## Improve the documentation
 
-The documentation is present in the [README][readme] and thorough the
-source code. On release, it gets updated on [pkg.go.dev][pkg.go.dev]. To make a
-change to the documentation, create a pull request with your proposed
-changes. For simple changes like that, the easiest way to go is probably
-the "Fork this project and edit the file" button on Github, displayed at
-the top right of the file. Unless it's a trivial change (for example a
-typo), provide a little bit of context in your pull request description or
-commit message.
+The best way to share your knowledge and experience with go-toml is to improve
+the documentation. Fix a typo, clarify an interface, add an example, anything
+goes!
 
-### Report a bug
+The documentation is present in the [README][readme] and thorough the source
+code. On release, it gets updated on [pkg.go.dev][pkg.go.dev]. To make a change
+to the documentation, create a pull request with your proposed changes. For
+simple changes like that, the easiest way to go is probably the "Fork this
+project and edit the file" button on Github, displayed at the top right of the
+file. Unless it's a trivial change (for example a typo), provide a little bit of
+context in your pull request description or commit message.
 
-Found a bug! Sorry to hear that :(. Help us and other track them down and
-fix by reporting it. [File a new bug report][bug-report] on the [issues
-tracker][issues-tracker]. The template should provide enough guidance on
-what to include. When in doubt: add more details! By reducing ambiguity and
-providing more information, it decreases back and forth and saves everyone
-time.
+## Report a bug
 
-### Code changes
+Found a bug! Sorry to hear that :(. Help us and other track them down and fix by
+reporting it. [File a new bug report][bug-report] on the [issues
+tracker][issues-tracker]. The template should provide enough guidance on what to
+include. When in doubt: add more details! By reducing ambiguity and providing
+more information, it decreases back and forth and saves everyone time.
+
+## Code changes
 
 Want to contribute a patch? Very happy to hear that!
 
 First, some high-level rules:
 
-* A short proposal with some POC code is better than a lengthy piece of
-  text with no code. Code speaks louder than words.
-* No backward-incompatible patch will be accepted unless discussed.
-  Sometimes it's hard, and Go's lack of versioning by default does not
-  help, but we try not to break people's programs unless we absolutely have
+- A short proposal with some POC code is better than a lengthy piece of text
+  with no code. Code speaks louder than words. That being said, bigger changes
+  should probably start with a [discussion][discussions].
+- No backward-incompatible patch will be accepted unless discussed. Sometimes
+  it's hard, but we try not to break people's programs unless we absolutely have
   to.
-* If you are writing a new feature or extending an existing one, make sure
-  to write some documentation.
-* Bug fixes need to be accompanied with regression tests.
-* New code needs to be tested.
-* Your commit messages need to explain why the change is needed, even if
-  already included in the PR description.
+- If you are writing a new feature or extending an existing one, make sure to
+  write some documentation.
+- Bug fixes need to be accompanied with regression tests.
+- New code needs to be tested.
+- Your commit messages need to explain why the change is needed, even if already
+  included in the PR description.
 
-It does sound like a lot, but those best practices are here to save time
-overall and continuously improve the quality of the project, which is
-something everyone benefits from.
+It does sound like a lot, but those best practices are here to save time overall
+and continuously improve the quality of the project, which is something everyone
+benefits from.
 
-#### Get started
+### Get started
 
 The fairly standard code contribution process looks like that:
 
@@ -76,50 +76,114 @@ The fairly standard code contribution process looks like that:
 2. Make your changes, commit on any branch you like.
 3. [Open up a pull request][pull-request]
 4. Review, potential ask for changes.
-5. Merge. You're in!
+5. Merge.
 
 Feel free to ask for help! You can create draft pull requests to gather
 some early feedback!
 
-#### Run the tests
+### Run the tests
 
-You can run tests for go-toml using Go's test tool: `go test ./...`.
-When creating a pull requests, all tests will be ran on Linux on a few Go
-versions (Travis CI), and on Windows using the latest Go version
-(AppVeyor).
+You can run tests for go-toml using Go's test tool: `go test -race ./...`.
 
-#### Style
+During the pull request process, all tests will be ran on Linux, Windows, and
+MacOS on the last two versions of Go.
 
-Try to look around and follow the same format and structure as the rest of
-the code. We enforce using `go fmt` on the whole code base.
+However, given GitHub's new policy to _not_ run Actions on pull requests until a
+maintainer clicks on button, it is highly recommended that you run them locally
+as you make changes.
+
+### Check coverage
+
+We use `go tool cover` to compute test coverage. Most code editors have a way to
+run and display code coverage, but at the end of the day, we do this:
+
+```
+go test -covermode=atomic -coverprofile=coverage.out
+go tool cover -func=coverage.out
+```
+
+and verify that the overall percentage of tested code does not go down. This is
+a requirement. As a rule of thumb, all lines of code touched by your changes
+should be covered. On Unix you can use `./ci.sh coverage -d v2` to check if your
+code lowers the coverage.
+
+### Verify performance
+
+Go-toml aims to stay efficient. We rely on a set of scenarios executed with Go's
+builtin benchmark systems. Because of their noisy nature, containers provided by
+Github Actions cannot be reliably used for benchmarking. As a result, you are
+responsible for checking that your changes do not incur a performance penalty.
+You can run their following to execute benchmarks:
+
+```
+go test ./... -bench=. -count=10
+```
+
+Benchmark results should be compared against each other with
+[benchstat][benchstat]. Typical flow looks like this:
+
+1. On the `v2` branch, run `go test ./... -bench=. -count 10` and save output to
+   a file (for example `old.txt`).
+2. Make some code changes.
+3. Run `go test ....` again, and save the output to an other file (for example
+   `new.txt`).
+4. Run `benchstat old.txt new.txt` to check that time/op does not go up in any
+   test.
+
+On Unix you can use `./ci.sh benchmark -d v2` to verify how your code impacts
+performance.
+
+It is highly encouraged to add the benchstat results to your pull request
+description. Pull requests that lower performance will receive more scrutiny.
+
+[benchstat]: https://pkg.go.dev/golang.org/x/perf/cmd/benchstat
+
+### Style
+
+Try to look around and follow the same format and structure as the rest of the
+code. We enforce using `go fmt` on the whole code base.
 
 ---
 
-### Maintainers-only
+## Maintainers-only
 
-#### Merge pull request
+### Merge pull request
 
 Checklist:
 
-* Passing CI.
-* Does not introduce backward-incompatible changes (unless discussed).
-* Has relevant doc changes.
-* Has relevant unit tests.
+- Passing CI.
+- Does not introduce backward-incompatible changes (unless discussed).
+- Has relevant doc changes.
+- Benchstat does not show performance regression.
+- Pull request is [labeled appropriately][pr-labels].
+- Title will be understandable in the changelog.
 
 1. Merge using "squash and merge".
 2. Make sure to edit the commit message to keep all the useful information
    nice and clean.
 3. Make sure the commit title is clear and contains the PR number (#123).
 
-#### New release
+### New release
 
-1. Go to [releases][releases]. Click on "X commits to master since this
-   release".
-2. Make note of all the changes. Look for backward incompatible changes,
-   new features, and bug fixes.
-3. Pick the new version using the above and semver.
-4. Create a [new release][new-release].
-5. Follow the same format as [1.1.0][release-110].
+1. Decide on the next version number. Use semver.
+2. Generate release notes using [`gh`][gh]. Example:
+```
+$ gh api -X POST \
+  -F tag_name='v2.0.0-beta.5' \
+  -F target_commitish='v2' \
+  -F previous_tag_name='v2.0.0-beta.4' \
+  --jq '.body' \
+  repos/pelletier/go-toml/releases/generate-notes
+```
+3. Look for "Other changes". That would indicate a pull request not labeled
+   properly. Tweak labels and pull request titles until changelog looks good for
+   users.
+4. [Draft new release][new-release].
+5. Fill tag and target with the same value used to generate the changelog.
+6. Set title to the new tag value.
+7. Paste the generated changelog.
+8. Check "create discussion", in the "Releases" category.
+9. Check pre-release if new version is an alpha or beta.
 
 [issues-tracker]: https://github.com/pelletier/go-toml/issues
 [bug-report]: https://github.com/pelletier/go-toml/issues/new?template=bug_report.md
@@ -127,6 +191,6 @@ Checklist:
 [readme]: ./README.md
 [fork]: https://help.github.com/articles/fork-a-repo
 [pull-request]: https://help.github.com/en/articles/creating-a-pull-request
-[releases]: https://github.com/pelletier/go-toml/releases
 [new-release]: https://github.com/pelletier/go-toml/releases/new
-[release-110]: https://github.com/pelletier/go-toml/releases/tag/v1.1.0
+[gh]: https://github.com/cli/cli
+[pr-labels]: https://github.com/pelletier/go-toml/blob/v2/.github/release.yml

Dockerfile

@@ -0,0 +1,5 @@
+FROM scratch
+ENV PATH "$PATH:/bin"
+COPY tomll /bin/tomll
+COPY tomljson /bin/tomljson
+COPY jsontoml /bin/jsontoml

README.md

@@ -1,57 +1,555 @@
-# go-toml V2
+# go-toml v2
 
-Development branch. Use at your own risk.
+Go library for the [TOML](https://toml.io/en/) format.
 
-[👉 Discussion on github](https://github.com/pelletier/go-toml/discussions/471).
+This library supports [TOML v1.0.0](https://toml.io/en/v1.0.0).
 
-* `toml.Unmarshal()` should work as well as v1.
+## Development status
 
-## Must do
+This is the upcoming major version of go-toml. It is currently in active
+development. As of release v2.0.0-beta.1, the library has reached feature parity
+with v1, and fixes a lot known bugs and performance issues along the way.
 
-### Unmarshal
+If you do not need the advanced document editing features of v1, you are
+encouraged to try out this version.
 
-- [x] Unmarshal into maps.
-- [x] Support Array Tables.
-- [x] Unmarshal into pointers.
-- [x] Support Date / times.
-- [x] Support struct tags annotations.
-- [x] Support Arrays.
-- [x] Support Unmarshaler interface.
-- [x] Original go-toml unmarshal tests pass.
-- [x] Benchmark!
-- [x] Abstract AST.
-- [x] Original go-toml testgen tests pass.
-- [x] Track file position (line, column) for errors.
-- [ ] Strict mode.
-- [ ] Document Unmarshal / Decode
+[👉 Roadmap for v2](https://github.com/pelletier/go-toml/discussions/506)
 
-### Marshal
+[🐞 Bug Reports](https://github.com/pelletier/go-toml/issues)
 
-- [x] Minimal implementation
-- [x] Multiline strings
-- [ ] Multiline arrays
-- [ ] `inline` tag for tables
-- [ ] Optional indentation
-- [ ] Option to pick default quotes
+[💬 Anything else](https://github.com/pelletier/go-toml/discussions)
 
-### Document
+## Documentation
 
-- [ ] Gather requirements and design API.
+Full API, examples, and implementation notes are available in the Go
+documentation.
 
-## Ideas
+[![Go Reference](https://pkg.go.dev/badge/github.com/pelletier/go-toml/v2.svg)](https://pkg.go.dev/github.com/pelletier/go-toml/v2)
 
-- [ ] Allow types to implement a `ASTUnmarshaler` interface to unmarshal
-      straight from the AST?
-- [x] Rewrite AST to use a single array as storage instead of one allocation per
-      node.
-- [ ] Provide "minimal allocations" option that uses `unsafe` to reuse the input
-      byte array as storage for strings.
-- [x] Cache reflection operations per type.
-- [ ] Optimize tracker pass.
+## Import
 
-## Differences with v1
+```go
+import "github.com/pelletier/go-toml/v2"
+```
 
-* [unmarshal](https://github.com/pelletier/go-toml/discussions/488)
+See [Modules](#Modules).
+
+## Features
+
+### Stdlib behavior
+
+As much as possible, this library is designed to behave similarly as the
+standard library's `encoding/json`.
+
+### Performance
+
+While go-toml favors usability, it is written with performance in mind. Most
+operations should not be shockingly slow. See [benchmarks](#benchmarks).
+
+### Strict mode
+
+`Decoder` can be set to "strict mode", which makes it error when some parts of
+the TOML document was not prevent in the target structure. This is a great way
+to check for typos. [See example in the documentation][strict].
+
+[strict]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#example-Decoder.DisallowUnknownFields
+
+### Contextualized errors
+
+When most decoding errors occur, go-toml returns [`DecodeError`][decode-err]),
+which contains a human readable contextualized version of the error. For
+example:
+
+```
+2| key1 = "value1"
+3| key2 = "missing2"
+ | ~~~~ missing field
+4| key3 = "missing3"
+5| key4 = "value4"
+```
+
+[decode-err]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#DecodeError
+
+### Local date and time support
+
+TOML supports native [local date/times][ldt]. It allows to represent a given
+date, time, or date-time without relation to a timezone or offset. To support
+this use-case, go-toml provides [`LocalDate`][tld], [`LocalTime`][tlt], and
+[`LocalDateTime`][tldt]. Those types can be transformed to and from `time.Time`,
+making them convenient yet unambiguous structures for their respective TOML
+representation.
+
+[ldt]: https://toml.io/en/v1.0.0#local-date-time
+[tld]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalDate
+[tlt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalTime
+[tldt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalDateTime
+
+## Getting started
+
+Given the following struct, let's see how to read it and write it as TOML:
+
+```go
+type MyConfig struct {
+      Version int
+      Name    string
+      Tags    []string
+}
+```
+
+### Unmarshaling
+
+[`Unmarshal`][unmarshal] reads a TOML document and fills a Go structure with its
+content. For example:
+
+```go
+doc := `
+version = 2
+name = "go-toml"
+tags = ["go", "toml"]
+`
+
+var cfg MyConfig
+err := toml.Unmarshal([]byte(doc), &cfg)
+if err != nil {
+      panic(err)
+}
+fmt.Println("version:", cfg.Version)
+fmt.Println("name:", cfg.Name)
+fmt.Println("tags:", cfg.Tags)
+
+// Output:
+// version: 2
+// name: go-toml
+// tags: [go toml]
+```
+
+[unmarshal]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Unmarshal
+
+### Marshaling
+
+[`Marshal`][marshal] is the opposite of Unmarshal: it represents a Go structure
+as a TOML document:
+
+```go
+cfg := MyConfig{
+      Version: 2,
+      Name:    "go-toml",
+      Tags:    []string{"go", "toml"},
+}
+
+b, err := toml.Marshal(cfg)
+if err != nil {
+      panic(err)
+}
+fmt.Println(string(b))
+
+// Output:
+// Version = 2
+// Name = 'go-toml'
+// Tags = ['go', 'toml']
+```
+
+[marshal]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Marshal
+
+## Benchmarks
+
+Execution time speedup compared to other Go TOML libraries:
+
+<table>
+    <thead>
+        <tr><th>Benchmark</th><th>go-toml v1</th><th>BurntSushi/toml</th></tr>
+    </thead>
+    <tbody>
+        <tr><td>Marshal/HugoFrontMatter-2</td><td>1.9x</td><td>1.9x</td></tr>
+        <tr><td>Marshal/ReferenceFile/map-2</td><td>1.7x</td><td>1.9x</td></tr>
+        <tr><td>Marshal/ReferenceFile/struct-2</td><td>2.4x</td><td>2.6x</td></tr>
+        <tr><td>Unmarshal/HugoFrontMatter-2</td><td>2.9x</td><td>2.5x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/map-2</td><td>2.7x</td><td>2.6x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/struct-2</td><td>4.8x</td><td>5.1x</td></tr>
+     </tbody>
+</table>
+<details><summary>See more</summary>
+<p>The table above has the results of the most common use-cases. The table below
+contains the results of all benchmarks, including unrealistic ones. It is
+provided for completeness.</p>
+
+<table>
+    <thead>
+        <tr><th>Benchmark</th><th>go-toml v1</th><th>BurntSushi/toml</th></tr>
+    </thead>
+    <tbody>
+        <tr><td>Marshal/SimpleDocument/map-2</td><td>1.7x</td><td>2.1x</td></tr>
+        <tr><td>Marshal/SimpleDocument/struct-2</td><td>2.5x</td><td>2.8x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/map-2</td><td>4.1x</td><td>3.1x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/struct-2</td><td>6.4x</td><td>4.3x</td></tr>
+        <tr><td>UnmarshalDataset/example-2</td><td>3.4x</td><td>3.2x</td></tr>
+        <tr><td>UnmarshalDataset/code-2</td><td>2.2x</td><td>2.5x</td></tr>
+        <tr><td>UnmarshalDataset/twitter-2</td><td>2.8x</td><td>2.7x</td></tr>
+        <tr><td>UnmarshalDataset/citm_catalog-2</td><td>2.2x</td><td>2.0x</td></tr>
+        <tr><td>UnmarshalDataset/canada-2</td><td>1.8x</td><td>1.4x</td></tr>
+        <tr><td>UnmarshalDataset/config-2</td><td>4.4x</td><td>2.9x</td></tr>
+        <tr><td>[Geo mean]</td><td>2.8x</td><td>2.6x</td></tr>
+     </tbody>
+</table>
+<p>This table can be generated with <code>./ci.sh benchmark -a -html</code>.</p>
+</details>
+
+## Modules
+
+go-toml uses Go's standard modules system.
+
+Installation instructions:
+
+- Go ≥ 1.16: Nothing to do. Use the import in your code. The `go` command deals
+  with it automatically.
+- Go ≥ 1.13: `GO111MODULE=on go get github.com/pelletier/go-toml/v2`.
+
+In case of trouble: [Go Modules FAQ][mod-faq].
+
+[mod-faq]: https://github.com/golang/go/wiki/Modules#why-does-installing-a-tool-via-go-get-fail-with-error-cannot-find-main-module
+
+## Tools
+
+Go-toml provides three handy command line tools:
+
+ * `tomljson`: Reads a TOML file and outputs its JSON representation.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/tomljson@latest
+    $ tomljson --help
+    ```
+
+ * `jsontoml`: Reads a JSON file and outputs a TOML representation.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/jsontoml@latest
+    $ jsontoml --help
+    ```
+
+ * `tomll`: Lints and reformats a TOML file.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/tomll@latest
+    $ tomll --help
+    ```
+
+### Docker image
+
+Those tools are also available as a [Docker image][docker]. For example, to use
+`tomljson`:
+
+```
+docker run -i ghcr.io/pelletier/go-toml:v2 tomljson < example.toml
+```
+
+Multiple versions are availble on [ghcr.io][docker].
+
+[docker]: https://github.com/pelletier/go-toml/pkgs/container/go-toml
+
+## Migrating from v1
+
+This section describes the differences between v1 and v2, with some pointers on
+how to get the original behavior when possible.
+
+### Decoding / Unmarshal
+
+#### Automatic field name guessing
+
+When unmarshaling to a struct, if a key in the TOML document does not exactly
+match the name of a struct field or any of the `toml`-tagged field, v1 tries
+multiple variations of the key ([code][v1-keys]).
+
+V2 instead does a case-insensitive matching, like `encoding/json`.
+
+This could impact you if you are relying on casing to differentiate two fields,
+and one of them is a not using the `toml` struct tag. The recommended solution
+is to be specific about tag names for those fields using the `toml` struct tag.
+
+[v1-keys]: https://github.com/pelletier/go-toml/blob/a2e52561804c6cd9392ebf0048ca64fe4af67a43/marshal.go#L775-L781
+
+#### Ignore preexisting value in interface
+
+When decoding into a non-nil `interface{}`, go-toml v1 uses the type of the
+element in the interface to decode the object. For example:
+
+```go
+type inner struct {
+  B interface{}
+}
+type doc struct {
+  A interface{}
+}
+
+d := doc{
+  A: inner{
+    B: "Before",
+  },
+}
+
+data := `
+[A]
+B = "After"
+`
+
+toml.Unmarshal([]byte(data), &d)
+fmt.Printf("toml v1: %#v\n", d)
+
+// toml v1: main.doc{A:main.inner{B:"After"}}
+```
+
+In this case, field `A` is of type `interface{}`, containing a `inner` struct.
+V1 sees that type and uses it when decoding the object.
+
+When decoding an object into an `interface{}`, V2 instead disregards whatever
+value the `interface{}` may contain and replaces it with a
+`map[string]interface{}`. With the same data structure as above, here is what
+the result looks like:
+
+```go
+toml.Unmarshal([]byte(data), &d)
+fmt.Printf("toml v2: %#v\n", d)
+
+// toml v2: main.doc{A:map[string]interface {}{"B":"After"}}
+```
+
+This is to match `encoding/json`'s behavior. There is no way to make the v2
+decoder behave like v1.
+
+#### Values out of array bounds ignored
+
+When decoding into an array, v1 returns an error when the number of elements
+contained in the doc is superior to the capacity of the array. For example:
+
+```go
+type doc struct {
+  A [2]string
+}
+d := doc{}
+err := toml.Unmarshal([]byte(`A = ["one", "two", "many"]`), &d)
+fmt.Println(err)
+
+// (1, 1): unmarshal: TOML array length (3) exceeds destination array length (2)
+```
+
+In the same situation, v2 ignores the last value:
+
+```go
+err := toml.Unmarshal([]byte(`A = ["one", "two", "many"]`), &d)
+fmt.Println("err:", err, "d:", d)
+// err: <nil> d: {[one two]}
+```
+
+This is to match `encoding/json`'s behavior. There is no way to make the v2
+decoder behave like v1.
+
+#### Support for `toml.Unmarshaler` has been dropped
+
+This method was not widely used, poorly defined, and added a lot of complexity.
+A similar effect can be achieved by implementing the `encoding.TextUnmarshaler`
+interface and use strings.
+
+#### Support for `default` struct tag has been dropped
+
+This feature adds complexity and a poorly defined API for an effect that can be
+accomplished outside of the library.
+
+It does not seem like other format parsers in Go support that feature (the
+project referenced in the original ticket #202 has not been updated since 2017).
+Given that go-toml v2 should not touch values not in the document, the same
+effect can be achieved by pre-filling the struct with defaults (libraries like
+[go-defaults][go-defaults] can help). Also, string representation is not well
+defined for all types: it creates issues like #278.
+
+The recommended replacement is pre-filling the struct before unmarshaling.
+
+[go-defaults]: https://github.com/mcuadros/go-defaults
+
+#### `toml.Tree` replacement
+
+This structure was the initial attempt at providing a document model for
+go-toml. It allows manipulating the structure of any document, encoding and
+decoding from their TOML representation. While a more robust feature was
+initially planned in go-toml v2, this has been ultimately [removed from
+scope][nodoc] of this library, with no plan to add it back at the moment. The
+closest equivalent at the moment would be to unmarshal into an `interface{}` and
+use type assertions and/or reflection to manipulate the arbitrary
+structure. However this would fall short of providing all of the TOML features
+such as adding comments and be specific about whitespace.
+
+
+#### `toml.Position` are not retrievable anymore
+
+The API for retrieving the position (line, column) of a specific TOML element do
+not exist anymore. This was done to minimize the amount of concepts introduced
+by the library (query path), and avoid the performance hit related to storing
+positions in the absence of a document model, for a feature that seemed to have
+little use. Errors however have gained more detailed position
+information. Position retrieval seems better fitted for a document model, which
+has been [removed from the scope][nodoc] of go-toml v2 at the moment.
+
+### Encoding / Marshal
+
+#### Default struct fields order
+
+V1 emits struct fields order alphabetically by default. V2 struct fields are
+emitted in order they are defined. For example:
+
+```go
+type S struct {
+	B string
+	A string
+}
+
+data := S{
+	B: "B",
+	A: "A",
+}
+
+b, _ := tomlv1.Marshal(data)
+fmt.Println("v1:\n" + string(b))
+
+b, _ = tomlv2.Marshal(data)
+fmt.Println("v2:\n" + string(b))
+
+// Output:
+// v1:
+// A = "A"
+// B = "B"
+
+// v2:
+// B = 'B'
+// A = 'A'
+```
+
+There is no way to make v2 encoder behave like v1. A workaround could be to
+manually sort the fields alphabetically in the struct definition, or generate
+struct types using `reflect.StructOf`.
+
+#### No indentation by default
+
+V1 automatically indents content of tables by default. V2 does not. However the
+same behavior can be obtained using [`Encoder.SetIndentTables`][sit]. For example:
+
+```go
+data := map[string]interface{}{
+	"table": map[string]string{
+		"key": "value",
+	},
+}
+
+b, _ := tomlv1.Marshal(data)
+fmt.Println("v1:\n" + string(b))
+
+b, _ = tomlv2.Marshal(data)
+fmt.Println("v2:\n" + string(b))
+
+buf := bytes.Buffer{}
+enc := tomlv2.NewEncoder(&buf)
+enc.SetIndentTables(true)
+enc.Encode(data)
+fmt.Println("v2 Encoder:\n" + string(buf.Bytes()))
+
+// Output:
+// v1:
+//
+// [table]
+//   key = "value"
+//
+// v2:
+// [table]
+// key = 'value'
+//
+//
+// v2 Encoder:
+// [table]
+//   key = 'value'
+```
+
+[sit]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Encoder.SetIndentTables
+
+#### Keys and strings are single quoted
+
+V1 always uses double quotes (`"`) around strings and keys that cannot be
+represented bare (unquoted). V2 uses single quotes instead by default (`'`),
+unless a character cannot be represented, then falls back to double quotes. As a
+result of this change, `Encoder.QuoteMapKeys` has been removed, as it is not
+useful anymore.
+
+There is no way to make v2 encoder behave like v1.
+
+#### `TextMarshaler` emits as a string, not TOML
+
+Types that implement [`encoding.TextMarshaler`][tm] can emit arbitrary TOML in
+v1. The encoder would append the result to the output directly. In v2 the result
+is wrapped in a string. As a result, this interface cannot be implemented by the
+root object.
+
+There is no way to make v2 encoder behave like v1.
+
+[tm]: https://golang.org/pkg/encoding/#TextMarshaler
+
+#### `Encoder.CompactComments` has been removed
+
+Emitting compact comments is now the default behavior of go-toml. This option
+is not necessary anymore.
+
+#### Struct tags have been merged
+
+V1 used to provide multiple struct tags: `comment`, `commented`, `multiline`,
+`toml`, and `omitempty`. To behave more like the standard library, v2 has merged
+`toml`, `multiline`, and `omitempty`. For example:
+
+```go
+type doc struct {
+	// v1
+	F string `toml:"field" multiline:"true" omitempty:"true"`
+	// v2
+	F string `toml:"field,multiline,omitempty"`
+}
+```
+
+Has a result, the `Encoder.SetTag*` methods have been removed, as there is just
+one tag now.
+
+
+#### `commented` tag has been removed
+
+There is no replacement for the `commented` tag. This feature would be better
+suited in a proper document model for go-toml v2, which has been [cut from
+scope][nodoc] at the moment.
+
+#### `Encoder.ArraysWithOneElementPerLine` has been renamed
+
+The new name is `Encoder.SetArraysMultiline`. The behavior should be the same.
+
+#### `Encoder.Indentation` has been renamed
+
+The new name is `Encoder.SetIndentSymbol`. The behavior should be the same.
+
+
+#### Embedded structs behave like stdlib
+
+V1 defaults to merging embedded struct fields into the embedding struct. This
+behavior was unexpected because it does not follow the standard library. To
+avoid breaking backward compatibility, the `Encoder.PromoteAnonymous` method was
+added to make the encoder behave correctly. Given backward compatibility is not
+a problem anymore, v2 does the right thing by default: it follows the behavior
+of `encoding/json`. `Encoder.PromoteAnonymous` has been removed.
+
+[nodoc]: https://github.com/pelletier/go-toml/discussions/506#discussioncomment-1526038
+
+### `query`
+
+go-toml v1 provided the [`go-toml/query`][query] package. It allowed to run
+JSONPath-style queries on TOML files. This feature is not available in v2. For a
+replacement, check out [dasel][dasel].
+
+This package has been removed because it was essentially not supported anymore
+(last commit May 2020), increased the complexity of the code base, and more
+complete solutions exist out there.
+
+[query]: https://github.com/pelletier/go-toml/tree/f99d6bbca119636aeafcf351ee52b3d202782627/query
+[dasel]: https://github.com/TomWright/dasel
 
 ## License
 

SECURITY.md

@@ -0,0 +1,19 @@
+# Security Policy
+
+## Supported Versions
+
+Use this section to tell people about which versions of your project are
+currently being supported with security updates.
+
+| Version    | Supported          |
+| ---------- | ------------------ |
+| Latest 2.x | :white_check_mark: |
+| All 1.x    | :x:                |
+| All 0.x    | :x:                |
+
+## Reporting a Vulnerability
+
+Email a vulnerability report to `security@pelletier.codes`. Make sure to include
+as many details as possible to reproduce the vulnerability. This is a
+side-project: I will try to get back to you as quickly as possible, time
+permitting in my personal life. Providing a working patch helps very much!

benchmark/bench_datasets_test.go

@@ -8,6 +8,7 @@ import (
 	"path/filepath"
 	"testing"
 
+	"github.com/pelletier/go-toml/v2"
 	"github.com/stretchr/testify/require"
 )
 
@@ -30,65 +31,50 @@ var bench_inputs = []struct {
 
 func TestUnmarshalDatasetCode(t *testing.T) {
 	for _, tc := range bench_inputs {
-		buf := fixture(t, tc.name)
 		t.Run(tc.name, func(t *testing.T) {
-			for _, r := range runners {
-				if r.name == "bs" && tc.name == "canada" {
-					t.Skip("skipping: burntsushi can't handle mixed arrays")
-				}
+			buf := fixture(t, tc.name)
 
-				t.Run(r.name, func(t *testing.T) {
-					var v interface{}
-					check(t, r.unmarshal(buf, &v))
+			var v interface{}
+			require.NoError(t, toml.Unmarshal(buf, &v))
 
-					b, err := json.Marshal(v)
-					check(t, err)
-					require.Equal(t, len(b), tc.jsonLen)
-				})
-			}
+			b, err := json.Marshal(v)
+			require.NoError(t, err)
+			require.Equal(t, len(b), tc.jsonLen)
 		})
 	}
 }
 
 func BenchmarkUnmarshalDataset(b *testing.B) {
 	for _, tc := range bench_inputs {
-		buf := fixture(b, tc.name)
 		b.Run(tc.name, func(b *testing.B) {
-			bench(b, func(r runner, b *testing.B) {
-				if r.name == "bs" && tc.name == "canada" {
-					b.Skip("skipping: burntsushi can't handle mixed arrays")
-				}
-
-				b.SetBytes(int64(len(buf)))
-				b.ReportAllocs()
-				b.ResetTimer()
-				for i := 0; i < b.N; i++ {
-					var v interface{}
-					check(b, r.unmarshal(buf, &v))
-				}
-			})
+			buf := fixture(b, tc.name)
+			b.SetBytes(int64(len(buf)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				var v interface{}
+				require.NoError(b, toml.Unmarshal(buf, &v))
+			}
 		})
 	}
 }
 
 // fixture returns the uncompressed contents of path.
 func fixture(tb testing.TB, path string) []byte {
-	f, err := os.Open(filepath.Join("testdata", path+".toml.gz"))
-	check(tb, err)
+	tb.Helper()
+
+	file := path + ".toml.gz"
+	f, err := os.Open(filepath.Join("testdata", file))
+	if os.IsNotExist(err) {
+		tb.Skip("benchmark fixture not found:", file)
+	}
+	require.NoError(tb, err)
 	defer f.Close()
 
 	gz, err := gzip.NewReader(f)
-	check(tb, err)
+	require.NoError(tb, err)
 
 	buf, err := ioutil.ReadAll(gz)
-	check(tb, err)
-
+	require.NoError(tb, err)
 	return buf
 }
-
-func check(tb testing.TB, err error) {
-	if err != nil {
-		tb.Helper()
-		tb.Fatal(err)
-	}
-}

benchmark/benchmark.toml

@@ -186,7 +186,7 @@ key3 = 1979-05-27T00:32:00.999999-07:00
 key1 = [ 1, 2, 3 ]
 key2 = [ "red", "yellow", "green" ]
 key3 = [ [ 1, 2 ], [3, 4, 5] ]
-#key4 = [ [ 1, 2 ], ["a", "b", "c"] ] # this is ok
+key4 = [ [ 1, 2 ], ["a", "b", "c"] ] # this is ok
 
 # Arrays can also be multiline. So in addition to ignoring whitespace, arrays
 # also ignore newlines between the brackets.  Terminating commas are ok before

benchmark/benchmark_test.go

@@ -1,43 +1,103 @@
 package benchmark_test
 
 import (
+	"bytes"
 	"io/ioutil"
 	"testing"
 	"time"
 
-	tomlbs "github.com/BurntSushi/toml"
-	tomlv1 "github.com/pelletier/go-toml-v1"
 	"github.com/pelletier/go-toml/v2"
 	"github.com/stretchr/testify/require"
 )
 
-type runner struct {
-	name      string
-	unmarshal func([]byte, interface{}) error
-}
+func TestUnmarshalSimple(t *testing.T) {
+	doc := []byte(`A = "hello"`)
+	d := struct {
+		A string
+	}{}
 
-var runners = []runner{
-	{"v2", toml.Unmarshal},
-	{"v1", tomlv1.Unmarshal},
-	{"bs", tomlbs.Unmarshal},
-}
-
-func bench(b *testing.B, f func(r runner, b *testing.B)) {
-	for _, r := range runners {
-		b.Run(r.name, func(b *testing.B) {
-			f(r, b)
-		})
+	err := toml.Unmarshal(doc, &d)
+	if err != nil {
+		panic(err)
 	}
 }
 
-func BenchmarkUnmarshalSimple(b *testing.B) {
-	bench(b, func(r runner, b *testing.B) {
-		d := struct {
-			A string
-		}{}
+func BenchmarkUnmarshal(b *testing.B) {
+	b.Run("SimpleDocument", func(b *testing.B) {
 		doc := []byte(`A = "hello"`)
+
+		b.Run("struct", func(b *testing.B) {
+			b.SetBytes(int64(len(doc)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				d := struct {
+					A string
+				}{}
+
+				err := toml.Unmarshal(doc, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+
+		b.Run("map", func(b *testing.B) {
+			b.SetBytes(int64(len(doc)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				d := map[string]interface{}{}
+				err := toml.Unmarshal(doc, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+	})
+
+	b.Run("ReferenceFile", func(b *testing.B) {
+		bytes, err := ioutil.ReadFile("benchmark.toml")
+		if err != nil {
+			b.Fatal(err)
+		}
+
+		b.Run("struct", func(b *testing.B) {
+			b.SetBytes(int64(len(bytes)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				d := benchmarkDoc{}
+				err := toml.Unmarshal(bytes, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+
+		b.Run("map", func(b *testing.B) {
+			b.SetBytes(int64(len(bytes)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				d := map[string]interface{}{}
+				err := toml.Unmarshal(bytes, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+	})
+
+	b.Run("HugoFrontMatter", func(b *testing.B) {
+		b.SetBytes(int64(len(hugoFrontMatterbytes)))
+		b.ReportAllocs()
+		b.ResetTimer()
 		for i := 0; i < b.N; i++ {
-			err := r.unmarshal(doc, &d)
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(hugoFrontMatterbytes, &d)
 			if err != nil {
 				panic(err)
 			}
@@ -45,6 +105,137 @@ func BenchmarkUnmarshalSimple(b *testing.B) {
 	})
 }
 
+func marshal(v interface{}) ([]byte, error) {
+	var b bytes.Buffer
+	enc := toml.NewEncoder(&b)
+	err := enc.Encode(v)
+	return b.Bytes(), err
+}
+
+func BenchmarkMarshal(b *testing.B) {
+	b.Run("SimpleDocument", func(b *testing.B) {
+		doc := []byte(`A = "hello"`)
+
+		b.Run("struct", func(b *testing.B) {
+			d := struct {
+				A string
+			}{}
+
+			err := toml.Unmarshal(doc, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+
+		b.Run("map", func(b *testing.B) {
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(doc, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+	})
+
+	b.Run("ReferenceFile", func(b *testing.B) {
+		bytes, err := ioutil.ReadFile("benchmark.toml")
+		if err != nil {
+			b.Fatal(err)
+		}
+
+		b.Run("struct", func(b *testing.B) {
+			d := benchmarkDoc{}
+			err := toml.Unmarshal(bytes, &d)
+			if err != nil {
+				panic(err)
+			}
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+
+		b.Run("map", func(b *testing.B) {
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(bytes, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+	})
+
+	b.Run("HugoFrontMatter", func(b *testing.B) {
+		d := map[string]interface{}{}
+		err := toml.Unmarshal(hugoFrontMatterbytes, &d)
+		if err != nil {
+			panic(err)
+		}
+
+		b.ReportAllocs()
+		b.ResetTimer()
+
+		var out []byte
+
+		for i := 0; i < b.N; i++ {
+			out, err = marshal(d)
+			if err != nil {
+				panic(err)
+			}
+		}
+
+		b.SetBytes(int64(len(out)))
+	})
+}
+
 type benchmarkDoc struct {
 	Table struct {
 		Key      string
@@ -58,7 +249,7 @@ type benchmarkDoc struct {
 			}
 			Point struct {
 				X int64
-				U int64
+				Y int64
 			}
 		}
 	}
@@ -131,6 +322,7 @@ type benchmarkDoc struct {
 		Key2 []string
 		Key3 [][]int64
 		// TODO: Key4 not supported by go-toml's Unmarshal
+		Key4 []interface{}
 		Key5 []int64
 		Key6 []int64
 	}
@@ -142,38 +334,321 @@ type benchmarkDoc struct {
 	Fruit []struct {
 		Name     string
 		Physical struct {
-			Color   string
-			Shape   string
-			Variety []struct {
-				Name string
-			}
+			Color string
+			Shape string
+		}
+		Variety []struct {
+			Name string
 		}
 	}
 }
 
-func BenchmarkReferenceFile(b *testing.B) {
-	bench(b, func(r runner, b *testing.B) {
-		bytes, err := ioutil.ReadFile("benchmark.toml")
-		if err != nil {
-			b.Fatal(err)
-		}
-		b.SetBytes(int64(len(bytes)))
-		b.ReportAllocs()
-		b.ResetTimer()
-		for i := 0; i < b.N; i++ {
-			d := benchmarkDoc{}
-			err := r.unmarshal(bytes, &d)
-			if err != nil {
-				panic(err)
-			}
-		}
-	})
-}
-
-func TestReferenceFile(t *testing.T) {
+func TestUnmarshalReferenceFile(t *testing.T) {
 	bytes, err := ioutil.ReadFile("benchmark.toml")
 	require.NoError(t, err)
 	d := benchmarkDoc{}
 	err = toml.Unmarshal(bytes, &d)
 	require.NoError(t, err)
+
+	expected := benchmarkDoc{
+		Table: struct {
+			Key      string
+			Subtable struct{ Key string }
+			Inline   struct {
+				Name struct {
+					First string
+					Last  string
+				}
+				Point struct {
+					X int64
+					Y int64
+				}
+			}
+		}{
+			Key: "value",
+			Subtable: struct{ Key string }{
+				Key: "another value",
+			},
+			// note: x.y.z.w is purposefully missing
+			Inline: struct {
+				Name struct {
+					First string
+					Last  string
+				}
+				Point struct {
+					X int64
+					Y int64
+				}
+			}{
+				Name: struct {
+					First string
+					Last  string
+				}{
+					First: "Tom",
+					Last:  "Preston-Werner",
+				},
+				Point: struct {
+					X int64
+					Y int64
+				}{
+					X: 1,
+					Y: 2,
+				},
+			},
+		},
+		String: struct {
+			Basic     struct{ Basic string }
+			Multiline struct {
+				Key1      string
+				Key2      string
+				Key3      string
+				Continued struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}
+			}
+			Literal struct {
+				Winpath   string
+				Winpath2  string
+				Quoted    string
+				Regex     string
+				Multiline struct {
+					Regex2 string
+					Lines  string
+				}
+			}
+		}{
+			Basic: struct{ Basic string }{
+				Basic: "I'm a string. \"You can quote me\". Name\tJos\u00E9\nLocation\tSF.",
+			},
+			Multiline: struct {
+				Key1      string
+				Key2      string
+				Key3      string
+				Continued struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}
+			}{
+				Key1: "One\nTwo",
+				Key2: "One\nTwo",
+				Key3: "One\nTwo",
+
+				Continued: struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}{
+					Key1: `The quick brown fox jumps over the lazy dog.`,
+					Key2: `The quick brown fox jumps over the lazy dog.`,
+					Key3: `The quick brown fox jumps over the lazy dog.`,
+				},
+			},
+			Literal: struct {
+				Winpath   string
+				Winpath2  string
+				Quoted    string
+				Regex     string
+				Multiline struct {
+					Regex2 string
+					Lines  string
+				}
+			}{
+				Winpath:  `C:\Users\nodejs\templates`,
+				Winpath2: `\\ServerX\admin$\system32\`,
+				Quoted:   `Tom "Dubs" Preston-Werner`,
+				Regex:    `<\i\c*\s*>`,
+
+				Multiline: struct {
+					Regex2 string
+					Lines  string
+				}{
+					Regex2: `I [dw]on't need \d{2} apples`,
+					Lines: `The first newline is
+trimmed in raw strings.
+   All other whitespace
+   is preserved.
+`,
+				},
+			},
+		},
+		Integer: struct {
+			Key1        int64
+			Key2        int64
+			Key3        int64
+			Key4        int64
+			Underscores struct {
+				Key1 int64
+				Key2 int64
+				Key3 int64
+			}
+		}{
+			Key1: 99,
+			Key2: 42,
+			Key3: 0,
+			Key4: -17,
+
+			Underscores: struct {
+				Key1 int64
+				Key2 int64
+				Key3 int64
+			}{
+				Key1: 1000,
+				Key2: 5349221,
+				Key3: 12345,
+			},
+		},
+		Float: struct {
+			Fractional struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}
+			Exponent struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}
+			Both        struct{ Key float64 }
+			Underscores struct {
+				Key1 float64
+				Key2 float64
+			}
+		}{
+			Fractional: struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}{
+				Key1: 1.0,
+				Key2: 3.1415,
+				Key3: -0.01,
+			},
+			Exponent: struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}{
+				Key1: 5e+22,
+				Key2: 1e6,
+				Key3: -2e-2,
+			},
+			Both: struct{ Key float64 }{
+				Key: 6.626e-34,
+			},
+			Underscores: struct {
+				Key1 float64
+				Key2 float64
+			}{
+				Key1: 9224617.445991228313,
+				Key2: 1e100,
+			},
+		},
+		Boolean: struct {
+			True  bool
+			False bool
+		}{
+			True:  true,
+			False: false,
+		},
+		Datetime: struct {
+			Key1 time.Time
+			Key2 time.Time
+			Key3 time.Time
+		}{
+			Key1: time.Date(1979, 5, 27, 7, 32, 0, 0, time.UTC),
+			Key2: time.Date(1979, 5, 27, 0, 32, 0, 0, time.FixedZone("", -7*3600)),
+			Key3: time.Date(1979, 5, 27, 0, 32, 0, 999999000, time.FixedZone("", -7*3600)),
+		},
+		Array: struct {
+			Key1 []int64
+			Key2 []string
+			Key3 [][]int64
+			Key4 []interface{}
+			Key5 []int64
+			Key6 []int64
+		}{
+			Key1: []int64{1, 2, 3},
+			Key2: []string{"red", "yellow", "green"},
+			Key3: [][]int64{{1, 2}, {3, 4, 5}},
+			Key4: []interface{}{
+				[]interface{}{int64(1), int64(2)},
+				[]interface{}{"a", "b", "c"},
+			},
+			Key5: []int64{1, 2, 3},
+			Key6: []int64{1, 2},
+		},
+		Products: []struct {
+			Name  string
+			Sku   int64
+			Color string
+		}{
+			{
+				Name: "Hammer",
+				Sku:  738594937,
+			},
+			{},
+			{
+				Name:  "Nail",
+				Sku:   284758393,
+				Color: "gray",
+			},
+		},
+		Fruit: []struct {
+			Name     string
+			Physical struct {
+				Color string
+				Shape string
+			}
+			Variety []struct{ Name string }
+		}{
+			{
+				Name: "apple",
+				Physical: struct {
+					Color string
+					Shape string
+				}{
+					Color: "red",
+					Shape: "round",
+				},
+				Variety: []struct{ Name string }{
+					{Name: "red delicious"},
+					{Name: "granny smith"},
+				},
+			},
+			{
+				Name: "banana",
+				Variety: []struct{ Name string }{
+					{Name: "plantain"},
+				},
+			},
+		},
+	}
+
+	require.Equal(t, expected, d)
 }
+
+var hugoFrontMatterbytes = []byte(`
+categories = ["Development", "VIM"]
+date = "2012-04-06"
+description = "spf13-vim is a cross platform distribution of vim plugins and resources for Vim."
+slug = "spf13-vim-3-0-release-and-new-website"
+tags = [".vimrc", "plugins", "spf13-vim", "vim"]
+title = "spf13-vim 3.0 release and new website"
+include_toc = true
+show_comments = false
+
+[[cascade]]
+  background = "yosemite.jpg"
+  [cascade._target]
+    kind = "page"
+    lang = "en"
+    path = "/blog/**"
+
+[[cascade]]
+  background = "goldenbridge.jpg"
+  [cascade._target]
+    kind = "section"
+`)

benchmark/go.mod

@@ -1,14 +0,0 @@
-module github.com/pelletier/go-toml/v2/benchmark
-
-go 1.16
-
-replace github.com/pelletier/go-toml/v2 => ../
-
-replace github.com/pelletier/go-toml-v1 => github.com/pelletier/go-toml v1.8.1
-
-require (
-	github.com/BurntSushi/toml v0.3.1
-	github.com/pelletier/go-toml-v1 v0.0.0-00010101000000-000000000000
-	github.com/pelletier/go-toml/v2 v2.0.0-00010101000000-000000000000
-	github.com/stretchr/testify v1.7.0
-)

benchmark/go.sum

@@ -1,16 +0,0 @@
-github.com/BurntSushi/toml v0.3.1 h1:WXkYYl6Yr3qBf1K79EBnL4mak0OimBfB0XUf9Vl28OQ=
-github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
-github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
-github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/pelletier/go-toml v1.8.1 h1:1Nf83orprkJyknT6h7zbuEGUEjcyVlCxSUGTENmNCRM=
-github.com/pelletier/go-toml v1.8.1/go.mod h1:T2/BmBdy8dvIRq1a/8aqjN41wvWlN4lrapLU/GW4pbc=
-github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
-github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/testify v1.7.0 h1:nwc3DEeHmmLAfoZucVR881uASk0Mfjw8xYJ99tb5CcY=
-github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
-gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c h1:dUUwHk2QECo/6vqA44rthZ8ie2QXMNeKRTHCNY2nXvo=
-gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

benchmark_test.go

@@ -0,0 +1,71 @@
+package toml
+
+import (
+	"bytes"
+	"testing"
+)
+
+var valid10Ascii = []byte("1234567890")
+var valid10Utf8 = []byte("日本語a")
+var valid1kUtf8 = bytes.Repeat([]byte("0123456789日本語日本語日本語日abcdefghijklmnopqrstuvwx"), 16)
+var valid1MUtf8 = bytes.Repeat(valid1kUtf8, 1024)
+var valid1kAscii = bytes.Repeat([]byte("012345678998jhjklasDJKLAAdjdfjsdklfjdslkabcdefghijklmnopqrstuvwx"), 16)
+var valid1MAscii = bytes.Repeat(valid1kAscii, 1024)
+
+func BenchmarkScanComments(b *testing.B) {
+	wrap := func(x []byte) []byte {
+		return []byte("# " + string(x) + "\n")
+	}
+
+	inputs := map[string][]byte{
+		"10Valid":     wrap(valid10Ascii),
+		"1kValid":     wrap(valid1kAscii),
+		"1MValid":     wrap(valid1MAscii),
+		"10ValidUtf8": wrap(valid10Utf8),
+		"1kValidUtf8": wrap(valid1kUtf8),
+		"1MValidUtf8": wrap(valid1MUtf8),
+	}
+
+	for name, input := range inputs {
+		b.Run(name, func(b *testing.B) {
+			b.SetBytes(int64(len(input)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				scanComment(input)
+			}
+		})
+	}
+}
+
+func BenchmarkParseLiteralStringValid(b *testing.B) {
+	wrap := func(x []byte) []byte {
+		return []byte("'" + string(x) + "'")
+	}
+
+	inputs := map[string][]byte{
+		"10Valid":     wrap(valid10Ascii),
+		"1kValid":     wrap(valid1kAscii),
+		"1MValid":     wrap(valid1MAscii),
+		"10ValidUtf8": wrap(valid10Utf8),
+		"1kValidUtf8": wrap(valid1kUtf8),
+		"1MValidUtf8": wrap(valid1MUtf8),
+	}
+
+	for name, input := range inputs {
+		b.Run(name, func(b *testing.B) {
+			p := parser{}
+			b.SetBytes(int64(len(input)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				_, _, _, err := p.parseLiteralString(input)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+	}
+}

ci.sh

@@ -0,0 +1,271 @@
+#!/usr/bin/env bash
+
+
+stderr() {
+    echo "$@" 1>&2
+}
+
+usage() {
+    b=$(basename "$0")
+    echo $b: ERROR: "$@" 1>&2
+
+    cat 1>&2 <<EOF
+
+DESCRIPTION
+
+    $(basename "$0") is the script to run continuous integration commands for
+    go-toml on unix.
+
+    Requires Go and Git to be available in the PATH. Expects to be ran from the
+    root of go-toml's Git repository.
+
+USAGE
+
+    $b COMMAND [OPTIONS...]
+
+COMMANDS
+
+benchmark [OPTIONS...] [BRANCH]
+
+    Run benchmarks.
+
+    ARGUMENTS
+
+        BRANCH Optional. Defines which Git branch to use when running
+               benchmarks.
+
+    OPTIONS
+
+        -d      Compare benchmarks of HEAD with BRANCH using benchstats. In
+                this form the BRANCH argument is required.
+
+        -a      Compare benchmarks of HEAD against go-toml v1 and
+                BurntSushi/toml.
+
+        -html   When used with -a, emits the output as HTML, ready to be
+                embedded in the README.
+
+coverage [OPTIONS...] [BRANCH]
+
+    Generates code coverage.
+
+    ARGUMENTS
+
+        BRANCH  Optional. Defines which Git branch to use when reporting
+                coverage. Defaults to HEAD.
+
+    OPTIONS
+
+        -d      Compare coverage of HEAD with the one of BRANCH. In this form,
+                the BRANCH argument is required. Exit code is non-zero when
+                coverage percentage decreased.
+EOF
+    exit 1
+}
+
+cover() {
+    branch="${1}"
+    dir="$(mktemp -d)"
+
+    stderr "Executing coverage for ${branch} at ${dir}"
+
+    if [ "${branch}" = "HEAD" ]; then
+	    cp -r . "${dir}/"
+    else
+	    git worktree add "$dir" "$branch"
+    fi
+
+    pushd "$dir"
+    go test -covermode=atomic -coverprofile=coverage.out ./...
+    go tool cover -func=coverage.out
+    popd
+
+    if [ "${branch}" != "HEAD" ]; then
+	    git worktree remove --force "$dir"
+    fi
+}
+
+coverage() {
+    case "$1" in
+	-d)
+	    shift
+	    target="${1?Need to provide a target branch argument}"
+
+	    output_dir="$(mktemp -d)"
+	    target_out="${output_dir}/target.txt"
+	    head_out="${output_dir}/head.txt"
+	    
+	    cover "${target}" > "${target_out}"
+	    cover "HEAD" > "${head_out}"
+
+	    cat "${target_out}"
+	    cat "${head_out}"
+
+	    echo ""
+
+	    target_pct="$(cat ${target_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    head_pct="$(cat ${head_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    echo "Results: ${target} ${target_pct}% HEAD ${head_pct}%"
+
+	    delta_pct=$(echo "$head_pct - $target_pct" | bc -l)
+	    echo "Delta: ${delta_pct}"
+
+	    if [[ $delta_pct = \-* ]]; then
+		echo "Regression!";
+		return 1
+	    fi
+	    return 0
+	    ;;
+    esac
+
+    cover "${1-HEAD}"
+}
+
+bench() {
+    branch="${1}"
+    out="${2}"
+    replace="${3}"
+    dir="$(mktemp -d)"
+
+    stderr "Executing benchmark for ${branch} at ${dir}"
+
+    if [ "${branch}" = "HEAD" ]; then
+    	cp -r . "${dir}/"
+    else
+	    git worktree add "$dir" "$branch"
+    fi
+
+    pushd "$dir"
+
+    if [ "${replace}" != "" ]; then
+        find ./benchmark/ -iname '*.go' -exec sed -i -E "s|github.com/pelletier/go-toml/v2|${replace}|g" {} \;
+        go get "${replace}"
+    fi
+
+    export GOMAXPROCS=2
+    nice -n -19 taskset --cpu-list 0,1 go test '-bench=^Benchmark(Un)?[mM]arshal' -count=5 -run=Nothing ./... | tee "${out}"
+    popd
+
+    if [ "${branch}" != "HEAD" ]; then
+	    git worktree remove --force "$dir"
+    fi
+}
+
+fmktemp() {
+    if mktemp --version|grep GNU >/dev/null; then
+        mktemp --suffix=-$1;
+    else
+        mktemp -t $1;
+    fi
+}
+
+benchstathtml() {
+python3 - $1 <<'EOF'
+import sys
+
+lines = []
+stop = False
+
+with open(sys.argv[1]) as f:
+    for line in f.readlines():
+        line = line.strip()
+        if line == "":
+            stop = True
+        if not stop:
+            lines.append(line.split(','))
+
+results = []
+for line in reversed(lines[1:]):
+    v2 = float(line[1])
+    results.append([
+        line[0].replace("-32", ""),
+        "%.1fx" % (float(line[3])/v2),  # v1
+        "%.1fx" % (float(line[5])/v2),  # bs
+    ])
+# move geomean to the end
+results.append(results[0])
+del results[0]
+
+
+def printtable(data):
+    print("""
+<table>
+    <thead>
+        <tr><th>Benchmark</th><th>go-toml v1</th><th>BurntSushi/toml</th></tr>
+    </thead>
+    <tbody>""")
+
+    for r in data:
+        print("        <tr><td>{}</td><td>{}</td><td>{}</td></tr>".format(*r))
+
+    print("""     </tbody>
+</table>""")
+
+
+def match(x):
+    return "ReferenceFile" in x[0] or "HugoFrontMatter" in x[0]
+
+above = [x for x in results if match(x)]
+below = [x for x in results if not match(x)]
+
+printtable(above)
+print("<details><summary>See more</summary>")
+print("""<p>The table above has the results of the most common use-cases. The table below
+contains the results of all benchmarks, including unrealistic ones. It is
+provided for completeness.</p>""")
+printtable(below)
+print('<p>This table can be generated with <code>./ci.sh benchmark -a -html</code>.</p>')
+print("</details>")
+
+EOF
+}
+
+benchmark() {
+    case "$1" in
+    -d)
+        shift
+     	target="${1?Need to provide a target branch argument}"
+
+        old=`fmktemp ${target}`
+        bench "${target}" "${old}"
+
+        new=`fmktemp HEAD`
+        bench HEAD "${new}"
+
+        benchstat "${old}" "${new}"
+        return 0
+        ;;
+    -a)
+        shift
+
+        v2stats=`fmktemp go-toml-v2`
+        bench HEAD "${v2stats}" "github.com/pelletier/go-toml/v2"
+        v1stats=`fmktemp go-toml-v1`
+        bench HEAD "${v1stats}" "github.com/pelletier/go-toml"
+        bsstats=`fmktemp bs-toml`
+        bench HEAD "${bsstats}" "github.com/BurntSushi/toml"
+
+        cp "${v2stats}" go-toml-v2.txt
+        cp "${v1stats}" go-toml-v1.txt
+        cp "${bsstats}" bs-toml.txt
+
+        if [ "$1" = "-html" ]; then
+            tmpcsv=`fmktemp csv`
+            benchstat -csv -geomean go-toml-v2.txt go-toml-v1.txt bs-toml.txt > $tmpcsv
+            benchstathtml $tmpcsv
+        else
+            benchstat -geomean go-toml-v2.txt go-toml-v1.txt bs-toml.txt
+        fi
+
+        rm -f go-toml-v2.txt go-toml-v1.txt bs-toml.txt
+        return $?
+    esac
+
+    bench "${1-HEAD}" `mktemp`
+}
+
+case "$1" in
+    coverage) shift; coverage $@;;
+    benchmark) shift; benchmark $@;;
+    *) usage "bad argument $1";;
+esac

cmd/gotoml-test-decoder/main.go

@@ -0,0 +1,30 @@
+package main
+
+import (
+	"flag"
+	"log"
+	"os"
+	"path"
+
+	"github.com/pelletier/go-toml/v2/internal/testsuite"
+)
+
+func main() {
+	log.SetFlags(0)
+	flag.Usage = usage
+	flag.Parse()
+	if flag.NArg() != 0 {
+		flag.Usage()
+	}
+
+	err := testsuite.DecodeStdin()
+	if err != nil {
+		log.Fatal(err)
+	}
+}
+
+func usage() {
+	log.Printf("Usage: %s < toml-file\n", path.Base(os.Args[0]))
+	flag.PrintDefaults()
+	os.Exit(1)
+}

cmd/jsontoml/main.go

@@ -0,0 +1,55 @@
+// Package jsontoml is a program that converts JSON to TOML.
+//
+// Usage
+//
+// Reading from stdin:
+//
+//   cat file.json | jsontoml > file.toml
+//
+// Reading from a file:
+//
+//   jsontoml file.json > file.toml
+//
+// Installation
+//
+// Using Go:
+//
+//   go install github.com/pelletier/go-toml/v2/cmd/jsontoml@latest
+package main
+
+import (
+	"encoding/json"
+	"io"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/internal/cli"
+)
+
+const usage = `jsontoml can be used in two ways:
+Reading from stdin:
+  cat file.json | jsontoml > file.toml
+
+Reading from a file:
+  jsontoml file.json > file.toml
+`
+
+func main() {
+	p := cli.Program{
+		Usage: usage,
+		Fn:    convert,
+	}
+	p.Execute()
+}
+
+func convert(r io.Reader, w io.Writer) error {
+	var v interface{}
+
+	d := json.NewDecoder(r)
+	err := d.Decode(&v)
+	if err != nil {
+		return err
+	}
+
+	e := toml.NewEncoder(w)
+	return e.Encode(v)
+}

cmd/jsontoml/main_test.go

@@ -0,0 +1,49 @@
+package main
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConvert(t *testing.T) {
+	examples := []struct {
+		name     string
+		input    string
+		expected string
+		errors   bool
+	}{
+		{
+			name: "valid json",
+			input: `
+{
+  "mytoml": {
+    "a": 42
+  }
+}`,
+			expected: `[mytoml]
+a = 42.0
+
+`,
+		},
+		{
+			name:   "invalid json",
+			input:  `{ foo`,
+			errors: true,
+		},
+	}
+
+	for _, e := range examples {
+		b := new(bytes.Buffer)
+		err := convert(strings.NewReader(e.input), b)
+		if e.errors {
+			require.Error(t, err)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, e.expected, b.String())
+		}
+	}
+}

cmd/tomljson/main.go

@@ -0,0 +1,63 @@
+// Package tomljson is a program that converts TOML to JSON.
+//
+// Usage
+//
+// Reading from stdin:
+//
+//   cat file.toml | tomljson > file.json
+//
+// Reading from a file:
+//
+//   tomljson file.toml > file.json
+//
+// Installation
+//
+// Using Go:
+//
+//   go install github.com/pelletier/go-toml/v2/cmd/tomljson@latest
+package main
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/internal/cli"
+)
+
+const usage = `tomljson can be used in two ways:
+Reading from stdin:
+  cat file.toml | tomljson > file.json
+
+Reading from a file:
+  tomljson file.toml > file.json
+`
+
+func main() {
+	p := cli.Program{
+		Usage: usage,
+		Fn:    convert,
+	}
+	p.Execute()
+}
+
+func convert(r io.Reader, w io.Writer) error {
+	var v interface{}
+
+	d := toml.NewDecoder(r)
+	err := d.Decode(&v)
+	if err != nil {
+		var derr *toml.DecodeError
+		if errors.As(err, &derr) {
+			row, col := derr.Position()
+			return fmt.Errorf("%s\nerror occurred at row %d column %d", derr.String(), row, col)
+		}
+		return err
+	}
+
+	e := json.NewEncoder(w)
+	e.SetIndent("", "  ")
+	return e.Encode(v)
+}

cmd/tomljson/main_test.go

@@ -0,0 +1,61 @@
+package main
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConvert(t *testing.T) {
+	examples := []struct {
+		name     string
+		input    io.Reader
+		expected string
+		errors   bool
+	}{
+		{
+			name: "valid toml",
+			input: strings.NewReader(`
+[mytoml]
+a = 42`),
+			expected: `{
+  "mytoml": {
+    "a": 42
+  }
+}
+`,
+		},
+		{
+			name:   "invalid toml",
+			input:  strings.NewReader(`bad = []]`),
+			errors: true,
+		},
+		{
+			name:   "bad reader",
+			input:  &badReader{},
+			errors: true,
+		},
+	}
+
+	for _, e := range examples {
+		b := new(bytes.Buffer)
+		err := convert(e.input, b)
+		if e.errors {
+			require.Error(t, err)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, e.expected, b.String())
+		}
+	}
+}
+
+type badReader struct{}
+
+func (r *badReader) Read([]byte) (int, error) {
+	return 0, fmt.Errorf("reader failed on purpose")
+}

cmd/tomll/main.go

@@ -0,0 +1,58 @@
+// Package tomll is a linter program for TOML.
+//
+// Usage
+//
+// Reading from stdin, writing to stdout:
+//
+//   cat file.toml | tomll
+//
+// Reading and updating a list of files in place:
+//
+//   tomll a.toml b.toml c.toml
+//
+// Installation
+//
+// Using Go:
+//
+//   go install github.com/pelletier/go-toml/v2/cmd/tomll@latest
+package main
+
+import (
+	"io"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/internal/cli"
+)
+
+const usage = `tomll can be used in two ways:
+
+Reading from stdin, writing to stdout:
+  cat file.toml | tomll > file.toml
+
+Reading and updating a list of files in place:
+  tomll a.toml b.toml c.toml
+
+When given a list of files, tomll will modify all files in place without asking.
+`
+
+func main() {
+	p := cli.Program{
+		Usage:   usage,
+		Fn:      convert,
+		Inplace: true,
+	}
+	p.Execute()
+}
+
+func convert(r io.Reader, w io.Writer) error {
+	var v interface{}
+
+	d := toml.NewDecoder(r)
+	err := d.Decode(&v)
+	if err != nil {
+		return err
+	}
+
+	e := toml.NewEncoder(w)
+	return e.Encode(v)
+}

cmd/tomll/main_test.go

@@ -0,0 +1,46 @@
+package main
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConvert(t *testing.T) {
+	examples := []struct {
+		name     string
+		input    string
+		expected string
+		errors   bool
+	}{
+		{
+			name: "valid toml",
+			input: `
+mytoml.a = 42.0
+`,
+			expected: `[mytoml]
+a = 42.0
+
+`,
+		},
+		{
+			name:   "invalid toml",
+			input:  `[what`,
+			errors: true,
+		},
+	}
+
+	for _, e := range examples {
+		b := new(bytes.Buffer)
+		err := convert(strings.NewReader(e.input), b)
+		if e.errors {
+			require.Error(t, err)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, e.expected, b.String())
+		}
+	}
+}

cmd/tomltestgen/main.go

@@ -0,0 +1,223 @@
+// tomltestgen retrieves a given version of the language-agnostic TOML test suite in
+// https://github.com/BurntSushi/toml-test and generates go-toml unit tests.
+//
+// Within the go-toml package, run `go generate`.  Otherwise, use:
+//
+//     go run github.com/pelletier/go-toml/cmd/tomltestgen -o toml_testgen_test.go
+package main
+
+import (
+	"archive/zip"
+	"bytes"
+	"flag"
+	"fmt"
+	"go/format"
+	"io"
+	"io/ioutil"
+	"log"
+	"net/http"
+	"os"
+	"regexp"
+	"strconv"
+	"strings"
+	"text/template"
+	"time"
+)
+
+type invalid struct {
+	Name  string
+	Input string
+}
+
+type valid struct {
+	Name    string
+	Input   string
+	JsonRef string
+}
+
+type testsCollection struct {
+	Ref       string
+	Timestamp string
+	Invalid   []invalid
+	Valid     []valid
+	Count     int
+}
+
+const srcTemplate = "// Generated by tomltestgen for toml-test ref {{.Ref}} on {{.Timestamp}}\n" +
+	"package toml_test\n" +
+	" import (\n" +
+	"	\"testing\"\n" +
+	")\n" +
+
+	"{{range .Invalid}}\n" +
+	"func TestTOMLTest_Invalid_{{.Name}}(t *testing.T) {\n" +
+	"	input := {{.Input|gostr}}\n" +
+	"	testgenInvalid(t, input)\n" +
+	"}\n" +
+	"{{end}}\n" +
+	"\n" +
+	"{{range .Valid}}\n" +
+	"func TestTOMLTest_Valid_{{.Name}}(t *testing.T) {\n" +
+	"   input := {{.Input|gostr}}\n" +
+	"   jsonRef := {{.JsonRef|gostr}}\n" +
+	"   testgenValid(t, input, jsonRef)\n" +
+	"}\n" +
+	"{{end}}\n"
+
+func downloadTmpFile(url string) string {
+	log.Println("starting to download file from", url)
+	resp, err := http.Get(url)
+	if err != nil {
+		panic(err)
+	}
+	defer resp.Body.Close()
+
+	tmpfile, err := ioutil.TempFile("", "toml-test-*.zip")
+	if err != nil {
+		panic(err)
+	}
+	defer tmpfile.Close()
+
+	copiedLen, err := io.Copy(tmpfile, resp.Body)
+	if err != nil {
+		panic(err)
+	}
+	if resp.ContentLength > 0 && copiedLen != resp.ContentLength {
+		panic(fmt.Errorf("copied %d bytes, request body had %d", copiedLen, resp.ContentLength))
+	}
+	return tmpfile.Name()
+}
+
+func kebabToCamel(kebab string) string {
+	camel := ""
+	nextUpper := true
+	for _, c := range kebab {
+		if nextUpper {
+			camel += strings.ToUpper(string(c))
+			nextUpper = false
+		} else if c == '-' {
+			nextUpper = true
+		} else if c == '/' {
+			nextUpper = true
+			camel += "_"
+		} else {
+			camel += string(c)
+		}
+	}
+	return camel
+}
+
+func readFileFromZip(f *zip.File) string {
+	reader, err := f.Open()
+	if err != nil {
+		panic(err)
+	}
+	defer reader.Close()
+	bytes, err := ioutil.ReadAll(reader)
+	if err != nil {
+		panic(err)
+	}
+	return string(bytes)
+}
+
+func templateGoStr(input string) string {
+	return strconv.Quote(input)
+}
+
+var (
+	ref = flag.String("r", "master", "git reference")
+	out = flag.String("o", "", "output file")
+)
+
+func usage() {
+	_, _ = fmt.Fprintf(os.Stderr, "usage: tomltestgen [flags]\n")
+	flag.PrintDefaults()
+}
+
+func main() {
+	flag.Usage = usage
+	flag.Parse()
+
+	url := "https://codeload.github.com/BurntSushi/toml-test/zip/" + *ref
+	resultFile := downloadTmpFile(url)
+	defer os.Remove(resultFile)
+	log.Println("file written to", resultFile)
+
+	zipReader, err := zip.OpenReader(resultFile)
+	if err != nil {
+		panic(err)
+	}
+	defer zipReader.Close()
+
+	collection := testsCollection{
+		Ref:       *ref,
+		Timestamp: time.Now().Format(time.RFC3339),
+	}
+
+	zipFilesMap := map[string]*zip.File{}
+
+	for _, f := range zipReader.File {
+		zipFilesMap[f.Name] = f
+	}
+
+	testFileRegexp := regexp.MustCompile(`([^/]+/tests/(valid|invalid)/(.+))\.(toml)`)
+	for _, f := range zipReader.File {
+		groups := testFileRegexp.FindStringSubmatch(f.Name)
+		if len(groups) > 0 {
+			name := kebabToCamel(groups[3])
+			testType := groups[2]
+
+			log.Printf("> [%s] %s\n", testType, name)
+
+			tomlContent := readFileFromZip(f)
+
+			switch testType {
+			case "invalid":
+				collection.Invalid = append(collection.Invalid, invalid{
+					Name:  name,
+					Input: tomlContent,
+				})
+				collection.Count++
+			case "valid":
+				baseFilePath := groups[1]
+				jsonFilePath := baseFilePath + ".json"
+				jsonContent := readFileFromZip(zipFilesMap[jsonFilePath])
+
+				collection.Valid = append(collection.Valid, valid{
+					Name:    name,
+					Input:   tomlContent,
+					JsonRef: jsonContent,
+				})
+				collection.Count++
+			default:
+				panic(fmt.Sprintf("unknown test type: %s", testType))
+			}
+		}
+	}
+
+	log.Printf("Collected %d tests from toml-test\n", collection.Count)
+
+	funcMap := template.FuncMap{
+		"gostr": templateGoStr,
+	}
+	t := template.Must(template.New("src").Funcs(funcMap).Parse(srcTemplate))
+	buf := new(bytes.Buffer)
+	err = t.Execute(buf, collection)
+	if err != nil {
+		panic(err)
+	}
+	outputBytes, err := format.Source(buf.Bytes())
+	if err != nil {
+		panic(err)
+	}
+
+	if *out == "" {
+		fmt.Println(string(outputBytes))
+		return
+	}
+
+	err = os.WriteFile(*out, outputBytes, 0644)
+	if err != nil {
+		panic(err)
+	}
+}

decode.go

@@ -1,11 +1,9 @@
 package toml
 
 import (
-	"errors"
 	"fmt"
 	"math"
 	"strconv"
-	"strings"
 	"time"
 )
 
@@ -19,7 +17,7 @@ func parseInteger(b []byte) (int64, error) {
 		case 'o':
 			return parseIntOct(b)
 		default:
-			return 0, newDecodeError(b[1:2], "invalid base: '%c'", b[1])
+			panic(fmt.Errorf("invalid base '%c', should have been checked by scanIntOrFloat", b[1]))
 		}
 	}
 
@@ -41,34 +39,33 @@ func parseLocalDate(b []byte) (LocalDate, error) {
 
 	date.Year, err = parseDecimalDigits(b[0:4])
 	if err != nil {
-		return date, err
+		return LocalDate{}, err
 	}
 
-	v, err := parseDecimalDigits(b[5:7])
+	date.Month, err = parseDecimalDigits(b[5:7])
 	if err != nil {
-		return date, err
+		return LocalDate{}, err
 	}
 
-	date.Month = time.Month(v)
-
 	date.Day, err = parseDecimalDigits(b[8:10])
 	if err != nil {
-		return date, err
+		return LocalDate{}, err
+	}
+
+	if !isValidDate(date.Year, date.Month, date.Day) {
+		return LocalDate{}, newDecodeError(b, "impossible date")
 	}
 
 	return date, nil
 }
 
-var errNotDigit = errors.New("not a digit")
-
 func parseDecimalDigits(b []byte) (int, error) {
 	v := 0
 
-	for _, c := range b {
-		if !isDigit(c) {
-			return 0, fmt.Errorf("%s: %w", b, errNotDigit)
+	for i, c := range b {
+		if c < '0' || c > '9' {
+			return 0, newDecodeError(b[i:i+1], "expected digit (0-9)")
 		}
-
 		v *= 10
 		v += int(c - '0')
 	}
@@ -76,13 +73,12 @@ func parseDecimalDigits(b []byte) (int, error) {
 	return v, nil
 }
 
-var errParseDateTimeMissingInfo = errors.New("date-time missing timezone information")
-
 func parseDateTime(b []byte) (time.Time, error) {
 	// offset-date-time = full-date time-delim full-time
 	// full-time      = partial-time time-offset
 	// time-offset    = "Z" / time-numoffset
 	// time-numoffset = ( "+" / "-" ) time-hour ":" time-minute
+
 	dt, b, err := parseLocalDateTime(b)
 	if err != nil {
 		return time.Time{}, err
@@ -91,10 +87,11 @@ func parseDateTime(b []byte) (time.Time, error) {
 	var zone *time.Location
 
 	if len(b) == 0 {
-		return time.Time{}, errParseDateTimeMissingInfo
+		// parser should have checked that when assigning the date time node
+		panic("date time should have a timezone")
 	}
 
-	if b[0] == 'Z' {
+	if b[0] == 'Z' || b[0] == 'z' {
 		b = b[1:]
 		zone = time.UTC
 	} else {
@@ -102,19 +99,39 @@ func parseDateTime(b []byte) (time.Time, error) {
 		if len(b) != dateTimeByteLen {
 			return time.Time{}, newDecodeError(b, "invalid date-time timezone")
 		}
-		direction := 1
+		var direction int
 		switch b[0] {
-		case '+':
 		case '-':
 			direction = -1
+		case '+':
+			direction = +1
 		default:
-			return time.Time{}, newDecodeError(b[0:1], "invalid timezone offset character")
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset character")
+		}
+
+		if b[3] != ':' {
+			return time.Time{}, newDecodeError(b[3:4], "expected a : separator")
+		}
+
+		hours, err := parseDecimalDigits(b[1:3])
+		if err != nil {
+			return time.Time{}, err
+		}
+		if hours > 23 {
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset hours")
+		}
+
+		minutes, err := parseDecimalDigits(b[4:6])
+		if err != nil {
+			return time.Time{}, err
+		}
+		if minutes > 59 {
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset minutes")
 		}
 
-		hours := digitsToInt(b[1:3])
-		minutes := digitsToInt(b[4:6])
 		seconds := direction * (hours*3600 + minutes*60)
 		zone = time.FixedZone("", seconds)
+		b = b[dateTimeByteLen:]
 	}
 
 	if len(b) > 0 {
@@ -122,64 +139,59 @@ func parseDateTime(b []byte) (time.Time, error) {
 	}
 
 	t := time.Date(
-		dt.Date.Year,
-		dt.Date.Month,
-		dt.Date.Day,
-		dt.Time.Hour,
-		dt.Time.Minute,
-		dt.Time.Second,
-		dt.Time.Nanosecond,
+		dt.Year,
+		time.Month(dt.Month),
+		dt.Day,
+		dt.Hour,
+		dt.Minute,
+		dt.Second,
+		dt.Nanosecond,
 		zone)
 
 	return t, nil
 }
 
-var (
-	errParseLocalDateTimeWrongLength = errors.New(
-		"local datetimes are expected to have the format YYYY-MM-DDTHH:MM:SS[.NNNNNN]",
-	)
-	errParseLocalDateTimeWrongSeparator = errors.New("datetime separator is expected to be T or a space")
-)
-
 func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
 	var dt LocalDateTime
 
-	const localDateTimeByteLen = 11
-	if len(b) < localDateTimeByteLen {
-		return dt, nil, errParseLocalDateTimeWrongLength
+	const localDateTimeByteMinLen = 11
+	if len(b) < localDateTimeByteMinLen {
+		return dt, nil, newDecodeError(b, "local datetimes are expected to have the format YYYY-MM-DDTHH:MM:SS[.NNNNNNNNN]")
 	}
 
 	date, err := parseLocalDate(b[:10])
 	if err != nil {
 		return dt, nil, err
 	}
-	dt.Date = date
+	dt.LocalDate = date
 
 	sep := b[10]
-	if sep != 'T' && sep != ' ' {
-		return dt, nil, errParseLocalDateTimeWrongSeparator
+	if sep != 'T' && sep != ' ' && sep != 't' {
+		return dt, nil, newDecodeError(b[10:11], "datetime separator is expected to be T or a space")
 	}
 
 	t, rest, err := parseLocalTime(b[11:])
 	if err != nil {
 		return dt, nil, err
 	}
-	dt.Time = t
+	dt.LocalTime = t
 
 	return dt, rest, nil
 }
 
-var errParseLocalTimeWrongLength = errors.New("times are expected to have the format HH:MM:SS[.NNNNNN]")
-
 // parseLocalTime is a bit different because it also returns the remaining
 // []byte that is didn't need. This is to allow parseDateTime to parse those
 // remaining bytes as a timezone.
 func parseLocalTime(b []byte) (LocalTime, []byte, error) {
-	var t LocalTime
+	var (
+		nspow = [10]int{0, 1e8, 1e7, 1e6, 1e5, 1e4, 1e3, 1e2, 1e1, 1e0}
+		t     LocalTime
+	)
 
+	// check if b matches to have expected format HH:MM:SS[.NNNNNN]
 	const localTimeByteLen = 8
 	if len(b) < localTimeByteLen {
-		return t, nil, errParseLocalTimeWrongLength
+		return t, nil, newDecodeError(b, "times are expected to have the format HH:MM:SS[.NNNNNN]")
 	}
 
 	var err error
@@ -189,6 +201,9 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		return t, nil, err
 	}
 
+	if t.Hour > 23 {
+		return t, nil, newDecodeError(b[0:2], "hour cannot be greater 23")
+	}
 	if b[2] != ':' {
 		return t, nil, newDecodeError(b[2:3], "expecting colon between hours and minutes")
 	}
@@ -197,7 +212,9 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 	if err != nil {
 		return t, nil, err
 	}
-
+	if t.Minute > 59 {
+		return t, nil, newDecodeError(b[3:5], "minutes cannot be greater 59")
+	}
 	if b[5] != ':' {
 		return t, nil, newDecodeError(b[5:6], "expecting colon between minutes and seconds")
 	}
@@ -207,175 +224,317 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		return t, nil, err
 	}
 
-	if len(b) >= 15 && b[8] == '.' {
-		t.Nanosecond, err = parseDecimalDigits(b[9:15])
-		if err != nil {
-			return t, nil, err
-		}
-
-		return t, b[15:], nil
+	if t.Second > 60 {
+		return t, nil, newDecodeError(b[6:8], "seconds cannot be greater 60")
 	}
 
-	return t, b[8:], nil
-}
+	b = b[8:]
 
-var (
-	errParseFloatStartDot = errors.New("float cannot start with a dot")
-	errParseFloatEndDot   = errors.New("float cannot end with a dot")
-)
+	if len(b) >= 1 && b[0] == '.' {
+		frac := 0
+		precision := 0
+		digits := 0
+
+		for i, c := range b[1:] {
+			if !isDigit(c) {
+				if i == 0 {
+					return t, nil, newDecodeError(b[0:1], "need at least one digit after fraction point")
+				}
+				break
+			}
+			digits++
+
+			const maxFracPrecision = 9
+			if i >= maxFracPrecision {
+				// go-toml allows decoding fractional seconds
+				// beyond the supported precision of 9
+				// digits. It truncates the fractional component
+				// to the supported precision and ignores the
+				// remaining digits.
+				//
+				// https://github.com/pelletier/go-toml/discussions/707
+				continue
+			}
+
+			frac *= 10
+			frac += int(c - '0')
+			precision++
+		}
+
+		if precision == 0 {
+			return t, nil, newDecodeError(b[:1], "nanoseconds need at least one digit")
+		}
+
+		t.Nanosecond = frac * nspow[precision]
+		t.Precision = precision
+
+		return t, b[1+digits:], nil
+	}
+	return t, b, nil
+}
 
 //nolint:cyclop
 func parseFloat(b []byte) (float64, error) {
-	//nolint:godox
-	// TODO: inefficient
 	if len(b) == 4 && (b[0] == '+' || b[0] == '-') && b[1] == 'n' && b[2] == 'a' && b[3] == 'n' {
 		return math.NaN(), nil
 	}
 
-	tok := string(b)
-
-	err := numberContainsInvalidUnderscore(tok)
+	cleaned, err := checkAndRemoveUnderscoresFloats(b)
 	if err != nil {
 		return 0, err
 	}
 
-	cleanedVal := cleanupNumberToken(tok)
-	if cleanedVal[0] == '.' {
-		return 0, errParseFloatStartDot
+	if cleaned[0] == '.' {
+		return 0, newDecodeError(b, "float cannot start with a dot")
 	}
 
-	if cleanedVal[len(cleanedVal)-1] == '.' {
-		return 0, errParseFloatEndDot
+	if cleaned[len(cleaned)-1] == '.' {
+		return 0, newDecodeError(b, "float cannot end with a dot")
 	}
 
-	f, err := strconv.ParseFloat(cleanedVal, 64)
+	dotAlreadySeen := false
+	for i, c := range cleaned {
+		if c == '.' {
+			if dotAlreadySeen {
+				return 0, newDecodeError(b[i:i+1], "float can have at most one decimal point")
+			}
+			if !isDigit(cleaned[i-1]) {
+				return 0, newDecodeError(b[i-1:i+1], "float decimal point must be preceded by a digit")
+			}
+			if !isDigit(cleaned[i+1]) {
+				return 0, newDecodeError(b[i:i+2], "float decimal point must be followed by a digit")
+			}
+			dotAlreadySeen = true
+		}
+	}
+
+	start := 0
+	if cleaned[0] == '+' || cleaned[0] == '-' {
+		start = 1
+	}
+	if cleaned[start] == '0' && isDigit(cleaned[start+1]) {
+		return 0, newDecodeError(b, "float integer part cannot have leading zeroes")
+	}
+
+	f, err := strconv.ParseFloat(string(cleaned), 64)
 	if err != nil {
-		return 0, fmt.Errorf("coudn't ParseFloat %w", err)
+		return 0, newDecodeError(b, "unable to parse float: %w", err)
 	}
 
 	return f, nil
 }
 
 func parseIntHex(b []byte) (int64, error) {
-	tok := string(b)
-	cleanedVal := cleanupNumberToken(tok)
-
-	err := hexNumberContainsInvalidUnderscore(cleanedVal)
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
 
-	i, err := strconv.ParseInt(cleanedVal[2:], 16, 64)
+	i, err := strconv.ParseInt(string(cleaned), 16, 64)
 	if err != nil {
-		return 0, fmt.Errorf("coudn't ParseIntHex %w", err)
+		return 0, newDecodeError(b, "couldn't parse hexadecimal number: %w", err)
 	}
 
 	return i, nil
 }
 
 func parseIntOct(b []byte) (int64, error) {
-	tok := string(b)
-	cleanedVal := cleanupNumberToken(tok)
-
-	err := numberContainsInvalidUnderscore(cleanedVal)
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
 
-	i, err := strconv.ParseInt(cleanedVal[2:], 8, 64)
+	i, err := strconv.ParseInt(string(cleaned), 8, 64)
 	if err != nil {
-		return 0, fmt.Errorf("coudn't ParseIntOct %w", err)
+		return 0, newDecodeError(b, "couldn't parse octal number: %w", err)
 	}
 
 	return i, nil
 }
 
 func parseIntBin(b []byte) (int64, error) {
-	tok := string(b)
-	cleanedVal := cleanupNumberToken(tok)
-
-	err := numberContainsInvalidUnderscore(cleanedVal)
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
 
-	i, err := strconv.ParseInt(cleanedVal[2:], 2, 64)
+	i, err := strconv.ParseInt(string(cleaned), 2, 64)
 	if err != nil {
-		return 0, fmt.Errorf("coudn't ParseIntBin %w", err)
+		return 0, newDecodeError(b, "couldn't parse binary number: %w", err)
 	}
 
 	return i, nil
 }
 
+func isSign(b byte) bool {
+	return b == '+' || b == '-'
+}
+
 func parseIntDec(b []byte) (int64, error) {
-	tok := string(b)
-	cleanedVal := cleanupNumberToken(tok)
-
-	err := numberContainsInvalidUnderscore(cleanedVal)
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b)
 	if err != nil {
 		return 0, err
 	}
 
-	i, err := strconv.ParseInt(cleanedVal, 10, 64)
+	startIdx := 0
+
+	if isSign(cleaned[0]) {
+		startIdx++
+	}
+
+	if len(cleaned) > startIdx+1 && cleaned[startIdx] == '0' {
+		return 0, newDecodeError(b, "leading zero not allowed on decimal number")
+	}
+
+	i, err := strconv.ParseInt(string(cleaned), 10, 64)
 	if err != nil {
-		return 0, fmt.Errorf("coudn't parseIntDec %w", err)
+		return 0, newDecodeError(b, "couldn't parse decimal number: %w", err)
 	}
 
 	return i, nil
 }
 
-func numberContainsInvalidUnderscore(value string) error {
-	// For large numbers, you may use underscores between digits to enhance
-	// readability. Each underscore must be surrounded by at least one digit on
-	// each side.
-	hasBefore := false
-
-	for idx, r := range value {
-		if r == '_' {
-			if !hasBefore || idx+1 >= len(value) {
-				// can't end with an underscore
-				return errInvalidUnderscore
-			}
-		}
-		hasBefore = isDigitRune(r)
+func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
+	start := 0
+	if b[start] == '+' || b[start] == '-' {
+		start++
 	}
 
-	return nil
-}
-
-func hexNumberContainsInvalidUnderscore(value string) error {
-	hasBefore := false
-
-	for idx, r := range value {
-		if r == '_' {
-			if !hasBefore || idx+1 >= len(value) {
-				// can't end with an underscore
-				return errInvalidUnderscoreHex
-			}
-		}
-		hasBefore = isHexDigit(r)
+	if len(b) == start {
+		return b, nil
 	}
 
-	return nil
+	if b[start] == '_' {
+		return nil, newDecodeError(b[start:start+1], "number cannot start with underscore")
+	}
+
+	if b[len(b)-1] == '_' {
+		return nil, newDecodeError(b[len(b)-1:], "number cannot end with underscore")
+	}
+
+	// fast path
+	i := 0
+	for ; i < len(b); i++ {
+		if b[i] == '_' {
+			break
+		}
+	}
+	if i == len(b) {
+		return b, nil
+	}
+
+	before := false
+	cleaned := make([]byte, i, len(b))
+	copy(cleaned, b)
+
+	for i++; i < len(b); i++ {
+		c := b[i]
+		if c == '_' {
+			if !before {
+				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+			}
+			before = false
+		} else {
+			before = true
+			cleaned = append(cleaned, c)
+		}
+	}
+
+	return cleaned, nil
 }
 
-func cleanupNumberToken(value string) string {
-	cleanedVal := strings.ReplaceAll(value, "_", "")
+func checkAndRemoveUnderscoresFloats(b []byte) ([]byte, error) {
+	if b[0] == '_' {
+		return nil, newDecodeError(b[0:1], "number cannot start with underscore")
+	}
 
-	return cleanedVal
+	if b[len(b)-1] == '_' {
+		return nil, newDecodeError(b[len(b)-1:], "number cannot end with underscore")
+	}
+
+	// fast path
+	i := 0
+	for ; i < len(b); i++ {
+		if b[i] == '_' {
+			break
+		}
+	}
+	if i == len(b) {
+		return b, nil
+	}
+
+	before := false
+	cleaned := make([]byte, 0, len(b))
+
+	for i := 0; i < len(b); i++ {
+		c := b[i]
+
+		switch c {
+		case '_':
+			if !before {
+				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+			}
+			if i < len(b)-1 && (b[i+1] == 'e' || b[i+1] == 'E') {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore before exponent")
+			}
+			before = false
+		case '+', '-':
+			// signed exponents
+			cleaned = append(cleaned, c)
+			before = false
+		case 'e', 'E':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after exponent")
+			}
+			cleaned = append(cleaned, c)
+		case '.':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after decimal point")
+			}
+			if i > 0 && b[i-1] == '_' {
+				return nil, newDecodeError(b[i-1:i], "cannot have underscore before decimal point")
+			}
+			cleaned = append(cleaned, c)
+		default:
+			before = true
+			cleaned = append(cleaned, c)
+		}
+	}
+
+	return cleaned, nil
 }
 
-func isHexDigit(r rune) bool {
-	return isDigitRune(r) ||
-		(r >= 'a' && r <= 'f') ||
-		(r >= 'A' && r <= 'F')
+// isValidDate checks if a provided date is a date that exists.
+func isValidDate(year int, month int, day int) bool {
+	return month > 0 && month < 13 && day > 0 && day <= daysIn(month, year)
 }
 
-func isDigitRune(r rune) bool {
-	return r >= '0' && r <= '9'
+// daysBefore[m] counts the number of days in a non-leap year
+// before month m begins. There is an entry for m=12, counting
+// the number of days before January of next year (365).
+var daysBefore = [...]int32{
+	0,
+	31,
+	31 + 28,
+	31 + 28 + 31,
+	31 + 28 + 31 + 30,
+	31 + 28 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31 + 30 + 31,
 }
 
-var (
-	errInvalidUnderscore    = errors.New("invalid use of _ in number")
-	errInvalidUnderscoreHex = errors.New("invalid use of _ in hex number")
-)
+func daysIn(m int, year int) int {
+	if m == 2 && isLeap(year) {
+		return 29
+	}
+	return int(daysBefore[m] - daysBefore[m-1])
+}
+
+func isLeap(year int) bool {
+	return year%4 == 0 && (year%100 != 0 || year%400 == 0)
+}

doc.go

@@ -0,0 +1,2 @@
+// Package toml is a library to read and write TOML documents.
+package toml

errors.go

@@ -5,7 +5,7 @@ import (
 	"strconv"
 	"strings"
 
-	"github.com/pelletier/go-toml/v2/internal/unsafe"
+	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
 // DecodeError represents an error encountered during the parsing or decoding
@@ -18,15 +18,49 @@ type DecodeError struct {
 	message string
 	line    int
 	column  int
+	key     Key
 
 	human string
 }
 
+// StrictMissingError occurs in a TOML document that does not have a
+// corresponding field in the target value. It contains all the missing fields
+// in Errors.
+//
+// Emitted by Decoder when DisallowUnknownFields() was called.
+type StrictMissingError struct {
+	// One error per field that could not be found.
+	Errors []DecodeError
+}
+
+// Error returns the canonical string for this error.
+func (s *StrictMissingError) Error() string {
+	return "strict mode: fields in the document are missing in the target struct"
+}
+
+// String returns a human readable description of all errors.
+func (s *StrictMissingError) String() string {
+	var buf strings.Builder
+
+	for i, e := range s.Errors {
+		if i > 0 {
+			buf.WriteString("\n---\n")
+		}
+
+		buf.WriteString(e.String())
+	}
+
+	return buf.String()
+}
+
+type Key []string
+
 // internal version of DecodeError that is used as the base to create a
 // DecodeError with full context.
 type decodeError struct {
 	highlight []byte
 	message   string
+	key       Key // optional
 }
 
 func (de *decodeError) Error() string {
@@ -36,13 +70,13 @@ func (de *decodeError) Error() string {
 func newDecodeError(highlight []byte, format string, args ...interface{}) error {
 	return &decodeError{
 		highlight: highlight,
-		message:   fmt.Sprintf(format, args...),
+		message:   fmt.Errorf(format, args...).Error(),
 	}
 }
 
 // Error returns the error message contained in the DecodeError.
 func (e *DecodeError) Error() string {
-	return e.message
+	return "toml: " + e.message
 }
 
 // String returns the human-readable contextualized error. This string is multi-line.
@@ -56,7 +90,13 @@ func (e *DecodeError) Position() (row int, column int) {
 	return e.line, e.column
 }
 
-// decodeErrorFromHighlight creates a DecodeError referencing to a highlighted
+// Key that was being processed when the error occurred. The key is present only
+// if this DecodeError is part of a StrictMissingError.
+func (e *DecodeError) Key() Key {
+	return e.key
+}
+
+// decodeErrorFromHighlight creates a DecodeError referencing a highlighted
 // range of bytes from document.
 //
 // highlight needs to be a sub-slice of document, or this function panics.
@@ -64,14 +104,10 @@ func (e *DecodeError) Position() (row int, column int) {
 // The function copies all bytes used in DecodeError, so that document and
 // highlight can be freely deallocated.
 //nolint:funlen
-func wrapDecodeError(document []byte, de *decodeError) error {
-	if de == nil {
-		return nil
-	}
+func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
+	offset := danger.SubsliceOffset(document, de.highlight)
 
-	offset := unsafe.SubsliceOffset(document, de.highlight)
-
-	errMessage := de.message
+	errMessage := de.Error()
 	errLine, errColumn := positionAtEnd(document[:offset])
 	before, after := linesOfContext(document, de.highlight, offset, 3)
 
@@ -80,6 +116,7 @@ func wrapDecodeError(document []byte, de *decodeError) error {
 	maxLine := errLine + len(after) - 1
 	lineColumnWidth := len(strconv.Itoa(maxLine))
 
+	// Write the lines of context strictly before the error.
 	for i := len(before) - 1; i > 0; i-- {
 		line := errLine - i
 		buf.WriteString(formatLineNumber(line, lineColumnWidth))
@@ -93,6 +130,8 @@ func wrapDecodeError(document []byte, de *decodeError) error {
 		buf.WriteRune('\n')
 	}
 
+	// Write the document line that contains the error.
+
 	buf.WriteString(formatLineNumber(errLine, lineColumnWidth))
 	buf.WriteString("| ")
 
@@ -107,6 +146,10 @@ func wrapDecodeError(document []byte, de *decodeError) error {
 	}
 
 	buf.WriteRune('\n')
+
+	// Write the line with the error message itself (so it does not have a line
+	// number).
+
 	buf.WriteString(strings.Repeat(" ", lineColumnWidth))
 	buf.WriteString("| ")
 
@@ -121,6 +164,8 @@ func wrapDecodeError(document []byte, de *decodeError) error {
 		buf.WriteString(errMessage)
 	}
 
+	// Write the lines of context strictly after the error.
+
 	for i := 1; i < len(after); i++ {
 		buf.WriteRune('\n')
 		line := errLine + i
@@ -137,6 +182,7 @@ func wrapDecodeError(document []byte, de *decodeError) error {
 		message: errMessage,
 		line:    errLine,
 		column:  errColumn,
+		key:     de.key,
 		human:   buf.String(),
 	}
 }
@@ -193,7 +239,7 @@ forward:
 			rest = rest[o+1:]
 			o = 0
 
-		case o == len(rest)-1 && o > 0:
+		case o == len(rest)-1:
 			// add last line only if it's non-empty
 			afterLines = append(afterLines, rest)
 

errors_test.go

@@ -3,6 +3,7 @@ package toml
 import (
 	"bytes"
 	"errors"
+	"fmt"
 	"strings"
 	"testing"
 
@@ -11,7 +12,6 @@ import (
 
 //nolint:funlen
 func TestDecodeError(t *testing.T) {
-	t.Parallel()
 
 	examples := []struct {
 		desc     string
@@ -148,12 +148,19 @@ line 5`,
 6|
 7| line 4`,
 		},
+		{
+			desc: "handle remainder of the error line when there is only one line",
+			doc:  [3]string{`P=`, `[`, `#`},
+			msg:  "array is incomplete",
+			expected: `1| P=[#
+ |   ~ array is incomplete`,
+		},
 	}
 
 	for _, e := range examples {
 		e := e
 		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
+
 			b := bytes.Buffer{}
 			b.Write([]byte(e.doc[0]))
 			start := b.Len()
@@ -179,3 +186,41 @@ line 5`,
 		})
 	}
 }
+
+func TestDecodeError_Accessors(t *testing.T) {
+
+	e := DecodeError{
+		message: "foo",
+		line:    1,
+		column:  2,
+		key:     []string{"one", "two"},
+		human:   "bar",
+	}
+	assert.Equal(t, "toml: foo", e.Error())
+	r, c := e.Position()
+	assert.Equal(t, 1, r)
+	assert.Equal(t, 2, c)
+	assert.Equal(t, Key{"one", "two"}, e.Key())
+	assert.Equal(t, "bar", e.String())
+}
+
+func ExampleDecodeError() {
+	doc := `name = 123__456`
+
+	s := map[string]interface{}{}
+	err := Unmarshal([]byte(doc), &s)
+
+	fmt.Println(err)
+
+	var derr *DecodeError
+	if errors.As(err, &derr) {
+		fmt.Println(derr.String())
+		row, col := derr.Position()
+		fmt.Println("error occurred at row", row, "column", col)
+	}
+	// Output:
+	// toml: number must have at least one digit between underscores
+	// 1| name = 123__456
+	//  |           ~~ number must have at least one digit between underscores
+	// error occurred at row 1 column 11
+}

fast_test.go

@@ -0,0 +1,100 @@
+package toml_test
+
+import (
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+func TestFastSimple(t *testing.T) {
+	m := map[string]int64{}
+	err := toml.Unmarshal([]byte(`a = 42`), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]int64{"a": 42}, m)
+}
+
+func TestFastSimpleString(t *testing.T) {
+	m := map[string]string{}
+	err := toml.Unmarshal([]byte(`a = "hello"`), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]string{"a": "hello"}, m)
+}
+
+func TestFastSimpleInterface(t *testing.T) {
+	m := map[string]interface{}{}
+	err := toml.Unmarshal([]byte(`
+	a = "hello"
+	b = 42`), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]interface{}{
+		"a": "hello",
+		"b": int64(42),
+	}, m)
+}
+
+func TestFastMultipartKeyInterface(t *testing.T) {
+	m := map[string]interface{}{}
+	err := toml.Unmarshal([]byte(`
+	a.interim = "test"
+	a.b.c = "hello"
+	b = 42`), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]interface{}{
+		"a": map[string]interface{}{
+			"interim": "test",
+			"b": map[string]interface{}{
+				"c": "hello",
+			},
+		},
+		"b": int64(42),
+	}, m)
+}
+
+func TestFastExistingMap(t *testing.T) {
+	m := map[string]interface{}{
+		"ints": map[string]int{},
+	}
+	err := toml.Unmarshal([]byte(`
+	ints.one = 1
+	ints.two = 2
+	strings.yo = "hello"`), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]interface{}{
+		"ints": map[string]interface{}{
+			"one": int64(1),
+			"two": int64(2),
+		},
+		"strings": map[string]interface{}{
+			"yo": "hello",
+		},
+	}, m)
+}
+
+func TestFastArrayTable(t *testing.T) {
+	b := []byte(`
+	[root]
+	[[root.nested]]
+	name = 'Bob'
+	[[root.nested]]
+	name = 'Alice'
+	`)
+
+	m := map[string]interface{}{}
+
+	err := toml.Unmarshal(b, &m)
+	require.NoError(t, err)
+
+	require.Equal(t, map[string]interface{}{
+		"root": map[string]interface{}{
+			"nested": []interface{}{
+				map[string]interface{}{
+					"name": "Bob",
+				},
+				map[string]interface{}{
+					"name": "Alice",
+				},
+			},
+		},
+	}, m)
+}

go.mod

@@ -1,5 +1,5 @@
 module github.com/pelletier/go-toml/v2
 
-go 1.15
+go 1.16
 
-require github.com/stretchr/testify v1.7.0
+require github.com/stretchr/testify v1.7.1

go.sum

@@ -3,8 +3,8 @@ github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSs
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/testify v1.7.0 h1:nwc3DEeHmmLAfoZucVR881uASk0Mfjw8xYJ99tb5CcY=
-github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+github.com/stretchr/testify v1.7.1 h1:5TQK59W5E3v0r2duFAb7P95B6hEeOyEnHRa8MjYSMTY=
+github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c h1:dUUwHk2QECo/6vqA44rthZ8ie2QXMNeKRTHCNY2nXvo=

internal/ast/ast.go

@@ -2,6 +2,9 @@ package ast
 
 import (
 	"fmt"
+	"unsafe"
+
+	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
 // Iterator starts uninitialized, you need to call Next() first.
@@ -14,11 +17,11 @@ import (
 // }
 type Iterator struct {
 	started bool
-	node    Node
+	node    *Node
 }
 
-// Next moves the iterator forward and returns true if points to a node, false
-// otherwise.
+// Next moves the iterator forward and returns true if points to a
+// node, false otherwise.
 func (c *Iterator) Next() bool {
 	if !c.started {
 		c.started = true
@@ -28,8 +31,14 @@ func (c *Iterator) Next() bool {
 	return c.node.Valid()
 }
 
+// IsLast returns true if the current node of the iterator is the last
+// one.  Subsequent call to Next() will return false.
+func (c *Iterator) IsLast() bool {
+	return c.node.next == 0
+}
+
 // Node returns a copy of the node pointed at by the iterator.
-func (c *Iterator) Node() Node {
+func (c *Iterator) Node() *Node {
 	return c.node
 }
 
@@ -44,65 +53,69 @@ type Root struct {
 func (r *Root) Iterator() Iterator {
 	it := Iterator{}
 	if len(r.nodes) > 0 {
-		it.node = r.nodes[0]
+		it.node = &r.nodes[0]
 	}
 	return it
 }
 
-func (r *Root) at(idx int) Node {
-	// TODO: unsafe to point to the node directly
-	return r.nodes[idx]
+func (r *Root) at(idx Reference) *Node {
+	return &r.nodes[idx]
 }
 
-// Arrays have one child per element in the array.
-// InlineTables have one child per key-value pair in the table.
-// KeyValues have at least two children. The first one is the value. The
-// rest make a potentially dotted key.
-// Table and Array table have one child per element of the key they
-// represent (same as KeyValue, but without the last node being the value).
-// children []Node
+// Arrays have one child per element in the array.  InlineTables have
+// one child per key-value pair in the table.  KeyValues have at least
+// two children. The first one is the value. The rest make a
+// potentially dotted key.  Table and Array table have one child per
+// element of the key they represent (same as KeyValue, but without
+// the last node being the value).
 type Node struct {
 	Kind Kind
-	Data []byte // Raw bytes from the input
+	Raw  Range  // Raw bytes from the input.
+	Data []byte // Node value (either allocated or referencing the input).
 
-	// next idx (in the root array). 0 if last of the collection.
-	next int
-	// child idx (in the root array). 0 if no child.
-	child int
-	// pointer to the root array
-	root *Root
+	// References to other nodes, as offsets in the backing array
+	// from this node. References can go backward, so those can be
+	// negative.
+	next  int // 0 if last element
+	child int // 0 if no child
 }
 
-// Next returns a copy of the next node, or an invalid Node if there is no
-// next node.
-func (n Node) Next() Node {
-	if n.next <= 0 {
-		return noNode
-	}
-	return n.root.at(n.next)
+type Range struct {
+	Offset uint32
+	Length uint32
 }
 
-// Child returns a copy of the first child node of this node. Other children
-// can be accessed calling Next on the first child.
-// Returns an invalid Node if there is none.
-func (n Node) Child() Node {
-	if n.child <= 0 {
-		return noNode
+// Next returns a copy of the next node, or an invalid Node if there
+// is no next node.
+func (n *Node) Next() *Node {
+	if n.next == 0 {
+		return nil
 	}
-	return n.root.at(n.child)
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.next))
+}
+
+// Child returns a copy of the first child node of this node. Other
+// children can be accessed calling Next on the first child.  Returns
+// an invalid Node if there is none.
+func (n *Node) Child() *Node {
+	if n.child == 0 {
+		return nil
+	}
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.child))
 }
 
 // Valid returns true if the node's kind is set (not to Invalid).
-func (n Node) Valid() bool {
-	return n.Kind != Invalid
+func (n *Node) Valid() bool {
+	return n != nil
 }
 
-var noNode = Node{}
-
-// Key returns the child nodes making the Key on a supported node. Panics
-// otherwise.
-// They are guaranteed to be all be of the Kind Key. A simple key would return
-// just one element.
+// Key returns the child nodes making the Key on a supported
+// node. Panics otherwise.  They are guaranteed to be all be of the
+// Kind Key. A simple key would return just one element.
 func (n *Node) Key() Iterator {
 	switch n.Kind {
 	case KeyValue:
@@ -119,20 +132,13 @@ func (n *Node) Key() Iterator {
 }
 
 // Value returns a pointer to the value node of a KeyValue.
-// Guaranteed to be non-nil.
-// Panics if not called on a KeyValue node, or if the Children are malformed.
-func (n Node) Value() Node {
-	assertKind(KeyValue, n)
+// Guaranteed to be non-nil.  Panics if not called on a KeyValue node,
+// or if the Children are malformed.
+func (n *Node) Value() *Node {
 	return n.Child()
 }
 
 // Children returns an iterator over a node's children.
-func (n Node) Children() Iterator {
+func (n *Node) Children() Iterator {
 	return Iterator{node: n.Child()}
 }
-
-func assertKind(k Kind, n Node) {
-	if n.Kind != k {
-		panic(fmt.Errorf("method was expecting a %s, not a %s", k, n.Kind))
-	}
-}

internal/ast/builder.go

@@ -1,12 +1,11 @@
 package ast
 
-type Reference struct {
-	idx int
-	set bool
-}
+type Reference int
+
+const InvalidReference Reference = -1
 
 func (r Reference) Valid() bool {
-	return r.set
+	return r != InvalidReference
 }
 
 type Builder struct {
@@ -18,8 +17,8 @@ func (b *Builder) Tree() *Root {
 	return &b.tree
 }
 
-func (b *Builder) NodeAt(ref Reference) Node {
-	return b.tree.at(ref.idx)
+func (b *Builder) NodeAt(ref Reference) *Node {
+	return b.tree.at(ref)
 }
 
 func (b *Builder) Reset() {
@@ -28,33 +27,25 @@ func (b *Builder) Reset() {
 }
 
 func (b *Builder) Push(n Node) Reference {
-	n.root = &b.tree
 	b.lastIdx = len(b.tree.nodes)
 	b.tree.nodes = append(b.tree.nodes, n)
-	return Reference{
-		idx: b.lastIdx,
-		set: true,
-	}
+	return Reference(b.lastIdx)
 }
 
 func (b *Builder) PushAndChain(n Node) Reference {
-	n.root = &b.tree
 	newIdx := len(b.tree.nodes)
 	b.tree.nodes = append(b.tree.nodes, n)
 	if b.lastIdx >= 0 {
-		b.tree.nodes[b.lastIdx].next = newIdx
+		b.tree.nodes[b.lastIdx].next = newIdx - b.lastIdx
 	}
 	b.lastIdx = newIdx
-	return Reference{
-		idx: b.lastIdx,
-		set: true,
-	}
+	return Reference(b.lastIdx)
 }
 
 func (b *Builder) AttachChild(parent Reference, child Reference) {
-	b.tree.nodes[parent.idx].child = child.idx
+	b.tree.nodes[parent].child = int(child) - int(parent)
 }
 
 func (b *Builder) Chain(from Reference, to Reference) {
-	b.tree.nodes[from.idx].next = to.idx
+	b.tree.nodes[from].next = int(to) - int(from)
 }

internal/ast/kind.go

@@ -25,9 +25,9 @@ const (
 	Float
 	Integer
 	LocalDate
+	LocalTime
 	LocalDateTime
 	DateTime
-	Time
 )
 
 func (k Kind) String() string {
@@ -58,12 +58,12 @@ func (k Kind) String() string {
 		return "Integer"
 	case LocalDate:
 		return "LocalDate"
+	case LocalTime:
+		return "LocalTime"
 	case LocalDateTime:
 		return "LocalDateTime"
 	case DateTime:
 		return "DateTime"
-	case Time:
-		return "Time"
 	}
 	panic(fmt.Errorf("Kind.String() not implemented for '%d'", k))
 }

internal/cli/cli.go

@@ -0,0 +1,88 @@
+package cli
+
+import (
+	"bytes"
+	"errors"
+	"flag"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"os"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+type ConvertFn func(r io.Reader, w io.Writer) error
+
+type Program struct {
+	Usage string
+	Fn    ConvertFn
+	// Inplace allows the command to take more than one file as argument and
+	// perform convertion in place on each provided file.
+	Inplace bool
+}
+
+func (p *Program) Execute() {
+	flag.Usage = func() { fmt.Fprintf(os.Stderr, p.Usage) }
+	flag.Parse()
+	os.Exit(p.main(flag.Args(), os.Stdin, os.Stdout, os.Stderr))
+}
+
+func (p *Program) main(files []string, input io.Reader, output, error io.Writer) int {
+	err := p.run(files, input, output)
+	if err != nil {
+
+		var derr *toml.DecodeError
+		if errors.As(err, &derr) {
+			fmt.Fprintln(error, derr.String())
+			row, col := derr.Position()
+			fmt.Fprintln(error, "error occurred at row", row, "column", col)
+		} else {
+			fmt.Fprintln(error, err.Error())
+		}
+
+		return -1
+	}
+	return 0
+}
+
+func (p *Program) run(files []string, input io.Reader, output io.Writer) error {
+	if len(files) > 0 {
+		if p.Inplace {
+			return p.runAllFilesInPlace(files)
+		}
+		f, err := os.Open(files[0])
+		if err != nil {
+			return err
+		}
+		defer f.Close()
+		input = f
+	}
+	return p.Fn(input, output)
+}
+
+func (p *Program) runAllFilesInPlace(files []string) error {
+	for _, path := range files {
+		err := p.runFileInPlace(path)
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (p *Program) runFileInPlace(path string) error {
+	in, err := ioutil.ReadFile(path)
+	if err != nil {
+		return err
+	}
+
+	out := new(bytes.Buffer)
+
+	err = p.Fn(bytes.NewReader(in), out)
+	if err != nil {
+		return err
+	}
+
+	return ioutil.WriteFile(path, out.Bytes(), 0600)
+}

internal/cli/cli_test.go

@@ -0,0 +1,172 @@
+package cli
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"io/ioutil"
+	"os"
+	"path"
+	"strings"
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func processMain(args []string, input io.Reader, stdout, stderr io.Writer, f ConvertFn) int {
+	p := Program{Fn: f}
+	return p.main(args, input, stdout, stderr)
+}
+
+func TestProcessMainStdin(t *testing.T) {
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+	input := strings.NewReader("this is the input")
+
+	exit := processMain([]string{}, input, stdout, stderr, func(r io.Reader, w io.Writer) error {
+		return nil
+	})
+
+	assert.Equal(t, 0, exit)
+	assert.Empty(t, stdout.String())
+	assert.Empty(t, stderr.String())
+}
+
+func TestProcessMainStdinErr(t *testing.T) {
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+	input := strings.NewReader("this is the input")
+
+	exit := processMain([]string{}, input, stdout, stderr, func(r io.Reader, w io.Writer) error {
+		return fmt.Errorf("something bad")
+	})
+
+	assert.Equal(t, -1, exit)
+	assert.Empty(t, stdout.String())
+	assert.NotEmpty(t, stderr.String())
+}
+
+func TestProcessMainStdinDecodeErr(t *testing.T) {
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+	input := strings.NewReader("this is the input")
+
+	exit := processMain([]string{}, input, stdout, stderr, func(r io.Reader, w io.Writer) error {
+		var v interface{}
+		return toml.Unmarshal([]byte(`qwe = 001`), &v)
+	})
+
+	assert.Equal(t, -1, exit)
+	assert.Empty(t, stdout.String())
+	assert.Contains(t, stderr.String(), "error occurred at")
+}
+
+func TestProcessMainFileExists(t *testing.T) {
+	tmpfile, err := ioutil.TempFile("", "example")
+	require.NoError(t, err)
+	defer os.Remove(tmpfile.Name())
+	_, err = tmpfile.Write([]byte(`some data`))
+	require.NoError(t, err)
+
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+
+	exit := processMain([]string{tmpfile.Name()}, nil, stdout, stderr, func(r io.Reader, w io.Writer) error {
+		return nil
+	})
+
+	assert.Equal(t, 0, exit)
+	assert.Empty(t, stdout.String())
+	assert.Empty(t, stderr.String())
+}
+
+func TestProcessMainFileDoesNotExist(t *testing.T) {
+	stdout := new(bytes.Buffer)
+	stderr := new(bytes.Buffer)
+
+	exit := processMain([]string{"/lets/hope/this/does/not/exist"}, nil, stdout, stderr, func(r io.Reader, w io.Writer) error {
+		return nil
+	})
+
+	assert.Equal(t, -1, exit)
+	assert.Empty(t, stdout.String())
+	assert.NotEmpty(t, stderr.String())
+}
+
+func TestProcessMainFilesInPlace(t *testing.T) {
+	dir, err := ioutil.TempDir("", "")
+	require.NoError(t, err)
+	defer os.RemoveAll(dir)
+
+	path1 := path.Join(dir, "file1")
+	path2 := path.Join(dir, "file2")
+
+	err = ioutil.WriteFile(path1, []byte("content 1"), 0600)
+	require.NoError(t, err)
+	err = ioutil.WriteFile(path2, []byte("content 2"), 0600)
+	require.NoError(t, err)
+
+	p := Program{
+		Fn:      dummyFileFn,
+		Inplace: true,
+	}
+
+	exit := p.main([]string{path1, path2}, os.Stdin, os.Stdout, os.Stderr)
+
+	require.Equal(t, 0, exit)
+
+	v1, err := ioutil.ReadFile(path1)
+	require.NoError(t, err)
+	require.Equal(t, "1", string(v1))
+
+	v2, err := ioutil.ReadFile(path2)
+	require.NoError(t, err)
+	require.Equal(t, "2", string(v2))
+}
+
+func TestProcessMainFilesInPlaceErrRead(t *testing.T) {
+	p := Program{
+		Fn:      dummyFileFn,
+		Inplace: true,
+	}
+
+	exit := p.main([]string{"/this/path/is/invalid"}, os.Stdin, os.Stdout, os.Stderr)
+
+	require.Equal(t, -1, exit)
+}
+
+func TestProcessMainFilesInPlaceFailFn(t *testing.T) {
+	dir, err := ioutil.TempDir("", "")
+	require.NoError(t, err)
+	defer os.RemoveAll(dir)
+
+	path1 := path.Join(dir, "file1")
+
+	err = ioutil.WriteFile(path1, []byte("content 1"), 0600)
+	require.NoError(t, err)
+
+	p := Program{
+		Fn:      func(io.Reader, io.Writer) error { return fmt.Errorf("oh no") },
+		Inplace: true,
+	}
+
+	exit := p.main([]string{path1}, os.Stdin, os.Stdout, os.Stderr)
+
+	require.Equal(t, -1, exit)
+
+	v1, err := ioutil.ReadFile(path1)
+	require.NoError(t, err)
+	require.Equal(t, "content 1", string(v1))
+}
+
+func dummyFileFn(r io.Reader, w io.Writer) error {
+	b, err := ioutil.ReadAll(r)
+	if err != nil {
+		return err
+	}
+	v := strings.SplitN(string(b), " ", 2)[1]
+	_, err = w.Write([]byte(v))
+	return err
+}

internal/danger/danger.go

@@ -0,0 +1,65 @@
+package danger
+
+import (
+	"fmt"
+	"reflect"
+	"unsafe"
+)
+
+const maxInt = uintptr(int(^uint(0) >> 1))
+
+func SubsliceOffset(data []byte, subslice []byte) int {
+	datap := (*reflect.SliceHeader)(unsafe.Pointer(&data))
+	hlp := (*reflect.SliceHeader)(unsafe.Pointer(&subslice))
+
+	if hlp.Data < datap.Data {
+		panic(fmt.Errorf("subslice address (%d) is before data address (%d)", hlp.Data, datap.Data))
+	}
+	offset := hlp.Data - datap.Data
+
+	if offset > maxInt {
+		panic(fmt.Errorf("slice offset larger than int (%d)", offset))
+	}
+
+	intoffset := int(offset)
+
+	if intoffset > datap.Len {
+		panic(fmt.Errorf("slice offset (%d) is farther than data length (%d)", intoffset, datap.Len))
+	}
+
+	if intoffset+hlp.Len > datap.Len {
+		panic(fmt.Errorf("slice ends (%d+%d) is farther than data length (%d)", intoffset, hlp.Len, datap.Len))
+	}
+
+	return intoffset
+}
+
+func BytesRange(start []byte, end []byte) []byte {
+	if start == nil || end == nil {
+		panic("cannot call BytesRange with nil")
+	}
+	startp := (*reflect.SliceHeader)(unsafe.Pointer(&start))
+	endp := (*reflect.SliceHeader)(unsafe.Pointer(&end))
+
+	if startp.Data > endp.Data {
+		panic(fmt.Errorf("start pointer address (%d) is after end pointer address (%d)", startp.Data, endp.Data))
+	}
+
+	l := startp.Len
+	endLen := int(endp.Data-startp.Data) + endp.Len
+	if endLen > l {
+		l = endLen
+	}
+
+	if l > startp.Cap {
+		panic(fmt.Errorf("range length is larger than capacity"))
+	}
+
+	return start[:l]
+}
+
+func Stride(ptr unsafe.Pointer, size uintptr, offset int) unsafe.Pointer {
+	// TODO: replace with unsafe.Add when Go 1.17 is released
+	//   https://github.com/golang/go/issues/40481
+	return unsafe.Pointer(uintptr(ptr) + uintptr(int(size)*offset))
+}

internal/danger/danger_test.go

@@ -0,0 +1,178 @@
+package danger_test
+
+import (
+	"testing"
+	"unsafe"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/pelletier/go-toml/v2/internal/danger"
+)
+
+func TestSubsliceOffsetValid(t *testing.T) {
+	examples := []struct {
+		desc   string
+		test   func() ([]byte, []byte)
+		offset int
+	}{
+		{
+			desc: "simple",
+			test: func() ([]byte, []byte) {
+				data := []byte("hello")
+				return data, data[1:]
+			},
+			offset: 1,
+		},
+	}
+
+	for _, e := range examples {
+		t.Run(e.desc, func(t *testing.T) {
+			d, s := e.test()
+			offset := danger.SubsliceOffset(d, s)
+			assert.Equal(t, e.offset, offset)
+		})
+	}
+}
+
+func TestSubsliceOffsetInvalid(t *testing.T) {
+	examples := []struct {
+		desc string
+		test func() ([]byte, []byte)
+	}{
+		{
+			desc: "unrelated arrays",
+			test: func() ([]byte, []byte) {
+				return []byte("one"), []byte("two")
+			},
+		},
+		{
+			desc: "slice starts before data",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[5:], full[1:]
+			},
+		},
+		{
+			desc: "slice starts after data",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[:3], full[5:]
+			},
+		},
+		{
+			desc: "slice ends after data",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[:5], full[3:8]
+			},
+		},
+	}
+
+	for _, e := range examples {
+		t.Run(e.desc, func(t *testing.T) {
+			d, s := e.test()
+			require.Panics(t, func() {
+				danger.SubsliceOffset(d, s)
+			})
+		})
+	}
+}
+
+func TestStride(t *testing.T) {
+	a := []byte{1, 2, 3, 4}
+	x := &a[1]
+	n := (*byte)(danger.Stride(unsafe.Pointer(x), unsafe.Sizeof(byte(0)), 1))
+	require.Equal(t, &a[2], n)
+	n = (*byte)(danger.Stride(unsafe.Pointer(x), unsafe.Sizeof(byte(0)), -1))
+	require.Equal(t, &a[0], n)
+}
+
+func TestBytesRange(t *testing.T) {
+	type fn = func() ([]byte, []byte)
+	examples := []struct {
+		desc     string
+		test     fn
+		expected []byte
+	}{
+		{
+			desc: "simple",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[1:3], full[6:8]
+			},
+			expected: []byte("ello wo"),
+		},
+		{
+			desc: "full",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[0:1], full[len(full)-1:]
+			},
+			expected: []byte("hello world"),
+		},
+		{
+			desc: "end before start",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[len(full)-1:], full[0:1]
+			},
+		},
+		{
+			desc: "nils",
+			test: func() ([]byte, []byte) {
+				return nil, nil
+			},
+		},
+		{
+			desc: "nils start",
+			test: func() ([]byte, []byte) {
+				return nil, []byte("foo")
+			},
+		},
+		{
+			desc: "nils end",
+			test: func() ([]byte, []byte) {
+				return []byte("foo"), nil
+			},
+		},
+		{
+			desc: "start is end",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[1:3], full[1:3]
+			},
+			expected: []byte("el"),
+		},
+		{
+			desc: "end contained in start",
+			test: func() ([]byte, []byte) {
+				full := []byte("hello world")
+				return full[1:7], full[2:4]
+			},
+			expected: []byte("ello w"),
+		},
+		{
+			desc: "different backing arrays",
+			test: func() ([]byte, []byte) {
+				one := []byte("hello world")
+				two := []byte("hello world")
+				return one, two
+			},
+		},
+	}
+
+	for _, e := range examples {
+		t.Run(e.desc, func(t *testing.T) {
+			start, end := e.test()
+			if e.expected == nil {
+				require.Panics(t, func() {
+					danger.BytesRange(start, end)
+				})
+			} else {
+				res := danger.BytesRange(start, end)
+				require.Equal(t, e.expected, res)
+			}
+		})
+	}
+}

internal/danger/typeid.go

@@ -0,0 +1,23 @@
+package danger
+
+import (
+	"reflect"
+	"unsafe"
+)
+
+// typeID is used as key in encoder and decoder caches to enable using
+// the optimize runtime.mapaccess2_fast64 function instead of the more
+// expensive lookup if we were to use reflect.Type as map key.
+//
+// typeID holds the pointer to the reflect.Type value, which is unique
+// in the program.
+//
+// https://github.com/segmentio/encoding/blob/master/json/codec.go#L59-L61
+type TypeID unsafe.Pointer
+
+func MakeTypeID(t reflect.Type) TypeID {
+	// reflect.Type has the fields:
+	// typ unsafe.Pointer
+	// ptr unsafe.Pointer
+	return TypeID((*[2]unsafe.Pointer)(unsafe.Pointer(&t))[1])
+}

internal/imported_tests/marshal_imported_test.go

@@ -4,6 +4,7 @@ package imported_tests
 // defaults of v2.
 
 import (
+	"fmt"
 	"testing"
 	"time"
 
@@ -164,3 +165,34 @@ stringlist = []
 
 	require.Equal(t, string(expected), string(result))
 }
+
+type textMarshaler struct {
+	FirstName string
+	LastName  string
+}
+
+func (m textMarshaler) MarshalText() ([]byte, error) {
+	fullName := fmt.Sprintf("%s %s", m.FirstName, m.LastName)
+	return []byte(fullName), nil
+}
+
+func TestTextMarshaler(t *testing.T) {
+	type wrap struct {
+		TM textMarshaler
+	}
+
+	m := textMarshaler{FirstName: "Sally", LastName: "Fields"}
+
+	t.Run("at root", func(t *testing.T) {
+		_, err := toml.Marshal(m)
+		// in v2 we do not allow TextMarshaler at root
+		require.Error(t, err)
+	})
+
+	t.Run("leaf", func(t *testing.T) {
+		res, err := toml.Marshal(wrap{m})
+		require.NoError(t, err)
+
+		require.Equal(t, "TM = 'Sally Fields'\n", string(res))
+	})
+}

internal/imported_tests/unmarshal_imported_test.go

@@ -7,6 +7,7 @@ package imported_tests
 // marked as skipped until we figure out if that's something we want in v2.
 
 import (
+	"bytes"
 	"errors"
 	"fmt"
 	"reflect"
@@ -222,11 +223,13 @@ type testSubDoc struct {
 	unexported int    `toml:"shouldntBeHere"`
 }
 
-var biteMe = "Bite me"
-var float1 float32 = 12.3
-var float2 float32 = 45.6
-var float3 float32 = 78.9
-var subdoc = testSubDoc{"Second", 0}
+var (
+	biteMe         = "Bite me"
+	float1 float32 = 12.3
+	float2 float32 = 45.6
+	float3 float32 = 78.9
+	subdoc         = testSubDoc{"Second", 0}
+)
 
 var docData = testDoc{
 	Title:       "TOML Marshal Testing",
@@ -381,7 +384,7 @@ var intErrTomls = []string{
 }
 
 func TestErrUnmarshal(t *testing.T) {
-	var errTomls = []string{
+	errTomls := []string{
 		"bool = truly\ndate = 1979-05-27T07:32:00Z\nfloat = 123.4\nint = 5000\nstring = \"Bite me\"",
 		"bool = true\ndate = 1979-05-27T07:3200Z\nfloat = 123.4\nint = 5000\nstring = \"Bite me\"",
 		"bool = true\ndate = 1979-05-27T07:32:00Z\nfloat = 123a4\nint = 5000\nstring = \"Bite me\"",
@@ -454,35 +457,6 @@ func TestEmptytomlUnmarshal(t *testing.T) {
 	assert.Equal(t, emptyTestData, result)
 }
 
-func TestEmptyUnmarshalOmit(t *testing.T) {
-	t.Skipf("Have not figured yet if omitempty is a good idea")
-
-	type emptyMarshalTestStruct2 struct {
-		Title      string                  `toml:"title"`
-		Bool       bool                    `toml:"bool,omitempty"`
-		Int        int                     `toml:"int, omitempty"`
-		String     string                  `toml:"string,omitempty "`
-		StringList []string                `toml:"stringlist,omitempty"`
-		Ptr        *basicMarshalTestStruct `toml:"ptr,omitempty"`
-		Map        map[string]string       `toml:"map,omitempty"`
-	}
-
-	var emptyTestData2 = emptyMarshalTestStruct2{
-		Title:      "Placeholder",
-		Bool:       false,
-		Int:        0,
-		String:     "",
-		StringList: []string{},
-		Ptr:        nil,
-		Map:        map[string]string{},
-	}
-
-	result := emptyMarshalTestStruct2{}
-	err := toml.Unmarshal(emptyTestToml, &result)
-	require.NoError(t, err)
-	assert.Equal(t, emptyTestData2, result)
-}
-
 type pointerMarshalTestStruct struct {
 	Str       *string
 	List      *[]string
@@ -495,21 +469,23 @@ type pointerMarshalTestStruct struct {
 	DblPtr    *[]*[]*string
 }
 
-var pointerStr = "Hello"
-var pointerList = []string{"Hello back"}
-var pointerListPtr = []*string{&pointerStr}
-var pointerMap = map[string]string{"response": "Goodbye"}
-var pointerMapPtr = map[string]*string{"alternate": &pointerStr}
-var pointerTestData = pointerMarshalTestStruct{
-	Str:       &pointerStr,
-	List:      &pointerList,
-	ListPtr:   &pointerListPtr,
-	Map:       &pointerMap,
-	MapPtr:    &pointerMapPtr,
-	EmptyStr:  nil,
-	EmptyList: nil,
-	EmptyMap:  nil,
-}
+var (
+	pointerStr      = "Hello"
+	pointerList     = []string{"Hello back"}
+	pointerListPtr  = []*string{&pointerStr}
+	pointerMap      = map[string]string{"response": "Goodbye"}
+	pointerMapPtr   = map[string]*string{"alternate": &pointerStr}
+	pointerTestData = pointerMarshalTestStruct{
+		Str:       &pointerStr,
+		List:      &pointerList,
+		ListPtr:   &pointerListPtr,
+		Map:       &pointerMap,
+		MapPtr:    &pointerMapPtr,
+		EmptyStr:  nil,
+		EmptyList: nil,
+		EmptyMap:  nil,
+	}
+)
 
 var pointerTestToml = []byte(`List = ["Hello back"]
 ListPtr = ["Hello"]
@@ -537,15 +513,17 @@ func TestUnmarshalTypeMismatch(t *testing.T) {
 
 type nestedMarshalTestStruct struct {
 	String [][]string
-	//Struct [][]basicMarshalTestSubStruct
+	// Struct [][]basicMarshalTestSubStruct
 	StringPtr *[]*[]*string
 	// StructPtr *[]*[]*basicMarshalTestSubStruct
 }
 
-var str1 = "Three"
-var str2 = "Four"
-var strPtr = []*string{&str1, &str2}
-var strPtr2 = []*[]*string{&strPtr}
+var (
+	str1    = "Three"
+	str2    = "Four"
+	strPtr  = []*string{&str1, &str2}
+	strPtr2 = []*[]*string{&strPtr}
+)
 
 var nestedTestData = nestedMarshalTestStruct{
 	String:    [][]string{{"Five", "Six"}, {"One", "Two"}},
@@ -596,6 +574,7 @@ var nestedCustomMarshalerData = customMarshalerParent{
 var nestedCustomMarshalerToml = []byte(`friends = ["Sally Fields"]
 me = "Maiku Suteda"
 `)
+
 var nestedCustomMarshalerTomlForUnmarshal = []byte(`[friends]
 FirstName = "Sally"
 LastName = "Fields"`)
@@ -611,22 +590,12 @@ func (x *IntOrString) MarshalTOML() ([]byte, error) {
 	return []byte(s), nil
 }
 
-type textMarshaler struct {
-	FirstName string
-	LastName  string
-}
-
-func (m textMarshaler) MarshalText() ([]byte, error) {
-	fullName := fmt.Sprintf("%s %s", m.FirstName, m.LastName)
-	return []byte(fullName), nil
-}
-
 func TestUnmarshalTextMarshaler(t *testing.T) {
-	var nested = struct {
+	nested := struct {
 		Friends textMarshaler `toml:"friends"`
 	}{}
 
-	var expected = struct {
+	expected := struct {
 		Friends textMarshaler `toml:"friends"`
 	}{
 		Friends: textMarshaler{FirstName: "Sally", LastName: "Fields"},
@@ -958,6 +927,29 @@ func TestUnmarshalMapWithTypedKey(t *testing.T) {
 	}
 }
 
+func TestUnmarshalTypeTableHeader(t *testing.T) {
+	testToml := []byte(`
+		[test]
+		a = 1
+		`)
+
+	type header string
+	var result map[header]map[string]int
+	err := toml.Unmarshal(testToml, &result)
+	if err != nil {
+		t.Errorf("Received unexpected error: %s", err)
+		return
+	}
+
+	expected := map[header]map[string]int{
+		"test": map[string]int{"a": 1},
+	}
+
+	if !reflect.DeepEqual(result, expected) {
+		t.Errorf("Bad unmarshal: expected %v, got %v", expected, result)
+	}
+}
+
 func TestUnmarshalNonPointer(t *testing.T) {
 	a := 1
 	err := toml.Unmarshal([]byte{}, a)
@@ -1369,7 +1361,6 @@ func TestUnmarshalPreservesUnexportedFields(t *testing.T) {
 	t.Run("unexported field should not be set from toml", func(t *testing.T) {
 		var actual unexportedFieldPreservationTest
 		err := toml.Unmarshal([]byte(doc), &actual)
-
 		if err != nil {
 			t.Fatal("did not expect an error")
 		}
@@ -1403,7 +1394,6 @@ func TestUnmarshalPreservesUnexportedFields(t *testing.T) {
 			Nested3:    &unexportedFieldPreservationTestNested{"baz", "bax"},
 		}
 		err := toml.Unmarshal([]byte(doc), &actual)
-
 		if err != nil {
 			t.Fatal("did not expect an error")
 		}
@@ -1429,211 +1419,208 @@ func TestUnmarshalPreservesUnexportedFields(t *testing.T) {
 	})
 }
 
-//
-//func TestUnmarshalLocalDate(t *testing.T) {
-//	t.Run("ToLocalDate", func(t *testing.T) {
-//		type dateStruct struct {
-//			Date toml.LocalDate
-//		}
-//
-//		doc := `date = 1979-05-27`
-//
-//		var obj dateStruct
-//
-//		err := toml.Unmarshal([]byte(doc), &obj)
-//
-//		if err != nil {
-//			t.Fatal(err)
-//		}
-//
-//		if obj.Date.Year != 1979 {
-//			t.Errorf("expected year 1979, got %d", obj.Date.Year)
-//		}
-//		if obj.Date.Month != 5 {
-//			t.Errorf("expected month 5, got %d", obj.Date.Month)
-//		}
-//		if obj.Date.Day != 27 {
-//			t.Errorf("expected day 27, got %d", obj.Date.Day)
-//		}
-//	})
-//
-//	t.Run("ToLocalDate", func(t *testing.T) {
-//		type dateStruct struct {
-//			Date time.Time
-//		}
-//
-//		doc := `date = 1979-05-27`
-//
-//		var obj dateStruct
-//
-//		err := toml.Unmarshal([]byte(doc), &obj)
-//
-//		if err != nil {
-//			t.Fatal(err)
-//		}
-//
-//		if obj.Date.Year() != 1979 {
-//			t.Errorf("expected year 1979, got %d", obj.Date.Year())
-//		}
-//		if obj.Date.Month() != 5 {
-//			t.Errorf("expected month 5, got %d", obj.Date.Month())
-//		}
-//		if obj.Date.Day() != 27 {
-//			t.Errorf("expected day 27, got %d", obj.Date.Day())
-//		}
-//	})
-//}
-//
-//func TestUnmarshalLocalDateTime(t *testing.T) {
-//	examples := []struct {
-//		name string
-//		in   string
-//		out  toml.LocalDateTime
-//	}{
-//		{
-//			name: "normal",
-//			in:   "1979-05-27T07:32:00",
-//			out: toml.LocalDateTime{
-//				Date: toml.LocalDate{
-//					Year:  1979,
-//					Month: 5,
-//					Day:   27,
-//				},
-//				Time: toml.LocalTime{
-//					Hour:       7,
-//					Minute:     32,
-//					Second:     0,
-//					Nanosecond: 0,
-//				},
-//			}},
-//		{
-//			name: "with nanoseconds",
-//			in:   "1979-05-27T00:32:00.999999",
-//			out: toml.LocalDateTime{
-//				Date: toml.LocalDate{
-//					Year:  1979,
-//					Month: 5,
-//					Day:   27,
-//				},
-//				Time: toml.LocalTime{
-//					Hour:       0,
-//					Minute:     32,
-//					Second:     0,
-//					Nanosecond: 999999000,
-//				},
-//			},
-//		},
-//	}
-//
-//	for i, example := range examples {
-//		doc := fmt.Sprintf(`date = %s`, example.in)
-//
-//		t.Run(fmt.Sprintf("ToLocalDateTime_%d_%s", i, example.name), func(t *testing.T) {
-//			type dateStruct struct {
-//				Date toml.LocalDateTime
-//			}
-//
-//			var obj dateStruct
-//
-//			err := toml.Unmarshal([]byte(doc), &obj)
-//
-//			if err != nil {
-//				t.Fatal(err)
-//			}
-//
-//			if obj.Date != example.out {
-//				t.Errorf("expected '%s', got '%s'", example.out, obj.Date)
-//			}
-//		})
-//
-//		t.Run(fmt.Sprintf("ToTime_%d_%s", i, example.name), func(t *testing.T) {
-//			type dateStruct struct {
-//				Date time.Time
-//			}
-//
-//			var obj dateStruct
-//
-//			err := toml.Unmarshal([]byte(doc), &obj)
-//
-//			if err != nil {
-//				t.Fatal(err)
-//			}
-//
-//			if obj.Date.Year() != example.out.Date.Year {
-//				t.Errorf("expected year %d, got %d", example.out.Date.Year, obj.Date.Year())
-//			}
-//			if obj.Date.Month() != example.out.Date.Month {
-//				t.Errorf("expected month %d, got %d", example.out.Date.Month, obj.Date.Month())
-//			}
-//			if obj.Date.Day() != example.out.Date.Day {
-//				t.Errorf("expected day %d, got %d", example.out.Date.Day, obj.Date.Day())
-//			}
-//			if obj.Date.Hour() != example.out.Time.Hour {
-//				t.Errorf("expected hour %d, got %d", example.out.Time.Hour, obj.Date.Hour())
-//			}
-//			if obj.Date.Minute() != example.out.Time.Minute {
-//				t.Errorf("expected minute %d, got %d", example.out.Time.Minute, obj.Date.Minute())
-//			}
-//			if obj.Date.Second() != example.out.Time.Second {
-//				t.Errorf("expected second %d, got %d", example.out.Time.Second, obj.Date.Second())
-//			}
-//			if obj.Date.Nanosecond() != example.out.Time.Nanosecond {
-//				t.Errorf("expected nanoseconds %d, got %d", example.out.Time.Nanosecond, obj.Date.Nanosecond())
-//			}
-//		})
-//	}
-//}
-//
-//func TestUnmarshalLocalTime(t *testing.T) {
-//	examples := []struct {
-//		name string
-//		in   string
-//		out  toml.LocalTime
-//	}{
-//		{
-//			name: "normal",
-//			in:   "07:32:00",
-//			out: toml.LocalTime{
-//				Hour:       7,
-//				Minute:     32,
-//				Second:     0,
-//				Nanosecond: 0,
-//			},
-//		},
-//		{
-//			name: "with nanoseconds",
-//			in:   "00:32:00.999999",
-//			out: toml.LocalTime{
-//				Hour:       0,
-//				Minute:     32,
-//				Second:     0,
-//				Nanosecond: 999999000,
-//			},
-//		},
-//	}
-//
-//	for i, example := range examples {
-//		doc := fmt.Sprintf(`Time = %s`, example.in)
-//
-//		t.Run(fmt.Sprintf("ToLocalTime_%d_%s", i, example.name), func(t *testing.T) {
-//			type dateStruct struct {
-//				Time toml.LocalTime
-//			}
-//
-//			var obj dateStruct
-//
-//			err := toml.Unmarshal([]byte(doc), &obj)
-//
-//			if err != nil {
-//				t.Fatal(err)
-//			}
-//
-//			if obj.Time != example.out {
-//				t.Errorf("expected '%s', got '%s'", example.out, obj.Time)
-//			}
-//		})
-//	}
-//}
+func TestUnmarshalLocalDate(t *testing.T) {
+	t.Run("ToLocalDate", func(t *testing.T) {
+		type dateStruct struct {
+			Date toml.LocalDate
+		}
+
+		doc := `date = 1979-05-27`
+
+		var obj dateStruct
+
+		err := toml.Unmarshal([]byte(doc), &obj)
+		if err != nil {
+			t.Fatal(err)
+		}
+
+		if obj.Date.Year != 1979 {
+			t.Errorf("expected year 1979, got %d", obj.Date.Year)
+		}
+		if obj.Date.Month != 5 {
+			t.Errorf("expected month 5, got %d", obj.Date.Month)
+		}
+		if obj.Date.Day != 27 {
+			t.Errorf("expected day 27, got %d", obj.Date.Day)
+		}
+	})
+
+	t.Run("ToLocalDate", func(t *testing.T) {
+		type dateStruct struct {
+			Date time.Time
+		}
+
+		doc := `date = 1979-05-27`
+
+		var obj dateStruct
+
+		err := toml.Unmarshal([]byte(doc), &obj)
+		if err != nil {
+			t.Fatal(err)
+		}
+
+		if obj.Date.Year() != 1979 {
+			t.Errorf("expected year 1979, got %d", obj.Date.Year())
+		}
+		if obj.Date.Month() != 5 {
+			t.Errorf("expected month 5, got %d", obj.Date.Month())
+		}
+		if obj.Date.Day() != 27 {
+			t.Errorf("expected day 27, got %d", obj.Date.Day())
+		}
+	})
+}
+
+func TestUnmarshalLocalDateTime(t *testing.T) {
+	examples := []struct {
+		name string
+		in   string
+		out  toml.LocalDateTime
+	}{
+		{
+			name: "normal",
+			in:   "1979-05-27T07:32:00",
+			out: toml.LocalDateTime{
+				LocalDate: toml.LocalDate{
+					Year:  1979,
+					Month: 5,
+					Day:   27,
+				},
+				LocalTime: toml.LocalTime{
+					Hour:       7,
+					Minute:     32,
+					Second:     0,
+					Nanosecond: 0,
+				},
+			},
+		},
+		{
+			name: "with nanoseconds",
+			in:   "1979-05-27T00:32:00.999999",
+			out: toml.LocalDateTime{
+				LocalDate: toml.LocalDate{
+					Year:  1979,
+					Month: 5,
+					Day:   27,
+				},
+				LocalTime: toml.LocalTime{
+					Hour:       0,
+					Minute:     32,
+					Second:     0,
+					Nanosecond: 999999000,
+					Precision:  6,
+				},
+			},
+		},
+	}
+
+	for i, example := range examples {
+		doc := fmt.Sprintf(`date = %s`, example.in)
+
+		t.Run(fmt.Sprintf("ToLocalDateTime_%d_%s", i, example.name), func(t *testing.T) {
+			type dateStruct struct {
+				Date toml.LocalDateTime
+			}
+
+			var obj dateStruct
+
+			err := toml.Unmarshal([]byte(doc), &obj)
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			if obj.Date != example.out {
+				t.Errorf("expected '%s', got '%s'", example.out, obj.Date)
+			}
+		})
+
+		t.Run(fmt.Sprintf("ToTime_%d_%s", i, example.name), func(t *testing.T) {
+			type dateStruct struct {
+				Date time.Time
+			}
+
+			var obj dateStruct
+
+			err := toml.Unmarshal([]byte(doc), &obj)
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			if obj.Date.Year() != example.out.Year {
+				t.Errorf("expected year %d, got %d", example.out.Year, obj.Date.Year())
+			}
+			if obj.Date.Month() != time.Month(example.out.Month) {
+				t.Errorf("expected month %d, got %d", example.out.Month, obj.Date.Month())
+			}
+			if obj.Date.Day() != example.out.Day {
+				t.Errorf("expected day %d, got %d", example.out.Day, obj.Date.Day())
+			}
+			if obj.Date.Hour() != example.out.Hour {
+				t.Errorf("expected hour %d, got %d", example.out.Hour, obj.Date.Hour())
+			}
+			if obj.Date.Minute() != example.out.Minute {
+				t.Errorf("expected minute %d, got %d", example.out.Minute, obj.Date.Minute())
+			}
+			if obj.Date.Second() != example.out.Second {
+				t.Errorf("expected second %d, got %d", example.out.Second, obj.Date.Second())
+			}
+			if obj.Date.Nanosecond() != example.out.Nanosecond {
+				t.Errorf("expected nanoseconds %d, got %d", example.out.Nanosecond, obj.Date.Nanosecond())
+			}
+		})
+	}
+}
+
+func TestUnmarshalLocalTime(t *testing.T) {
+	examples := []struct {
+		name string
+		in   string
+		out  toml.LocalTime
+	}{
+		{
+			name: "normal",
+			in:   "07:32:00",
+			out: toml.LocalTime{
+				Hour:       7,
+				Minute:     32,
+				Second:     0,
+				Nanosecond: 0,
+			},
+		},
+		{
+			name: "with nanoseconds",
+			in:   "00:32:00.999999",
+			out: toml.LocalTime{
+				Hour:       0,
+				Minute:     32,
+				Second:     0,
+				Nanosecond: 999999000,
+				Precision:  6,
+			},
+		},
+	}
+
+	for i, example := range examples {
+		doc := fmt.Sprintf(`Time = %s`, example.in)
+
+		t.Run(fmt.Sprintf("ToLocalTime_%d_%s", i, example.name), func(t *testing.T) {
+			type dateStruct struct {
+				Time toml.LocalTime
+			}
+
+			var obj dateStruct
+
+			err := toml.Unmarshal([]byte(doc), &obj)
+			if err != nil {
+				t.Fatal(err)
+			}
+
+			if obj.Time != example.out {
+				t.Errorf("expected '%s', got '%s'", example.out, obj.Time)
+			}
+		})
+	}
+}
 
 // test case for issue #339
 func TestUnmarshalSameInnerField(t *testing.T) {
@@ -1955,66 +1942,80 @@ String2="2"`
 	assert.Error(t, err)
 }
 
+func decoder(doc string) *toml.Decoder {
+	return toml.NewDecoder(bytes.NewReader([]byte(doc)))
+}
+
+func strictDecoder(doc string) *toml.Decoder {
+	d := decoder(doc)
+	d.DisallowUnknownFields()
+	return d
+}
+
 func TestDecoderStrict(t *testing.T) {
-	t.Skip()
-	//	input := `
-	//[decoded]
-	//  key = ""
-	//
-	//[undecoded]
-	//  key = ""
-	//
-	//  [undecoded.inner]
-	//	key = ""
-	//
-	//  [[undecoded.array]]
-	//	key = ""
-	//
-	//  [[undecoded.array]]
-	//	key = ""
-	//
-	//`
-	//	var doc struct {
-	//		Decoded struct {
-	//			Key string
-	//		}
-	//	}
-	//
-	//	expected := `undecoded keys: ["undecoded.array.0.key" "undecoded.array.1.key" "undecoded.inner.key" "undecoded.key"]`
-	//
-	//	err := NewDecoder(bytes.NewReader([]byte(input))).Strict(true).Decode(&doc)
-	//	if err == nil {
-	//		t.Error("expected error, got none")
-	//	} else if err.Error() != expected {
-	//		t.Errorf("expect err: %s, got: %s", expected, err.Error())
-	//	}
-	//
-	//	if err := NewDecoder(bytes.NewReader([]byte(input))).Decode(&doc); err != nil {
-	//		t.Errorf("unexpected err: %s", err)
-	//	}
-	//
-	//	var m map[string]interface{}
-	//	if err := NewDecoder(bytes.NewReader([]byte(input))).Decode(&m); err != nil {
-	//		t.Errorf("unexpected err: %s", err)
-	//	}
+	input := `
+	[decoded]
+	 key = ""
+	
+	[undecoded]
+	 key = ""
+	
+	 [undecoded.inner]
+		key = ""
+	
+	 [[undecoded.array]]
+		key = ""
+	
+	 [[undecoded.array]]
+		key = ""
+	
+	`
+	var doc struct {
+		Decoded struct {
+			Key string
+		}
+	}
+
+	err := strictDecoder(input).Decode(&doc)
+	require.Error(t, err)
+	require.IsType(t, &toml.StrictMissingError{}, err)
+	se := err.(*toml.StrictMissingError)
+
+	keys := []toml.Key{}
+
+	for _, e := range se.Errors {
+		keys = append(keys, e.Key())
+	}
+
+	expectedKeys := []toml.Key{
+		{"undecoded"},
+		{"undecoded", "inner"},
+		{"undecoded", "array"},
+		{"undecoded", "array"},
+	}
+
+	require.Equal(t, expectedKeys, keys)
+
+	err = decoder(input).Decode(&doc)
+	require.NoError(t, err)
+
+	var m map[string]interface{}
+	err = decoder(input).Decode(&m)
 }
 
 func TestDecoderStrictValid(t *testing.T) {
-	t.Skip()
-	//	input := `
-	//[decoded]
-	//  key = ""
-	//`
-	//	var doc struct {
-	//		Decoded struct {
-	//			Key string
-	//		}
-	//	}
-	//
-	//	err := NewDecoder(bytes.NewReader([]byte(input))).Strict(true).Decode(&doc)
-	//	if err != nil {
-	//		t.Fatal("unexpected error:", err)
-	//	}
+	input := `
+	[decoded]
+	 key = ""
+	`
+	var doc struct {
+		Decoded struct {
+			Key string
+		}
+	}
+
+	err := strictDecoder(input).Decode(&doc)
+	require.NoError(t, err)
 }
 
 type docUnmarshalTOML struct {
@@ -2279,8 +2280,7 @@ func (d *durationString) UnmarshalTOML(v interface{}) error {
 	return nil
 }
 
-type config437Error struct {
-}
+type config437Error struct{}
 
 func (e *config437Error) UnmarshalTOML(v interface{}) error {
 	return errors.New("expected")

internal/testsuite/add.go

@@ -0,0 +1,74 @@
+package testsuite
+
+import (
+	"fmt"
+	"math"
+	"time"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+// addTag adds JSON tags to a data structure as expected by toml-test.
+func addTag(key string, tomlData interface{}) interface{} {
+	// Switch on the data type.
+	switch orig := tomlData.(type) {
+	default:
+		//return map[string]interface{}{}
+		panic(fmt.Sprintf("Unknown type: %T", tomlData))
+
+	// A table: we don't need to add any tags, just recurse for every table
+	// entry.
+	case map[string]interface{}:
+		typed := make(map[string]interface{}, len(orig))
+		for k, v := range orig {
+			typed[k] = addTag(k, v)
+		}
+		return typed
+
+	// An array: we don't need to add any tags, just recurse for every table
+	// entry.
+	case []map[string]interface{}:
+		typed := make([]map[string]interface{}, len(orig))
+		for i, v := range orig {
+			typed[i] = addTag("", v).(map[string]interface{})
+		}
+		return typed
+	case []interface{}:
+		typed := make([]interface{}, len(orig))
+		for i, v := range orig {
+			typed[i] = addTag("", v)
+		}
+		return typed
+
+	// Datetime: tag as datetime.
+	case toml.LocalTime:
+		return tag("time-local", orig.String())
+	case toml.LocalDate:
+		return tag("date-local", orig.String())
+	case toml.LocalDateTime:
+		return tag("datetime-local", orig.String())
+	case time.Time:
+		return tag("datetime", orig.Format("2006-01-02T15:04:05.999999999Z07:00"))
+
+	// Tag primitive values: bool, string, int, and float64.
+	case bool:
+		return tag("bool", fmt.Sprintf("%v", orig))
+	case string:
+		return tag("string", orig)
+	case int64:
+		return tag("integer", fmt.Sprintf("%d", orig))
+	case float64:
+		// Special case for nan since NaN == NaN is false.
+		if math.IsNaN(orig) {
+			return tag("float", "nan")
+		}
+		return tag("float", fmt.Sprintf("%v", orig))
+	}
+}
+
+func tag(typeName string, data interface{}) map[string]interface{} {
+	return map[string]interface{}{
+		"type":  typeName,
+		"value": data,
+	}
+}

internal/testsuite/json.go

@@ -0,0 +1,244 @@
+package testsuite
+
+import (
+	"fmt"
+	"strconv"
+	"strings"
+	"testing"
+	"time"
+)
+
+func CmpJSON(t *testing.T, key string, want, have interface{}) {
+	switch w := want.(type) {
+	case map[string]interface{}:
+		cmpJSONMaps(t, key, w, have)
+	case []interface{}:
+		cmpJSONArrays(t, key, w, have)
+	default:
+		t.Errorf(
+			"Key '%s' in expected output should be a map or a list of maps, but it's a %T",
+			key, want)
+	}
+}
+
+func cmpJSONMaps(t *testing.T, key string, want map[string]interface{}, have interface{}) {
+	haveMap, ok := have.(map[string]interface{})
+	if !ok {
+		mismatch(t, key, "table", want, haveMap)
+		return
+	}
+
+	// Check to make sure both or neither are values.
+	if isValue(want) && !isValue(haveMap) {
+		t.Fatalf("Key '%s' is supposed to be a value, but the parser reports it as a table", key)
+	}
+	if !isValue(want) && isValue(haveMap) {
+		t.Fatalf("Key '%s' is supposed to be a table, but the parser reports it as a value", key)
+	}
+	if isValue(want) && isValue(haveMap) {
+		cmpJSONValues(t, key, want, haveMap)
+		return
+	}
+
+	// Check that the keys of each map are equivalent.
+	for k := range want {
+		if _, ok := haveMap[k]; !ok {
+			bunk := kjoin(key, k)
+			t.Fatalf("Could not find key '%s' in parser output.", bunk)
+		}
+	}
+	for k := range haveMap {
+		if _, ok := want[k]; !ok {
+			bunk := kjoin(key, k)
+			t.Fatalf("Could not find key '%s' in expected output.", bunk)
+		}
+	}
+
+	// Okay, now make sure that each value is equivalent.
+	for k := range want {
+		CmpJSON(t, kjoin(key, k), want[k], haveMap[k])
+	}
+}
+
+func cmpJSONArrays(t *testing.T, key string, want, have interface{}) {
+	wantSlice, ok := want.([]interface{})
+	if !ok {
+		panic(fmt.Sprintf("'value' should be a JSON array when 'type=array', but it is a %T", want))
+	}
+
+	haveSlice, ok := have.([]interface{})
+	if !ok {
+		t.Fatalf("Malformed output from your encoder: 'value' is not a JSON array: %T", have)
+	}
+
+	if len(wantSlice) != len(haveSlice) {
+		t.Fatalf("Array lengths differ for key '%s':\n"+
+			"  Expected:     %d\n"+
+			"  Your encoder: %d",
+			key, len(wantSlice), len(haveSlice))
+	}
+	for i := 0; i < len(wantSlice); i++ {
+		CmpJSON(t, key, wantSlice[i], haveSlice[i])
+	}
+}
+
+func cmpJSONValues(t *testing.T, key string, want, have map[string]interface{}) {
+	wantType, ok := want["type"].(string)
+	if !ok {
+		panic(fmt.Sprintf("'type' should be a string, but it is a %T", want["type"]))
+	}
+
+	haveType, ok := have["type"].(string)
+	if !ok {
+		t.Fatalf("Malformed output from your encoder: 'type' is not a string: %T", have["type"])
+	}
+
+	if wantType != haveType {
+		valMismatch(t, key, wantType, haveType, want, have)
+	}
+
+	// If this is an array, then we've got to do some work to check equality.
+	if wantType == "array" {
+		cmpJSONArrays(t, key, want, have)
+		return
+	}
+
+	// Atomic values are always strings
+	wantVal, ok := want["value"].(string)
+	if !ok {
+		panic(fmt.Sprintf("'value' %v should be a string, but it is a %[1]T", want["value"]))
+	}
+
+	haveVal, ok := have["value"].(string)
+	if !ok {
+		panic(fmt.Sprintf("Malformed output from your encoder: %T is not a string", have["value"]))
+	}
+
+	// Excepting floats and datetimes, other values can be compared as strings.
+	switch wantType {
+	case "float":
+		cmpFloats(t, key, wantVal, haveVal)
+	case "datetime", "datetime-local", "date-local", "time-local":
+		cmpAsDatetimes(t, key, wantType, wantVal, haveVal)
+	default:
+		cmpAsStrings(t, key, wantVal, haveVal)
+	}
+}
+
+func cmpAsStrings(t *testing.T, key string, want, have string) {
+	if want != have {
+		t.Fatalf("Values for key '%s' don't match:\n"+
+			"  Expected:     %s\n"+
+			"  Your encoder: %s",
+			key, want, have)
+	}
+}
+
+func cmpFloats(t *testing.T, key string, want, have string) {
+	// Special case for NaN, since NaN != NaN.
+	if strings.HasSuffix(want, "nan") || strings.HasSuffix(have, "nan") {
+		if want != have {
+			t.Fatalf("Values for key '%s' don't match:\n"+
+				"  Expected:     %v\n"+
+				"  Your encoder: %v",
+				key, want, have)
+		}
+		return
+	}
+
+	wantF, err := strconv.ParseFloat(want, 64)
+	if err != nil {
+		panic(fmt.Sprintf("Could not read '%s' as a float value for key '%s'", want, key))
+	}
+
+	haveF, err := strconv.ParseFloat(have, 64)
+	if err != nil {
+		panic(fmt.Sprintf("Malformed output from your encoder: key '%s' is not a float: '%s'", key, have))
+	}
+
+	if wantF != haveF {
+		t.Fatalf("Values for key '%s' don't match:\n"+
+			"  Expected:     %v\n"+
+			"  Your encoder: %v",
+			key, wantF, haveF)
+	}
+}
+
+var datetimeRepl = strings.NewReplacer(
+	" ", "T",
+	"t", "T",
+	"z", "Z")
+
+var layouts = map[string]string{
+	"datetime":       time.RFC3339Nano,
+	"datetime-local": "2006-01-02T15:04:05.999999999",
+	"date-local":     "2006-01-02",
+	"time-local":     "15:04:05",
+}
+
+func cmpAsDatetimes(t *testing.T, key string, kind, want, have string) {
+	layout, ok := layouts[kind]
+	if !ok {
+		panic("should never happen")
+	}
+
+	wantT, err := time.Parse(layout, datetimeRepl.Replace(want))
+	if err != nil {
+		panic(fmt.Sprintf("Could not read '%s' as a datetime value for key '%s'", want, key))
+	}
+
+	haveT, err := time.Parse(layout, datetimeRepl.Replace(want))
+	if err != nil {
+		t.Fatalf("Malformed output from your encoder: key '%s' is not a datetime: '%s'", key, have)
+		return
+	}
+	if !wantT.Equal(haveT) {
+		t.Fatalf("Values for key '%s' don't match:\n"+
+			"  Expected:     %v\n"+
+			"  Your encoder: %v",
+			key, wantT, haveT)
+	}
+}
+
+func cmpAsDatetimesLocal(t *testing.T, key string, want, have string) {
+	if datetimeRepl.Replace(want) != datetimeRepl.Replace(have) {
+		t.Fatalf("Values for key '%s' don't match:\n"+
+			"  Expected:     %v\n"+
+			"  Your encoder: %v",
+			key, want, have)
+	}
+}
+
+func kjoin(old, key string) string {
+	if len(old) == 0 {
+		return key
+	}
+	return old + "." + key
+}
+
+func isValue(m map[string]interface{}) bool {
+	if len(m) != 2 {
+		return false
+	}
+	if _, ok := m["type"]; !ok {
+		return false
+	}
+	if _, ok := m["value"]; !ok {
+		return false
+	}
+	return true
+}
+
+func mismatch(t *testing.T, key string, wantType string, want, have interface{}) {
+	t.Fatalf("Key '%s' is not an %s but %[4]T:\n"+
+		"  Expected:     %#[3]v\n"+
+		"  Your encoder: %#[4]v",
+		key, wantType, want, have)
+}
+
+func valMismatch(t *testing.T, key string, wantType, haveType string, want, have interface{}) {
+	t.Fatalf("Key '%s' is not an %s but %s:\n"+
+		"  Expected:     %#[3]v\n"+
+		"  Your encoder: %#[4]v",
+		key, wantType, want, have)
+}

internal/testsuite/parser.go

@@ -0,0 +1,69 @@
+package testsuite
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+type parser struct{}
+
+func (p parser) Decode(input string) (output string, outputIsError bool, retErr error) {
+	defer func() {
+		if r := recover(); r != nil {
+			switch rr := r.(type) {
+			case error:
+				retErr = rr
+			default:
+				retErr = fmt.Errorf("%s", rr)
+			}
+		}
+	}()
+
+	var v interface{}
+
+	if err := toml.Unmarshal([]byte(input), &v); err != nil {
+		return err.Error(), true, nil
+	}
+
+	j, err := json.MarshalIndent(addTag("", v), "", "  ")
+	if err != nil {
+		return "", false, retErr
+	}
+
+	return string(j), false, retErr
+}
+
+func (p parser) Encode(input string) (output string, outputIsError bool, retErr error) {
+	defer func() {
+		if r := recover(); r != nil {
+			switch rr := r.(type) {
+			case error:
+				retErr = rr
+			default:
+				retErr = fmt.Errorf("%s", rr)
+			}
+		}
+	}()
+
+	var tmp interface{}
+	err := json.Unmarshal([]byte(input), &tmp)
+	if err != nil {
+		return "", false, err
+	}
+
+	rm, err := rmTag(tmp)
+	if err != nil {
+		return err.Error(), true, retErr
+	}
+
+	buf := new(bytes.Buffer)
+	err = toml.NewEncoder(buf).Encode(rm)
+	if err != nil {
+		return err.Error(), true, retErr
+	}
+
+	return buf.String(), false, retErr
+}

internal/testsuite/rm.go

@@ -0,0 +1,110 @@
+package testsuite
+
+import (
+	"fmt"
+	"strconv"
+	"time"
+)
+
+// Remove JSON tags to a data structure as returned by toml-test.
+func rmTag(typedJson interface{}) (interface{}, error) {
+	// Check if key is in the table m.
+	in := func(key string, m map[string]interface{}) bool {
+		_, ok := m[key]
+		return ok
+	}
+
+	// Switch on the data type.
+	switch v := typedJson.(type) {
+
+	// Object: this can either be a TOML table or a primitive with tags.
+	case map[string]interface{}:
+		// This value represents a primitive: remove the tags and return just
+		// the primitive value.
+		if len(v) == 2 && in("type", v) && in("value", v) {
+			ut, err := untag(v)
+			if err != nil {
+				return ut, fmt.Errorf("tag.Remove: %w", err)
+			}
+			return ut, nil
+		}
+
+		// Table: remove tags on all children.
+		m := make(map[string]interface{}, len(v))
+		for k, v2 := range v {
+			var err error
+			m[k], err = rmTag(v2)
+			if err != nil {
+				return nil, err
+			}
+		}
+		return m, nil
+
+	// Array: remove tags from all itenm.
+	case []interface{}:
+		a := make([]interface{}, len(v))
+		for i := range v {
+			var err error
+			a[i], err = rmTag(v[i])
+			if err != nil {
+				return nil, err
+			}
+		}
+		return a, nil
+	}
+
+	// The top level must be an object or array.
+	return nil, fmt.Errorf("unrecognized JSON format '%T'", typedJson)
+}
+
+// Return a primitive: read the "type" and convert the "value" to that.
+func untag(typed map[string]interface{}) (interface{}, error) {
+	t := typed["type"].(string)
+	v := typed["value"].(string)
+	switch t {
+	case "string":
+		return v, nil
+	case "integer":
+		n, err := strconv.ParseInt(v, 10, 64)
+		if err != nil {
+			return nil, fmt.Errorf("untag: %w", err)
+		}
+		return n, nil
+	case "float":
+		f, err := strconv.ParseFloat(v, 64)
+		if err != nil {
+			return nil, fmt.Errorf("untag: %w", err)
+		}
+		return f, nil
+	case "datetime":
+		return parseTime(v, "2006-01-02T15:04:05.999999999Z07:00", false)
+	case "datetime-local":
+		return parseTime(v, "2006-01-02T15:04:05.999999999", true)
+	case "date-local":
+		return parseTime(v, "2006-01-02", true)
+	case "time-local":
+		return parseTime(v, "15:04:05.999999999", true)
+	case "bool":
+		switch v {
+		case "true":
+			return true, nil
+		case "false":
+			return false, nil
+		}
+		return nil, fmt.Errorf("untag: could not parse %q as a boolean", v)
+	}
+
+	return nil, fmt.Errorf("untag: unrecognized tag type %q", t)
+}
+
+func parseTime(v, format string, local bool) (t time.Time, err error) {
+	if local {
+		t, err = time.ParseInLocation(format, v, time.Local)
+	} else {
+		t, err = time.Parse(format, v)
+	}
+	if err != nil {
+		return time.Time{}, fmt.Errorf("Could not parse %q as a datetime: %w", v, err)
+	}
+	return t, nil
+}

internal/testsuite/testsuite.go

@@ -0,0 +1,50 @@
+// Package testsuite provides helper functions for interoperating with the
+// language-agnostic TOML test suite at github.com/BurntSushi/toml-test.
+package testsuite
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+// Marshal is a helpfer function for calling toml.Marshal
+//
+// Only needed to avoid package import loops.
+func Marshal(v interface{}) ([]byte, error) {
+	return toml.Marshal(v)
+}
+
+// Unmarshal is a helper function for calling toml.Unmarshal.
+//
+// Only needed to avoid package import loops.
+func Unmarshal(data []byte, v interface{}) error {
+	return toml.Unmarshal(data, v)
+}
+
+// ValueToTaggedJSON takes a data structure and returns the tagged JSON
+// representation.
+func ValueToTaggedJSON(doc interface{}) ([]byte, error) {
+	return json.MarshalIndent(addTag("", doc), "", "  ")
+}
+
+// DecodeStdin is a helper function for the toml-test binary interface.  TOML input
+// is read from STDIN and a resulting tagged JSON representation is written to
+// STDOUT.
+func DecodeStdin() error {
+	var decoded map[string]interface{}
+
+	if err := toml.NewDecoder(os.Stdin).Decode(&decoded); err != nil {
+		return fmt.Errorf("Error decoding TOML: %s", err)
+	}
+
+	j := json.NewEncoder(os.Stdout)
+	j.SetIndent("", "  ")
+	if err := j.Encode(addTag("", decoded)); err != nil {
+		return fmt.Errorf("Error encoding JSON: %s", err)
+	}
+
+	return nil
+}

internal/tracker/key.go

@@ -0,0 +1,50 @@
+package tracker
+
+import (
+	"github.com/pelletier/go-toml/v2/internal/ast"
+)
+
+// KeyTracker is a tracker that keeps track of the current Key as the AST is
+// walked.
+type KeyTracker struct {
+	k []string
+}
+
+// UpdateTable sets the state of the tracker with the AST table node.
+func (t *KeyTracker) UpdateTable(node *ast.Node) {
+	t.reset()
+	t.Push(node)
+}
+
+// UpdateArrayTable sets the state of the tracker with the AST array table node.
+func (t *KeyTracker) UpdateArrayTable(node *ast.Node) {
+	t.reset()
+	t.Push(node)
+}
+
+// Push the given key on the stack.
+func (t *KeyTracker) Push(node *ast.Node) {
+	it := node.Key()
+	for it.Next() {
+		t.k = append(t.k, string(it.Node().Data))
+	}
+}
+
+// Pop key from stack.
+func (t *KeyTracker) Pop(node *ast.Node) {
+	it := node.Key()
+	for it.Next() {
+		t.k = t.k[:len(t.k)-1]
+	}
+}
+
+// Key returns the current key
+func (t *KeyTracker) Key() []string {
+	k := make([]string, len(t.k))
+	copy(k, t.k)
+	return k
+}
+
+func (t *KeyTracker) reset() {
+	t.k = t.k[:0]
+}

internal/tracker/seen.go

@@ -0,0 +1,356 @@
+package tracker
+
+import (
+	"bytes"
+	"fmt"
+	"sync"
+
+	"github.com/pelletier/go-toml/v2/internal/ast"
+)
+
+type keyKind uint8
+
+const (
+	invalidKind keyKind = iota
+	valueKind
+	tableKind
+	arrayTableKind
+)
+
+func (k keyKind) String() string {
+	switch k {
+	case invalidKind:
+		return "invalid"
+	case valueKind:
+		return "value"
+	case tableKind:
+		return "table"
+	case arrayTableKind:
+		return "array table"
+	}
+	panic("missing keyKind string mapping")
+}
+
+// SeenTracker tracks which keys have been seen with which TOML type to flag
+// duplicates and mismatches according to the spec.
+//
+// Each node in the visited tree is represented by an entry. Each entry has an
+// identifier, which is provided by a counter. Entries are stored in the array
+// entries. As new nodes are discovered (referenced for the first time in the
+// TOML document), entries are created and appended to the array. An entry
+// points to its parent using its id.
+//
+// To find whether a given key (sequence of []byte) has already been visited,
+// the entries are linearly searched, looking for one with the right name and
+// parent id.
+//
+// Given that all keys appear in the document after their parent, it is
+// guaranteed that all descendants of a node are stored after the node, this
+// speeds up the search process.
+//
+// When encountering [[array tables]], the descendants of that node are removed
+// to allow that branch of the tree to be "rediscovered". To maintain the
+// invariant above, the deletion process needs to keep the order of entries.
+// This results in more copies in that case.
+type SeenTracker struct {
+	entries    []entry
+	currentIdx int
+}
+
+var pool sync.Pool
+
+func (s *SeenTracker) reset() {
+	// Always contains a root element at index 0.
+	s.currentIdx = 0
+	if len(s.entries) == 0 {
+		s.entries = make([]entry, 1, 2)
+	} else {
+		s.entries = s.entries[:1]
+	}
+	s.entries[0].child = -1
+	s.entries[0].next = -1
+}
+
+type entry struct {
+	// Use -1 to indicate no child or no sibling.
+	child int
+	next  int
+
+	name     []byte
+	kind     keyKind
+	explicit bool
+	kv       bool
+}
+
+// Find the index of the child of parentIdx with key k. Returns -1 if
+// it does not exist.
+func (s *SeenTracker) find(parentIdx int, k []byte) int {
+	for i := s.entries[parentIdx].child; i >= 0; i = s.entries[i].next {
+		if bytes.Equal(s.entries[i].name, k) {
+			return i
+		}
+	}
+	return -1
+}
+
+// Remove all descendants of node at position idx.
+func (s *SeenTracker) clear(idx int) {
+	if idx >= len(s.entries) {
+		return
+	}
+
+	for i := s.entries[idx].child; i >= 0; {
+		next := s.entries[i].next
+		n := s.entries[0].next
+		s.entries[0].next = i
+		s.entries[i].next = n
+		s.entries[i].name = nil
+		s.clear(i)
+		i = next
+	}
+
+	s.entries[idx].child = -1
+}
+
+func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit bool, kv bool) int {
+	e := entry{
+		child: -1,
+		next:  s.entries[parentIdx].child,
+
+		name:     name,
+		kind:     kind,
+		explicit: explicit,
+		kv:       kv,
+	}
+	var idx int
+	if s.entries[0].next >= 0 {
+		idx = s.entries[0].next
+		s.entries[0].next = s.entries[idx].next
+		s.entries[idx] = e
+	} else {
+		idx = len(s.entries)
+		s.entries = append(s.entries, e)
+	}
+
+	s.entries[parentIdx].child = idx
+
+	return idx
+}
+
+func (s *SeenTracker) setExplicitFlag(parentIdx int) {
+	for i := s.entries[parentIdx].child; i >= 0; i = s.entries[i].next {
+		if s.entries[i].kv {
+			s.entries[i].explicit = true
+			s.entries[i].kv = false
+		}
+		s.setExplicitFlag(i)
+	}
+}
+
+// CheckExpression takes a top-level node and checks that it does not contain
+// keys that have been seen in previous calls, and validates that types are
+// consistent.
+func (s *SeenTracker) CheckExpression(node *ast.Node) error {
+	if s.entries == nil {
+		s.reset()
+	}
+	switch node.Kind {
+	case ast.KeyValue:
+		return s.checkKeyValue(node)
+	case ast.Table:
+		return s.checkTable(node)
+	case ast.ArrayTable:
+		return s.checkArrayTable(node)
+	default:
+		panic(fmt.Errorf("this should not be a top level node type: %s", node.Kind))
+	}
+}
+
+func (s *SeenTracker) checkTable(node *ast.Node) error {
+	if s.currentIdx >= 0 {
+		s.setExplicitFlag(s.currentIdx)
+	}
+
+	it := node.Key()
+
+	parentIdx := 0
+
+	// This code is duplicated in checkArrayTable. This is because factoring
+	// it in a function requires to copy the iterator, or allocate it to the
+	// heap, which is not cheap.
+	for it.Next() {
+		if it.IsLast() {
+			break
+		}
+
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false, false)
+		} else {
+			entry := s.entries[idx]
+			if entry.kind == valueKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			}
+		}
+		parentIdx = idx
+	}
+
+	k := it.Node().Data
+	idx := s.find(parentIdx, k)
+
+	if idx >= 0 {
+		kind := s.entries[idx].kind
+		if kind != tableKind {
+			return fmt.Errorf("toml: key %s should be a table, not a %s", string(k), kind)
+		}
+		if s.entries[idx].explicit {
+			return fmt.Errorf("toml: table %s already exists", string(k))
+		}
+		s.entries[idx].explicit = true
+	} else {
+		idx = s.create(parentIdx, k, tableKind, true, false)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
+	if s.currentIdx >= 0 {
+		s.setExplicitFlag(s.currentIdx)
+	}
+
+	it := node.Key()
+
+	parentIdx := 0
+
+	for it.Next() {
+		if it.IsLast() {
+			break
+		}
+
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false, false)
+		} else {
+			entry := s.entries[idx]
+			if entry.kind == valueKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			}
+		}
+
+		parentIdx = idx
+	}
+
+	k := it.Node().Data
+	idx := s.find(parentIdx, k)
+
+	if idx >= 0 {
+		kind := s.entries[idx].kind
+		if kind != arrayTableKind {
+			return fmt.Errorf("toml: key %s already exists as a %s,  but should be an array table", kind, string(k))
+		}
+		s.clear(idx)
+	} else {
+		idx = s.create(parentIdx, k, arrayTableKind, true, false)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkKeyValue(node *ast.Node) error {
+	parentIdx := s.currentIdx
+	it := node.Key()
+
+	for it.Next() {
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false, true)
+		} else {
+			entry := s.entries[idx]
+			if it.IsLast() {
+				return fmt.Errorf("toml: key %s is already defined", string(k))
+			} else if entry.kind != tableKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			} else if entry.explicit {
+				return fmt.Errorf("toml: cannot redefine table %s that has already been explicitly defined", string(k))
+			}
+		}
+
+		parentIdx = idx
+	}
+
+	s.entries[parentIdx].kind = valueKind
+
+	value := node.Value()
+
+	switch value.Kind {
+	case ast.InlineTable:
+		return s.checkInlineTable(value)
+	case ast.Array:
+		return s.checkArray(value)
+	}
+
+	return nil
+}
+
+func (s *SeenTracker) checkArray(node *ast.Node) error {
+	it := node.Children()
+	for it.Next() {
+		n := it.Node()
+		switch n.Kind {
+		case ast.InlineTable:
+			err := s.checkInlineTable(n)
+			if err != nil {
+				return err
+			}
+		case ast.Array:
+			err := s.checkArray(n)
+			if err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
+func (s *SeenTracker) checkInlineTable(node *ast.Node) error {
+	if pool.New == nil {
+		pool.New = func() interface{} {
+			return &SeenTracker{}
+		}
+	}
+
+	s = pool.Get().(*SeenTracker)
+	s.reset()
+
+	it := node.Children()
+	for it.Next() {
+		n := it.Node()
+		err := s.checkKeyValue(n)
+		if err != nil {
+			return err
+		}
+	}
+
+	// As inline tables are self-contained, the tracker does not
+	// need to retain the details of what they contain. The
+	// keyValue element that creates the inline table is kept to
+	// mark the presence of the inline table and prevent
+	// redefinition of its keys: check* functions cannot walk into
+	// a value.
+	pool.Put(s)
+	return nil
+}

internal/tracker/seen_test.go

@@ -0,0 +1,15 @@
+package tracker
+
+import (
+	"testing"
+	"unsafe"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestEntrySize(t *testing.T) {
+	// Validate no regression on the size of entry{}. This is a critical bit for
+	// performance of unmarshaling documents. Should only be increased with care
+	// and a very good reason.
+	require.LessOrEqual(t, 48, int(unsafe.Sizeof(entry{})))
+}

internal/tracker/tracker.go

@@ -1,200 +1 @@
 package tracker
-
-import (
-	"fmt"
-
-	"github.com/pelletier/go-toml/v2/internal/ast"
-)
-
-type keyKind uint8
-
-const (
-	invalidKind keyKind = iota
-	valueKind
-	tableKind
-	arrayTableKind
-)
-
-func (k keyKind) String() string {
-	switch k {
-	case invalidKind:
-		return "invalid"
-	case valueKind:
-		return "value"
-	case tableKind:
-		return "table"
-	case arrayTableKind:
-		return "array table"
-	}
-	panic("missing keyKind string mapping")
-}
-
-// Tracks which keys have been seen with which TOML type to flag duplicates
-// and mismatches according to the spec.
-type Seen struct {
-	root    *info
-	current *info
-}
-
-type info struct {
-	parent   *info
-	kind     keyKind
-	children map[string]*info
-	explicit bool
-}
-
-func (i *info) Clear() {
-	i.children = nil
-}
-
-func (i *info) Has(k string) (*info, bool) {
-	c, ok := i.children[k]
-	return c, ok
-}
-
-func (i *info) SetKind(kind keyKind) {
-	i.kind = kind
-}
-
-func (i *info) CreateTable(k string, explicit bool) *info {
-	return i.createChild(k, tableKind, explicit)
-}
-
-func (i *info) CreateArrayTable(k string, explicit bool) *info {
-	return i.createChild(k, arrayTableKind, explicit)
-}
-
-func (i *info) createChild(k string, kind keyKind, explicit bool) *info {
-	if i.children == nil {
-		i.children = make(map[string]*info, 1)
-	}
-
-	x := &info{
-		parent:   i,
-		kind:     kind,
-		explicit: explicit,
-	}
-	i.children[k] = x
-	return x
-}
-
-// CheckExpression takes a top-level node and checks that it does not contain keys
-// that have been seen in previous calls, and validates that types are consistent.
-func (s *Seen) CheckExpression(node ast.Node) error {
-	if s.root == nil {
-		s.root = &info{
-			kind: tableKind,
-		}
-		s.current = s.root
-	}
-	switch node.Kind {
-	case ast.KeyValue:
-		return s.checkKeyValue(s.current, node)
-	case ast.Table:
-		return s.checkTable(node)
-	case ast.ArrayTable:
-		return s.checkArrayTable(node)
-	default:
-		panic(fmt.Errorf("this should not be a top level node type: %s", node.Kind))
-	}
-
-}
-func (s *Seen) checkTable(node ast.Node) error {
-	s.current = s.root
-
-	it := node.Key()
-	// handle the first parts of the key, excluding the last one
-	for it.Next() {
-		if !it.Node().Next().Valid() {
-			break
-		}
-
-		k := string(it.Node().Data)
-		child, found := s.current.Has(k)
-		if !found {
-			child = s.current.CreateTable(k, false)
-		}
-		s.current = child
-	}
-
-	// handle the last part of the key
-	k := string(it.Node().Data)
-
-	i, found := s.current.Has(k)
-	if found {
-		if i.kind != tableKind {
-			return fmt.Errorf("key %s should be a table", k)
-		}
-		if i.explicit {
-			return fmt.Errorf("table %s already exists", k)
-		}
-		i.explicit = true
-		s.current = i
-	} else {
-		s.current = s.current.CreateTable(k, true)
-	}
-
-	return nil
-}
-
-func (s *Seen) checkArrayTable(node ast.Node) error {
-	s.current = s.root
-
-	it := node.Key()
-
-	// handle the first parts of the key, excluding the last one
-	for it.Next() {
-		if !it.Node().Next().Valid() {
-			break
-		}
-
-		k := string(it.Node().Data)
-		child, found := s.current.Has(k)
-		if !found {
-			child = s.current.CreateTable(k, false)
-		}
-		s.current = child
-	}
-
-	// handle the last part of the key
-	k := string(it.Node().Data)
-
-	info, found := s.current.Has(k)
-	if found {
-		if info.kind != arrayTableKind {
-			return fmt.Errorf("key %s already exists but is not an array table", k)
-		}
-		info.Clear()
-	} else {
-		info = s.current.CreateArrayTable(k, true)
-	}
-
-	s.current = info
-	return nil
-}
-
-func (s *Seen) checkKeyValue(context *info, node ast.Node) error {
-	it := node.Key()
-
-	// handle the first parts of the key, excluding the last one
-	for it.Next() {
-		k := string(it.Node().Data)
-		child, found := context.Has(k)
-		if found {
-			if child.kind != tableKind {
-				return fmt.Errorf("expected %s to be a table, not a %s", k, child.kind)
-			}
-		} else {
-			child = context.CreateTable(k, false)
-		}
-		context = child
-	}
-
-	if node.Value().Kind == ast.InlineTable {
-		context.SetKind(tableKind)
-	} else {
-		context.SetKind(valueKind)
-	}
-
-	return nil
-}

internal/unsafe/unsafe.go

@@ -1,35 +0,0 @@
-package unsafe
-
-import (
-	"fmt"
-	"reflect"
-	"unsafe"
-)
-
-const maxInt = uintptr(int(^uint(0) >> 1))
-
-func SubsliceOffset(data []byte, subslice []byte) int {
-	datap := (*reflect.SliceHeader)(unsafe.Pointer(&data))
-	hlp := (*reflect.SliceHeader)(unsafe.Pointer(&subslice))
-
-	if hlp.Data < datap.Data {
-		panic(fmt.Errorf("subslice address (%d) is before data address (%d)", hlp.Data, datap.Data))
-	}
-	offset := hlp.Data - datap.Data
-
-	if offset > maxInt {
-		panic(fmt.Errorf("slice offset larger than int (%d)", offset))
-	}
-
-	intoffset := int(offset)
-
-	if intoffset > datap.Len {
-		panic(fmt.Errorf("slice offset (%d) is farther than data length (%d)", intoffset, datap.Len))
-	}
-
-	if intoffset+hlp.Len > datap.Len {
-		panic(fmt.Errorf("slice ends (%d+%d) is farther than data length (%d)", intoffset, hlp.Len, datap.Len))
-	}
-
-	return intoffset
-}

internal/unsafe/unsafe_test.go

@@ -1,79 +0,0 @@
-package unsafe_test
-
-import (
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-
-	"github.com/pelletier/go-toml/v2/internal/unsafe"
-)
-
-func TestUnsafeSubsliceOffsetValid(t *testing.T) {
-	examples := []struct {
-		desc   string
-		test   func() ([]byte, []byte)
-		offset int
-	}{
-		{
-			desc: "simple",
-			test: func() ([]byte, []byte) {
-				data := []byte("hello")
-				return data, data[1:]
-			},
-			offset: 1,
-		},
-	}
-
-	for _, e := range examples {
-		t.Run(e.desc, func(t *testing.T) {
-			d, s := e.test()
-			offset := unsafe.SubsliceOffset(d, s)
-			assert.Equal(t, e.offset, offset)
-		})
-	}
-}
-
-func TestUnsafeSubsliceOffsetInvalid(t *testing.T) {
-	examples := []struct {
-		desc string
-		test func() ([]byte, []byte)
-	}{
-		{
-			desc: "unrelated arrays",
-			test: func() ([]byte, []byte) {
-				return []byte("one"), []byte("two")
-			},
-		},
-		{
-			desc: "slice starts before data",
-			test: func() ([]byte, []byte) {
-				full := []byte("hello world")
-				return full[5:], full[1:]
-			},
-		},
-		{
-			desc: "slice starts after data",
-			test: func() ([]byte, []byte) {
-				full := []byte("hello world")
-				return full[:3], full[5:]
-			},
-		},
-		{
-			desc: "slice ends after data",
-			test: func() ([]byte, []byte) {
-				full := []byte("hello world")
-				return full[:5], full[3:8]
-			},
-		},
-	}
-
-	for _, e := range examples {
-		t.Run(e.desc, func(t *testing.T) {
-			d, s := e.test()
-			require.Panics(t, func() {
-				unsafe.SubsliceOffset(d, s)
-			})
-		})
-	}
-}

localtime.go

@@ -1,281 +1,120 @@
-// Implementation of TOML's local date/time.
-// Copied over from https://github.com/googleapis/google-cloud-go/blob/master/civil/civil.go
-// to avoid pulling all the Google dependencies.
-//
-// Copyright 2016 Google LLC
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-// Package civil implements types for civil time, a time-zone-independent
-// representation of time that follows the rules of the proleptic
-// Gregorian calendar with exactly 24-hour days, 60-minute hours, and 60-second
-// minutes.
-//
-// Because they lack location information, these types do not represent unique
-// moments or intervals of time. Use time.Time for that purpose.
 package toml
 
 import (
 	"fmt"
+	"strings"
 	"time"
 )
 
-// A LocalDate represents a date (year, month, day).
-//
-// This type does not include location information, and therefore does not
-// describe a unique 24-hour timespan.
+// LocalDate represents a calendar day in no specific timezone.
 type LocalDate struct {
-	Year  int        // Year (e.g., 2014).
-	Month time.Month // Month of the year (January = 1, ...).
-	Day   int        // Day of the month, starting at 1.
+	Year  int
+	Month int
+	Day   int
 }
 
-// LocalDateOf returns the LocalDate in which a time occurs in that time's location.
-func LocalDateOf(t time.Time) LocalDate {
-	var d LocalDate
-	d.Year, d.Month, d.Day = t.Date()
-	return d
+// AsTime converts d into a specific time instance at midnight in zone.
+func (d LocalDate) AsTime(zone *time.Location) time.Time {
+	return time.Date(d.Year, time.Month(d.Month), d.Day, 0, 0, 0, 0, zone)
 }
 
-// ParseLocalDate parses a string in RFC3339 full-date format and returns the date value it represents.
-func ParseLocalDate(s string) (LocalDate, error) {
-	t, err := time.Parse("2006-01-02", s)
-	if err != nil {
-		return LocalDate{}, err
-	}
-	return LocalDateOf(t), nil
-}
-
-// String returns the date in RFC3339 full-date format.
+// String returns RFC 3339 representation of d.
 func (d LocalDate) String() string {
 	return fmt.Sprintf("%04d-%02d-%02d", d.Year, d.Month, d.Day)
 }
 
-// IsValid reports whether the date is valid.
-func (d LocalDate) IsValid() bool {
-	return LocalDateOf(d.In(time.UTC)) == d
-}
-
-// In returns the time corresponding to time 00:00:00 of the date in the location.
-//
-// In is always consistent with time.LocalDate, even when time.LocalDate returns a time
-// on a different day. For example, if loc is America/Indiana/Vincennes, then both
-//     time.LocalDate(1955, time.May, 1, 0, 0, 0, 0, loc)
-// and
-//     civil.LocalDate{Year: 1955, Month: time.May, Day: 1}.In(loc)
-// return 23:00:00 on April 30, 1955.
-//
-// In panics if loc is nil.
-func (d LocalDate) In(loc *time.Location) time.Time {
-	return time.Date(d.Year, d.Month, d.Day, 0, 0, 0, 0, loc)
-}
-
-// AddDays returns the date that is n days in the future.
-// n can also be negative to go into the past.
-func (d LocalDate) AddDays(n int) LocalDate {
-	return LocalDateOf(d.In(time.UTC).AddDate(0, 0, n))
-}
-
-// DaysSince returns the signed number of days between the date and s, not including the end day.
-// This is the inverse operation to AddDays.
-func (d LocalDate) DaysSince(s LocalDate) (days int) {
-	// We convert to Unix time so we do not have to worry about leap seconds:
-	// Unix time increases by exactly 86400 seconds per day.
-	deltaUnix := d.In(time.UTC).Unix() - s.In(time.UTC).Unix()
-	return int(deltaUnix / 86400)
-}
-
-// Before reports whether d1 occurs before d2.
-func (d1 LocalDate) Before(d2 LocalDate) bool {
-	if d1.Year != d2.Year {
-		return d1.Year < d2.Year
-	}
-	if d1.Month != d2.Month {
-		return d1.Month < d2.Month
-	}
-	return d1.Day < d2.Day
-}
-
-// After reports whether d1 occurs after d2.
-func (d1 LocalDate) After(d2 LocalDate) bool {
-	return d2.Before(d1)
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of d.String().
+// MarshalText returns RFC 3339 representation of d.
 func (d LocalDate) MarshalText() ([]byte, error) {
 	return []byte(d.String()), nil
 }
 
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
-// The date is expected to be a string in a format accepted by ParseLocalDate.
-func (d *LocalDate) UnmarshalText(data []byte) error {
-	var err error
-	*d, err = ParseLocalDate(string(data))
-	return err
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalDate) UnmarshalText(b []byte) error {
+	res, err := parseLocalDate(b)
+	if err != nil {
+		return err
+	}
+	*d = res
+	return nil
 }
 
-// A LocalTime represents a time with nanosecond precision.
-//
-// This type does not include location information, and therefore does not
-// describe a unique moment in time.
-//
-// This type exists to represent the TIME type in storage-based APIs like BigQuery.
-// Most operations on Times are unlikely to be meaningful. Prefer the LocalDateTime type.
+// LocalTime represents a time of day of no specific day in no specific
+// timezone.
 type LocalTime struct {
-	Hour       int // The hour of the day in 24-hour format; range [0-23]
-	Minute     int // The minute of the hour; range [0-59]
-	Second     int // The second of the minute; range [0-59]
-	Nanosecond int // The nanosecond of the second; range [0-999999999]
+	Hour       int // Hour of the day: [0; 24[
+	Minute     int // Minute of the hour: [0; 60[
+	Second     int // Second of the minute: [0; 60[
+	Nanosecond int // Nanoseconds within the second:  [0, 1000000000[
+	Precision  int // Number of digits to display for Nanosecond.
 }
 
-// LocalTimeOf returns the LocalTime representing the time of day in which a time occurs
-// in that time's location. It ignores the date.
-func LocalTimeOf(t time.Time) LocalTime {
-	var tm LocalTime
-	tm.Hour, tm.Minute, tm.Second = t.Clock()
-	tm.Nanosecond = t.Nanosecond()
-	return tm
+// String returns RFC 3339 representation of d.
+// If d.Nanosecond and d.Precision are zero, the time won't have a nanosecond
+// component. If d.Nanosecond > 0 but d.Precision = 0, then the minimum number
+// of digits for nanoseconds is provided.
+func (d LocalTime) String() string {
+	s := fmt.Sprintf("%02d:%02d:%02d", d.Hour, d.Minute, d.Second)
+
+	if d.Precision > 0 {
+		s += fmt.Sprintf(".%09d", d.Nanosecond)[:d.Precision+1]
+	} else if d.Nanosecond > 0 {
+		// Nanoseconds are specified, but precision is not provided. Use the
+		// minimum.
+		s += strings.Trim(fmt.Sprintf(".%09d", d.Nanosecond), "0")
+	}
+
+	return s
 }
 
-// ParseLocalTime parses a string and returns the time value it represents.
-// ParseLocalTime accepts an extended form of the RFC3339 partial-time format. After
-// the HH:MM:SS part of the string, an optional fractional part may appear,
-// consisting of a decimal point followed by one to nine decimal digits.
-// (RFC3339 admits only one digit after the decimal point).
-func ParseLocalTime(s string) (LocalTime, error) {
-	t, err := time.Parse("15:04:05.999999999", s)
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalTime) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalTime) UnmarshalText(b []byte) error {
+	res, left, err := parseLocalTime(b)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
+	}
 	if err != nil {
-		return LocalTime{}, err
+		return err
 	}
-	return LocalTimeOf(t), nil
+	*d = res
+	return nil
 }
 
-// String returns the date in the format described in ParseLocalTime. If Nanoseconds
-// is zero, no fractional part will be generated. Otherwise, the result will
-// end with a fractional part consisting of a decimal point and nine digits.
-func (t LocalTime) String() string {
-	s := fmt.Sprintf("%02d:%02d:%02d", t.Hour, t.Minute, t.Second)
-	if t.Nanosecond == 0 {
-		return s
-	}
-	return s + fmt.Sprintf(".%09d", t.Nanosecond)
-}
-
-// IsValid reports whether the time is valid.
-func (t LocalTime) IsValid() bool {
-	// Construct a non-zero time.
-	tm := time.Date(2, 2, 2, t.Hour, t.Minute, t.Second, t.Nanosecond, time.UTC)
-	return LocalTimeOf(tm) == t
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of t.String().
-func (t LocalTime) MarshalText() ([]byte, error) {
-	return []byte(t.String()), nil
-}
-
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
-// The time is expected to be a string in a format accepted by ParseLocalTime.
-func (t *LocalTime) UnmarshalText(data []byte) error {
-	var err error
-	*t, err = ParseLocalTime(string(data))
-	return err
-}
-
-// A LocalDateTime represents a date and time.
-//
-// This type does not include location information, and therefore does not
-// describe a unique moment in time.
+// LocalDateTime represents a time of a specific day in no specific timezone.
 type LocalDateTime struct {
-	Date LocalDate
-	Time LocalTime
+	LocalDate
+	LocalTime
 }
 
-// Note: We deliberately do not embed LocalDate into LocalDateTime, to avoid promoting AddDays and Sub.
+// AsTime converts d into a specific time instance in zone.
+func (d LocalDateTime) AsTime(zone *time.Location) time.Time {
+	return time.Date(d.Year, time.Month(d.Month), d.Day, d.Hour, d.Minute, d.Second, d.Nanosecond, zone)
+}
 
-// LocalDateTimeOf returns the LocalDateTime in which a time occurs in that time's location.
-func LocalDateTimeOf(t time.Time) LocalDateTime {
-	return LocalDateTime{
-		Date: LocalDateOf(t),
-		Time: LocalTimeOf(t),
+// String returns RFC 3339 representation of d.
+func (d LocalDateTime) String() string {
+	return d.LocalDate.String() + "T" + d.LocalTime.String()
+}
+
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalDateTime) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalDateTime) UnmarshalText(data []byte) error {
+	res, left, err := parseLocalDateTime(data)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
 	}
-}
-
-// ParseLocalDateTime parses a string and returns the LocalDateTime it represents.
-// ParseLocalDateTime accepts a variant of the RFC3339 date-time format that omits
-// the time offset but includes an optional fractional time, as described in
-// ParseLocalTime. Informally, the accepted format is
-//     YYYY-MM-DDTHH:MM:SS[.FFFFFFFFF]
-// where the 'T' may be a lower-case 't'.
-func ParseLocalDateTime(s string) (LocalDateTime, error) {
-	t, err := time.Parse("2006-01-02T15:04:05.999999999", s)
 	if err != nil {
-		t, err = time.Parse("2006-01-02t15:04:05.999999999", s)
-		if err != nil {
-			return LocalDateTime{}, err
-		}
+		return err
 	}
-	return LocalDateTimeOf(t), nil
-}
 
-// String returns the date in the format described in ParseLocalDate.
-func (dt LocalDateTime) String() string {
-	return dt.Date.String() + "T" + dt.Time.String()
-}
-
-// IsValid reports whether the datetime is valid.
-func (dt LocalDateTime) IsValid() bool {
-	return dt.Date.IsValid() && dt.Time.IsValid()
-}
-
-// In returns the time corresponding to the LocalDateTime in the given location.
-//
-// If the time is missing or ambigous at the location, In returns the same
-// result as time.LocalDate. For example, if loc is America/Indiana/Vincennes, then
-// both
-//     time.LocalDate(1955, time.May, 1, 0, 30, 0, 0, loc)
-// and
-//     civil.LocalDateTime{
-//         civil.LocalDate{Year: 1955, Month: time.May, Day: 1}},
-//         civil.LocalTime{Minute: 30}}.In(loc)
-// return 23:30:00 on April 30, 1955.
-//
-// In panics if loc is nil.
-func (dt LocalDateTime) In(loc *time.Location) time.Time {
-	return time.Date(dt.Date.Year, dt.Date.Month, dt.Date.Day, dt.Time.Hour, dt.Time.Minute, dt.Time.Second, dt.Time.Nanosecond, loc)
-}
-
-// Before reports whether dt1 occurs before dt2.
-func (dt1 LocalDateTime) Before(dt2 LocalDateTime) bool {
-	return dt1.In(time.UTC).Before(dt2.In(time.UTC))
-}
-
-// After reports whether dt1 occurs after dt2.
-func (dt1 LocalDateTime) After(dt2 LocalDateTime) bool {
-	return dt2.Before(dt1)
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of dt.String().
-func (dt LocalDateTime) MarshalText() ([]byte, error) {
-	return []byte(dt.String()), nil
-}
-
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
-// The datetime is expected to be a string in a format accepted by ParseLocalDateTime
-func (dt *LocalDateTime) UnmarshalText(data []byte) error {
-	var err error
-	*dt, err = ParseLocalDateTime(string(data))
-	return err
+	*d = res
+	return nil
 }

localtime_test.go

@@ -1,446 +1,118 @@
-// Copyright 2016 Google LLC
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-package toml
+package toml_test
 
 import (
-	"encoding/json"
-	"reflect"
 	"testing"
 	"time"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
 )
 
-func cmpEqual(x, y interface{}) bool {
-	return reflect.DeepEqual(x, y)
+func TestLocalDate_AsTime(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 0, 0, 0, 0, time.UTC), cast)
 }
 
-func TestDates(t *testing.T) {
-	for _, test := range []struct {
-		date     LocalDate
-		loc      *time.Location
-		wantStr  string
-		wantTime time.Time
-	}{
-		{
-			date:     LocalDate{2014, 7, 29},
-			loc:      time.Local,
-			wantStr:  "2014-07-29",
-			wantTime: time.Date(2014, time.July, 29, 0, 0, 0, 0, time.Local),
-		},
-		{
-			date:     LocalDateOf(time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local)),
-			loc:      time.UTC,
-			wantStr:  "2014-08-20",
-			wantTime: time.Date(2014, 8, 20, 0, 0, 0, 0, time.UTC),
-		},
-		{
-			date:     LocalDateOf(time.Date(999, time.January, 26, 0, 0, 0, 0, time.Local)),
-			loc:      time.UTC,
-			wantStr:  "0999-01-26",
-			wantTime: time.Date(999, 1, 26, 0, 0, 0, 0, time.UTC),
-		},
-	} {
-		if got := test.date.String(); got != test.wantStr {
-			t.Errorf("%#v.String() = %q, want %q", test.date, got, test.wantStr)
-		}
-		if got := test.date.In(test.loc); !got.Equal(test.wantTime) {
-			t.Errorf("%#v.In(%v) = %v, want %v", test.date, test.loc, got, test.wantTime)
-		}
-	}
+func TestLocalDate_String(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	require.Equal(t, "2021-06-08", d.String())
 }
 
-func TestDateIsValid(t *testing.T) {
-	for _, test := range []struct {
-		date LocalDate
-		want bool
-	}{
-		{LocalDate{2014, 7, 29}, true},
-		{LocalDate{2000, 2, 29}, true},
-		{LocalDate{10000, 12, 31}, true},
-		{LocalDate{1, 1, 1}, true},
-		{LocalDate{0, 1, 1}, true},  // year zero is OK
-		{LocalDate{-1, 1, 1}, true}, // negative year is OK
-		{LocalDate{1, 0, 1}, false},
-		{LocalDate{1, 1, 0}, false},
-		{LocalDate{2016, 1, 32}, false},
-		{LocalDate{2016, 13, 1}, false},
-		{LocalDate{1, -1, 1}, false},
-		{LocalDate{1, 1, -1}, false},
-	} {
-		got := test.date.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.date, got, test.want)
-		}
-	}
+func TestLocalDate_MarshalText(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("2021-06-08"), b)
 }
 
-func TestParseDate(t *testing.T) {
-	for _, test := range []struct {
-		str  string
-		want LocalDate // if empty, expect an error
-	}{
-		{"2016-01-02", LocalDate{2016, 1, 2}},
-		{"2016-12-31", LocalDate{2016, 12, 31}},
-		{"0003-02-04", LocalDate{3, 2, 4}},
-		{"999-01-26", LocalDate{}},
-		{"", LocalDate{}},
-		{"2016-01-02x", LocalDate{}},
-	} {
-		got, err := ParseLocalDate(test.str)
-		if got != test.want {
-			t.Errorf("ParseLocalDate(%q) = %+v, want %+v", test.str, got, test.want)
-		}
-		if err != nil && test.want != (LocalDate{}) {
-			t.Errorf("Unexpected error %v from ParseLocalDate(%q)", err, test.str)
-		}
-	}
+func TestLocalDate_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalDate{}
+	err := d.UnmarshalText([]byte("2021-06-08"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalDate{2021, 6, 8}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
 }
 
-func TestDateArithmetic(t *testing.T) {
-	for _, test := range []struct {
-		desc  string
-		start LocalDate
-		end   LocalDate
-		days  int
-	}{
-		{
-			desc:  "zero days noop",
-			start: LocalDate{2014, 5, 9},
-			end:   LocalDate{2014, 5, 9},
-			days:  0,
-		},
-		{
-			desc:  "crossing a year boundary",
-			start: LocalDate{2014, 12, 31},
-			end:   LocalDate{2015, 1, 1},
-			days:  1,
-		},
-		{
-			desc:  "negative number of days",
-			start: LocalDate{2015, 1, 1},
-			end:   LocalDate{2014, 12, 31},
-			days:  -1,
-		},
-		{
-			desc:  "full leap year",
-			start: LocalDate{2004, 1, 1},
-			end:   LocalDate{2005, 1, 1},
-			days:  366,
-		},
-		{
-			desc:  "full non-leap year",
-			start: LocalDate{2001, 1, 1},
-			end:   LocalDate{2002, 1, 1},
-			days:  365,
-		},
-		{
-			desc:  "crossing a leap second",
-			start: LocalDate{1972, 6, 30},
-			end:   LocalDate{1972, 7, 1},
-			days:  1,
-		},
-		{
-			desc:  "dates before the unix epoch",
-			start: LocalDate{101, 1, 1},
-			end:   LocalDate{102, 1, 1},
-			days:  365,
-		},
-	} {
-		if got := test.start.AddDays(test.days); got != test.end {
-			t.Errorf("[%s] %#v.AddDays(%v) = %#v, want %#v", test.desc, test.start, test.days, got, test.end)
-		}
-		if got := test.end.DaysSince(test.start); got != test.days {
-			t.Errorf("[%s] %#v.Sub(%#v) = %v, want %v", test.desc, test.end, test.start, got, test.days)
-		}
-	}
+func TestLocalTime_String(t *testing.T) {
+	d := toml.LocalTime{20, 12, 1, 2, 9}
+	require.Equal(t, "20:12:01.000000002", d.String())
+	d = toml.LocalTime{20, 12, 1, 0, 0}
+	require.Equal(t, "20:12:01", d.String())
+	d = toml.LocalTime{20, 12, 1, 0, 9}
+	require.Equal(t, "20:12:01.000000000", d.String())
+	d = toml.LocalTime{20, 12, 1, 100, 0}
+	require.Equal(t, "20:12:01.0000001", d.String())
 }
 
-func TestDateBefore(t *testing.T) {
-	for _, test := range []struct {
-		d1, d2 LocalDate
-		want   bool
-	}{
-		{LocalDate{2016, 12, 31}, LocalDate{2017, 1, 1}, true},
-		{LocalDate{2016, 1, 1}, LocalDate{2016, 1, 1}, false},
-		{LocalDate{2016, 12, 30}, LocalDate{2016, 12, 31}, true},
-		{LocalDate{2016, 1, 30}, LocalDate{2016, 12, 31}, true},
-	} {
-		if got := test.d1.Before(test.d2); got != test.want {
-			t.Errorf("%v.Before(%v): got %t, want %t", test.d1, test.d2, got, test.want)
-		}
-	}
+func TestLocalTime_MarshalText(t *testing.T) {
+	d := toml.LocalTime{20, 12, 1, 2, 9}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("20:12:01.000000002"), b)
 }
 
-func TestDateAfter(t *testing.T) {
-	for _, test := range []struct {
-		d1, d2 LocalDate
-		want   bool
-	}{
-		{LocalDate{2016, 12, 31}, LocalDate{2017, 1, 1}, false},
-		{LocalDate{2016, 1, 1}, LocalDate{2016, 1, 1}, false},
-		{LocalDate{2016, 12, 30}, LocalDate{2016, 12, 31}, false},
-	} {
-		if got := test.d1.After(test.d2); got != test.want {
-			t.Errorf("%v.After(%v): got %t, want %t", test.d1, test.d2, got, test.want)
-		}
-	}
+func TestLocalTime_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalTime{}
+	err := d.UnmarshalText([]byte("20:12:01.000000002"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalTime{20, 12, 1, 2, 9}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
+
+	err = d.UnmarshalText([]byte("20:12:01.000000002 bad"))
+	require.Error(t, err)
 }
 
-func TestTimeToString(t *testing.T) {
-	for _, test := range []struct {
-		str       string
-		time      LocalTime
-		roundTrip bool // ParseLocalTime(str).String() == str?
-	}{
-		{"13:26:33", LocalTime{13, 26, 33, 0}, true},
-		{"01:02:03.000023456", LocalTime{1, 2, 3, 23456}, true},
-		{"00:00:00.000000001", LocalTime{0, 0, 0, 1}, true},
-		{"13:26:03.1", LocalTime{13, 26, 3, 100000000}, false},
-		{"13:26:33.0000003", LocalTime{13, 26, 33, 300}, false},
-	} {
-		gotTime, err := ParseLocalTime(test.str)
-		if err != nil {
-			t.Errorf("ParseLocalTime(%q): got error: %v", test.str, err)
-			continue
-		}
-		if gotTime != test.time {
-			t.Errorf("ParseLocalTime(%q) = %+v, want %+v", test.str, gotTime, test.time)
-		}
-		if test.roundTrip {
-			gotStr := test.time.String()
-			if gotStr != test.str {
-				t.Errorf("%#v.String() = %q, want %q", test.time, gotStr, test.str)
-			}
-		}
-	}
+func TestLocalTime_RoundTrip(t *testing.T) {
+	var d struct{ A toml.LocalTime }
+	err := toml.Unmarshal([]byte("a=20:12:01.500"), &d)
+	require.NoError(t, err)
+	require.Equal(t, "20:12:01.500", d.A.String())
 }
 
-func TestTimeOf(t *testing.T) {
-	for _, test := range []struct {
-		time time.Time
-		want LocalTime
-	}{
-		{time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local), LocalTime{15, 8, 43, 1}},
-		{time.Date(1, 1, 1, 0, 0, 0, 0, time.UTC), LocalTime{0, 0, 0, 0}},
-	} {
-		if got := LocalTimeOf(test.time); got != test.want {
-			t.Errorf("LocalTimeOf(%v) = %+v, want %+v", test.time, got, test.want)
-		}
+func TestLocalDateTime_AsTime(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
 	}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 20, 12, 1, 2, time.UTC), cast)
 }
 
-func TestTimeIsValid(t *testing.T) {
-	for _, test := range []struct {
-		time LocalTime
-		want bool
-	}{
-		{LocalTime{0, 0, 0, 0}, true},
-		{LocalTime{23, 0, 0, 0}, true},
-		{LocalTime{23, 59, 59, 999999999}, true},
-		{LocalTime{24, 59, 59, 999999999}, false},
-		{LocalTime{23, 60, 59, 999999999}, false},
-		{LocalTime{23, 59, 60, 999999999}, false},
-		{LocalTime{23, 59, 59, 1000000000}, false},
-		{LocalTime{-1, 0, 0, 0}, false},
-		{LocalTime{0, -1, 0, 0}, false},
-		{LocalTime{0, 0, -1, 0}, false},
-		{LocalTime{0, 0, 0, -1}, false},
-	} {
-		got := test.time.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.time, got, test.want)
-		}
+func TestLocalDateTime_String(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
 	}
+	require.Equal(t, "2021-06-08T20:12:01.000000002", d.String())
 }
 
-func TestDateTimeToString(t *testing.T) {
-	for _, test := range []struct {
-		str       string
-		dateTime  LocalDateTime
-		roundTrip bool // ParseLocalDateTime(str).String() == str?
-	}{
-		{"2016-03-22T13:26:33", LocalDateTime{LocalDate{2016, 03, 22}, LocalTime{13, 26, 33, 0}}, true},
-		{"2016-03-22T13:26:33.000000600", LocalDateTime{LocalDate{2016, 03, 22}, LocalTime{13, 26, 33, 600}}, true},
-		{"2016-03-22t13:26:33", LocalDateTime{LocalDate{2016, 03, 22}, LocalTime{13, 26, 33, 0}}, false},
-	} {
-		gotDateTime, err := ParseLocalDateTime(test.str)
-		if err != nil {
-			t.Errorf("ParseLocalDateTime(%q): got error: %v", test.str, err)
-			continue
-		}
-		if gotDateTime != test.dateTime {
-			t.Errorf("ParseLocalDateTime(%q) = %+v, want %+v", test.str, gotDateTime, test.dateTime)
-		}
-		if test.roundTrip {
-			gotStr := test.dateTime.String()
-			if gotStr != test.str {
-				t.Errorf("%#v.String() = %q, want %q", test.dateTime, gotStr, test.str)
-			}
-		}
+func TestLocalDateTime_MarshalText(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
 	}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("2021-06-08T20:12:01.000000002"), b)
 }
 
-func TestParseDateTimeErrors(t *testing.T) {
-	for _, str := range []string{
-		"",
-		"2016-03-22",           // just a date
-		"13:26:33",             // just a time
-		"2016-03-22 13:26:33",  // wrong separating character
-		"2016-03-22T13:26:33x", // extra at end
-	} {
-		if _, err := ParseLocalDateTime(str); err == nil {
-			t.Errorf("ParseLocalDateTime(%q) succeeded, want error", str)
-		}
-	}
-}
-
-func TestDateTimeOf(t *testing.T) {
-	for _, test := range []struct {
-		time time.Time
-		want LocalDateTime
-	}{
-		{time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local),
-			LocalDateTime{LocalDate{2014, 8, 20}, LocalTime{15, 8, 43, 1}}},
-		{time.Date(1, 1, 1, 0, 0, 0, 0, time.UTC),
-			LocalDateTime{LocalDate{1, 1, 1}, LocalTime{0, 0, 0, 0}}},
-	} {
-		if got := LocalDateTimeOf(test.time); got != test.want {
-			t.Errorf("LocalDateTimeOf(%v) = %+v, want %+v", test.time, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeIsValid(t *testing.T) {
-	// No need to be exhaustive here; it's just LocalDate.IsValid && LocalTime.IsValid.
-	for _, test := range []struct {
-		dt   LocalDateTime
-		want bool
-	}{
-		{LocalDateTime{LocalDate{2016, 3, 20}, LocalTime{0, 0, 0, 0}}, true},
-		{LocalDateTime{LocalDate{2016, -3, 20}, LocalTime{0, 0, 0, 0}}, false},
-		{LocalDateTime{LocalDate{2016, 3, 20}, LocalTime{24, 0, 0, 0}}, false},
-	} {
-		got := test.dt.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.dt, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeIn(t *testing.T) {
-	dt := LocalDateTime{LocalDate{2016, 1, 2}, LocalTime{3, 4, 5, 6}}
-	got := dt.In(time.UTC)
-	want := time.Date(2016, 1, 2, 3, 4, 5, 6, time.UTC)
-	if !got.Equal(want) {
-		t.Errorf("got %v, want %v", got, want)
-	}
-}
-
-func TestDateTimeBefore(t *testing.T) {
-	d1 := LocalDate{2016, 12, 31}
-	d2 := LocalDate{2017, 1, 1}
-	t1 := LocalTime{5, 6, 7, 8}
-	t2 := LocalTime{5, 6, 7, 9}
-	for _, test := range []struct {
-		dt1, dt2 LocalDateTime
-		want     bool
-	}{
-		{LocalDateTime{d1, t1}, LocalDateTime{d2, t1}, true},
-		{LocalDateTime{d1, t1}, LocalDateTime{d1, t2}, true},
-		{LocalDateTime{d2, t1}, LocalDateTime{d1, t1}, false},
-		{LocalDateTime{d2, t1}, LocalDateTime{d2, t1}, false},
-	} {
-		if got := test.dt1.Before(test.dt2); got != test.want {
-			t.Errorf("%v.Before(%v): got %t, want %t", test.dt1, test.dt2, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeAfter(t *testing.T) {
-	d1 := LocalDate{2016, 12, 31}
-	d2 := LocalDate{2017, 1, 1}
-	t1 := LocalTime{5, 6, 7, 8}
-	t2 := LocalTime{5, 6, 7, 9}
-	for _, test := range []struct {
-		dt1, dt2 LocalDateTime
-		want     bool
-	}{
-		{LocalDateTime{d1, t1}, LocalDateTime{d2, t1}, false},
-		{LocalDateTime{d1, t1}, LocalDateTime{d1, t2}, false},
-		{LocalDateTime{d2, t1}, LocalDateTime{d1, t1}, true},
-		{LocalDateTime{d2, t1}, LocalDateTime{d2, t1}, false},
-	} {
-		if got := test.dt1.After(test.dt2); got != test.want {
-			t.Errorf("%v.After(%v): got %t, want %t", test.dt1, test.dt2, got, test.want)
-		}
-	}
-}
-
-func TestMarshalJSON(t *testing.T) {
-	for _, test := range []struct {
-		value interface{}
-		want  string
-	}{
-		{LocalDate{1987, 4, 15}, `"1987-04-15"`},
-		{LocalTime{18, 54, 2, 0}, `"18:54:02"`},
-		{LocalDateTime{LocalDate{1987, 4, 15}, LocalTime{18, 54, 2, 0}}, `"1987-04-15T18:54:02"`},
-	} {
-		bgot, err := json.Marshal(test.value)
-		if err != nil {
-			t.Fatal(err)
-		}
-		if got := string(bgot); got != test.want {
-			t.Errorf("%#v: got %s, want %s", test.value, got, test.want)
-		}
-	}
-}
-
-func TestUnmarshalJSON(t *testing.T) {
-	var d LocalDate
-	var tm LocalTime
-	var dt LocalDateTime
-	for _, test := range []struct {
-		data string
-		ptr  interface{}
-		want interface{}
-	}{
-		{`"1987-04-15"`, &d, &LocalDate{1987, 4, 15}},
-		{`"1987-04-\u0031\u0035"`, &d, &LocalDate{1987, 4, 15}},
-		{`"18:54:02"`, &tm, &LocalTime{18, 54, 2, 0}},
-		{`"1987-04-15T18:54:02"`, &dt, &LocalDateTime{LocalDate{1987, 4, 15}, LocalTime{18, 54, 2, 0}}},
-	} {
-		if err := json.Unmarshal([]byte(test.data), test.ptr); err != nil {
-			t.Fatalf("%s: %v", test.data, err)
-		}
-		if !cmpEqual(test.ptr, test.want) {
-			t.Errorf("%s: got %#v, want %#v", test.data, test.ptr, test.want)
-		}
-	}
-
-	for _, bad := range []string{"", `""`, `"bad"`, `"1987-04-15x"`,
-		`19870415`,     // a JSON number
-		`11987-04-15x`, // not a JSON string
-
-	} {
-		if json.Unmarshal([]byte(bad), &d) == nil {
-			t.Errorf("%q, LocalDate: got nil, want error", bad)
-		}
-		if json.Unmarshal([]byte(bad), &tm) == nil {
-			t.Errorf("%q, LocalTime: got nil, want error", bad)
-		}
-		if json.Unmarshal([]byte(bad), &dt) == nil {
-			t.Errorf("%q, LocalDateTime: got nil, want error", bad)
-		}
-	}
+func TestLocalDateTime_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalDateTime{}
+	err := d.UnmarshalText([]byte("2021-06-08 20:12:01.000000002"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
+	}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
+
+	err = d.UnmarshalText([]byte("2021-06-08 20:12:01.000000002 bad"))
+	require.Error(t, err)
 }

marshaler_test.go

@@ -3,8 +3,11 @@ package toml_test
 import (
 	"bytes"
 	"encoding/json"
+	"fmt"
+	"math/big"
 	"strings"
 	"testing"
+	"time"
 
 	"github.com/pelletier/go-toml/v2"
 	"github.com/stretchr/testify/assert"
@@ -12,6 +15,18 @@ import (
 )
 
 func TestMarshal(t *testing.T) {
+	someInt := 42
+
+	type structInline struct {
+		A interface{} `toml:",inline"`
+	}
+
+	type comments struct {
+		One   int
+		Two   int   `comment:"Before kv"`
+		Three []int `comment:"Before array"`
+	}
+
 	examples := []struct {
 		desc     string
 		v        interface{}
@@ -65,8 +80,6 @@ hello = 'world'`,
 a = 'test'`,
 		},
 		{
-			// TODO: this test is flaky because output changes depending on
-			//   the map iteration order.
 			desc: "map in map in map and string with values",
 			v: map[string]interface{}{
 				"this": map[string]interface{}{
@@ -89,6 +102,16 @@ a = 'test'`,
 			},
 			expected: `array = ['one', 'two', 'three']`,
 		},
+		{
+			desc:     "empty string array",
+			v:        map[string][]string{},
+			expected: ``,
+		},
+		{
+			desc:     "map",
+			v:        map[string][]string{},
+			expected: ``,
+		},
 		{
 			desc: "nested string arrays",
 			v: map[string][][]string{
@@ -104,7 +127,7 @@ a = 'test'`,
 			expected: `array = ['a string', ['one', 'two'], 'last']`,
 		},
 		{
-			desc: "slice of maps",
+			desc: "array of maps",
 			v: map[string][]map[string]string{
 				"top": {
 					{"map1.1": "v1.1"},
@@ -157,7 +180,7 @@ K2 = 'v2'
 `,
 		},
 		{
-			desc: "structs in slice with interfaces",
+			desc: "structs in array with interfaces",
 			v: map[string]interface{}{
 				"root": map[string]interface{}{
 					"nested": []interface{}{
@@ -177,9 +200,9 @@ name = 'Alice'
 		{
 			desc: "string escapes",
 			v: map[string]interface{}{
-				"a": `'"\`,
+				"a": "'\b\f\r\t\"\\",
 			},
-			expected: `a = "'\"\\"`,
+			expected: `a = "'\b\f\r\t\"\\"`,
 		},
 		{
 			desc: "string utf8 low",
@@ -226,7 +249,7 @@ name = 'Alice'
 		{
 			desc: "multi-line forced",
 			v: struct {
-				A string `multiline:"true"`
+				A string `toml:",multiline"`
 			}{
 				A: "hello\nworld",
 			},
@@ -234,17 +257,395 @@ name = 'Alice'
 hello
 world"""`,
 		},
+		{
+			desc: "inline field",
+			v: struct {
+				A map[string]string `toml:",inline"`
+				B map[string]string
+			}{
+				A: map[string]string{
+					"isinline": "yes",
+				},
+				B: map[string]string{
+					"isinline": "no",
+				},
+			},
+			expected: `
+A = {isinline = 'yes'}
+[B]
+isinline = 'no'
+`,
+		},
+		{
+			desc: "mutiline array int",
+			v: struct {
+				A []int `toml:",multiline"`
+				B []int
+			}{
+				A: []int{1, 2, 3, 4},
+				B: []int{1, 2, 3, 4},
+			},
+			expected: `
+A = [
+  1,
+  2,
+  3,
+  4
+]
+B = [1, 2, 3, 4]
+`,
+		},
+		{
+			desc: "mutiline array in array",
+			v: struct {
+				A [][]int `toml:",multiline"`
+			}{
+				A: [][]int{{1, 2}, {3, 4}},
+			},
+			expected: `
+A = [
+  [1, 2],
+  [3, 4]
+]
+`,
+		},
+		{
+			desc: "nil interface not supported at root",
+			v:    nil,
+			err:  true,
+		},
+		{
+			desc: "nil interface not supported in slice",
+			v: map[string]interface{}{
+				"a": []interface{}{"a", nil, 2},
+			},
+			err: true,
+		},
+		{
+			desc: "nil pointer in slice uses zero value",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{nil},
+			},
+			expected: `A = [0]`,
+		},
+		{
+			desc: "nil pointer in slice uses zero value",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{nil},
+			},
+			expected: `A = [0]`,
+		},
+		{
+			desc: "pointer in slice",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{&someInt},
+			},
+			expected: `A = [42]`,
+		},
+		{
+			desc: "inline table in inline table",
+			v: structInline{
+				A: structInline{
+					A: structInline{
+						A: "hello",
+					},
+				},
+			},
+			expected: `A = {A = {A = 'hello'}}`,
+		},
+		{
+			desc: "empty slice in map",
+			v: map[string][]string{
+				"a": {},
+			},
+			expected: `a = []`,
+		},
+		{
+			desc: "map in slice",
+			v: map[string][]map[string]string{
+				"a": {{"hello": "world"}},
+			},
+			expected: `
+[[a]]
+hello = 'world'`,
+		},
+		{
+			desc: "newline in map in slice",
+			v: map[string][]map[string]string{
+				"a\n": {{"hello": "world"}},
+			},
+			err: true,
+		},
+		{
+			desc: "newline in map in slice",
+			v: map[string][]map[string]*customTextMarshaler{
+				"a": {{"hello": &customTextMarshaler{1}}},
+			},
+			err: true,
+		},
+		{
+			desc: "empty slice of empty struct",
+			v: struct {
+				A []struct{}
+			}{
+				A: []struct{}{},
+			},
+			expected: `A = []`,
+		},
+		{
+			desc: "nil field is ignored",
+			v: struct {
+				A interface{}
+			}{
+				A: nil,
+			},
+			expected: ``,
+		},
+		{
+			desc: "private fields are ignored",
+			v: struct {
+				Public  string
+				private string
+			}{
+				Public:  "shown",
+				private: "hidden",
+			},
+			expected: `Public = 'shown'`,
+		},
+		{
+			desc: "fields tagged - are ignored",
+			v: struct {
+				Public  string `toml:"-"`
+				private string
+			}{
+				Public: "hidden",
+			},
+			expected: ``,
+		},
+		{
+			desc: "nil value in map is ignored",
+			v: map[string]interface{}{
+				"A": nil,
+			},
+			expected: ``,
+		},
+		{
+			desc: "new line in table key",
+			v: map[string]interface{}{
+				"hello\nworld": 42,
+			},
+			err: true,
+		},
+		{
+			desc: "new line in parent of nested table key",
+			v: map[string]interface{}{
+				"hello\nworld": map[string]interface{}{
+					"inner": 42,
+				},
+			},
+			err: true,
+		},
+		{
+			desc: "new line in nested table key",
+			v: map[string]interface{}{
+				"parent": map[string]interface{}{
+					"in\ner": map[string]interface{}{
+						"foo": 42,
+					},
+				},
+			},
+			err: true,
+		},
+		{
+			desc: "invalid map key",
+			v:    map[int]interface{}{},
+			err:  true,
+		},
+		{
+			desc: "unhandled type",
+			v: struct {
+				A chan int
+			}{
+				A: make(chan int),
+			},
+			err: true,
+		},
+		{
+			desc: "time",
+			v: struct {
+				T time.Time
+			}{
+				T: time.Time{},
+			},
+			expected: `T = '0001-01-01T00:00:00Z'`,
+		},
+		{
+			desc: "bool",
+			v: struct {
+				A bool
+				B bool
+			}{
+				A: false,
+				B: true,
+			},
+			expected: `
+A = false
+B = true`,
+		},
+		{
+			desc: "numbers",
+			v: struct {
+				A float32
+				B uint64
+				C uint32
+				D uint16
+				E uint8
+				F uint
+				G int64
+				H int32
+				I int16
+				J int8
+				K int
+				L float64
+			}{
+				A: 1.1,
+				B: 42,
+				C: 42,
+				D: 42,
+				E: 42,
+				F: 42,
+				G: 42,
+				H: 42,
+				I: 42,
+				J: 42,
+				K: 42,
+				L: 2.2,
+			},
+			expected: `
+A = 1.1
+B = 42
+C = 42
+D = 42
+E = 42
+F = 42
+G = 42
+H = 42
+I = 42
+J = 42
+K = 42
+L = 2.2`,
+		},
+		{
+			desc: "comments",
+			v: struct {
+				Table comments `comment:"Before table"`
+			}{
+				Table: comments{
+					One:   1,
+					Two:   2,
+					Three: []int{1, 2, 3},
+				},
+			},
+			expected: `
+# Before table
+[Table]
+One = 1
+# Before kv
+Two = 2
+# Before array
+Three = [1, 2, 3]
+`,
+		},
 	}
 
 	for _, e := range examples {
+		e := e
 		t.Run(e.desc, func(t *testing.T) {
 			b, err := toml.Marshal(e.v)
 			if e.err {
 				require.Error(t, err)
-			} else {
-				require.NoError(t, err)
-				equalStringsIgnoreNewlines(t, e.expected, string(b))
+
+				return
 			}
+
+			require.NoError(t, err)
+			equalStringsIgnoreNewlines(t, e.expected, string(b))
+
+			// make sure the output is always valid TOML
+			defaultMap := map[string]interface{}{}
+			err = toml.Unmarshal(b, &defaultMap)
+			require.NoError(t, err)
+
+			testWithAllFlags(t, func(t *testing.T, flags int) {
+				t.Helper()
+
+				var buf bytes.Buffer
+				enc := toml.NewEncoder(&buf)
+				setFlags(enc, flags)
+
+				err := enc.Encode(e.v)
+				require.NoError(t, err)
+
+				inlineMap := map[string]interface{}{}
+				err = toml.Unmarshal(buf.Bytes(), &inlineMap)
+				require.NoError(t, err)
+
+				require.Equal(t, defaultMap, inlineMap)
+			})
+		})
+	}
+}
+
+type flagsSetters []struct {
+	name string
+	f    func(enc *toml.Encoder, flag bool) *toml.Encoder
+}
+
+var allFlags = flagsSetters{
+	{"arrays-multiline", (*toml.Encoder).SetArraysMultiline},
+	{"tables-inline", (*toml.Encoder).SetTablesInline},
+	{"indent-tables", (*toml.Encoder).SetIndentTables},
+}
+
+func setFlags(enc *toml.Encoder, flags int) {
+	for i := 0; i < len(allFlags); i++ {
+		enabled := flags&1 > 0
+		allFlags[i].f(enc, enabled)
+	}
+}
+
+func testWithAllFlags(t *testing.T, testfn func(t *testing.T, flags int)) {
+	t.Helper()
+	testWithFlags(t, 0, allFlags, testfn)
+}
+
+func testWithFlags(t *testing.T, flags int, setters flagsSetters, testfn func(t *testing.T, flags int)) {
+	t.Helper()
+
+	if len(setters) == 0 {
+		testfn(t, flags)
+
+		return
+	}
+
+	s := setters[0]
+
+	for _, enabled := range []bool{false, true} {
+		name := fmt.Sprintf("%s=%t", s.name, enabled)
+		newFlags := flags << 1
+
+		if enabled {
+			newFlags++
+		}
+
+		t.Run(name, func(t *testing.T) {
+			testWithFlags(t, newFlags, setters[1:], testfn)
 		})
 	}
 }
@@ -255,6 +656,191 @@ func equalStringsIgnoreNewlines(t *testing.T, expected string, actual string) {
 	assert.Equal(t, strings.Trim(expected, cutset), strings.Trim(actual, cutset))
 }
 
+//nolint:funlen
+func TestMarshalIndentTables(t *testing.T) {
+	examples := []struct {
+		desc     string
+		v        interface{}
+		expected string
+	}{
+		{
+			desc: "one kv",
+			v: map[string]interface{}{
+				"foo": "bar",
+			},
+			expected: `foo = 'bar'`,
+		},
+		{
+			desc: "one level table",
+			v: map[string]map[string]string{
+				"foo": {
+					"one": "value1",
+					"two": "value2",
+				},
+			},
+			expected: `
+[foo]
+  one = 'value1'
+  two = 'value2'
+`,
+		},
+		{
+			desc: "two levels table",
+			v: map[string]interface{}{
+				"root": "value0",
+				"level1": map[string]interface{}{
+					"one": "value1",
+					"level2": map[string]interface{}{
+						"two": "value2",
+					},
+				},
+			},
+			expected: `
+root = 'value0'
+[level1]
+  one = 'value1'
+  [level1.level2]
+    two = 'value2'
+`,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			var buf strings.Builder
+			enc := toml.NewEncoder(&buf)
+			enc.SetIndentTables(true)
+			err := enc.Encode(e.v)
+			require.NoError(t, err)
+			equalStringsIgnoreNewlines(t, e.expected, buf.String())
+		})
+	}
+}
+
+type customTextMarshaler struct {
+	value int64
+}
+
+func (c *customTextMarshaler) MarshalText() ([]byte, error) {
+	if c.value == 1 {
+		return nil, fmt.Errorf("cannot represent 1 because this is a silly test")
+	}
+	return []byte(fmt.Sprintf("::%d", c.value)), nil
+}
+
+func TestMarshalTextMarshaler_NoRoot(t *testing.T) {
+	c := customTextMarshaler{}
+	_, err := toml.Marshal(&c)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler_Error(t *testing.T) {
+	m := map[string]interface{}{"a": &customTextMarshaler{value: 1}}
+	_, err := toml.Marshal(m)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler_ErrorInline(t *testing.T) {
+	type s struct {
+		A map[string]interface{} `inline:"true"`
+	}
+
+	d := s{
+		A: map[string]interface{}{"a": &customTextMarshaler{value: 1}},
+	}
+
+	_, err := toml.Marshal(d)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler(t *testing.T) {
+	m := map[string]interface{}{"a": &customTextMarshaler{value: 2}}
+	r, err := toml.Marshal(m)
+	require.NoError(t, err)
+	equalStringsIgnoreNewlines(t, "a = '::2'", string(r))
+}
+
+type brokenWriter struct{}
+
+func (b *brokenWriter) Write([]byte) (int, error) {
+	return 0, fmt.Errorf("dead")
+}
+
+func TestEncodeToBrokenWriter(t *testing.T) {
+	w := brokenWriter{}
+	enc := toml.NewEncoder(&w)
+	err := enc.Encode(map[string]string{"hello": "world"})
+	require.Error(t, err)
+}
+
+func TestEncoderSetIndentSymbol(t *testing.T) {
+	var w strings.Builder
+	enc := toml.NewEncoder(&w)
+	enc.SetIndentTables(true)
+	enc.SetIndentSymbol(">>>")
+	err := enc.Encode(map[string]map[string]string{"parent": {"hello": "world"}})
+	require.NoError(t, err)
+	expected := `
+[parent]
+>>>hello = 'world'`
+	equalStringsIgnoreNewlines(t, expected, w.String())
+}
+
+func TestEncoderOmitempty(t *testing.T) {
+	type doc struct {
+		String  string            `toml:",omitempty,multiline"`
+		Bool    bool              `toml:",omitempty,multiline"`
+		Int     int               `toml:",omitempty,multiline"`
+		Int8    int8              `toml:",omitempty,multiline"`
+		Int16   int16             `toml:",omitempty,multiline"`
+		Int32   int32             `toml:",omitempty,multiline"`
+		Int64   int64             `toml:",omitempty,multiline"`
+		Uint    uint              `toml:",omitempty,multiline"`
+		Uint8   uint8             `toml:",omitempty,multiline"`
+		Uint16  uint16            `toml:",omitempty,multiline"`
+		Uint32  uint32            `toml:",omitempty,multiline"`
+		Uint64  uint64            `toml:",omitempty,multiline"`
+		Float32 float32           `toml:",omitempty,multiline"`
+		Float64 float64           `toml:",omitempty,multiline"`
+		MapNil  map[string]string `toml:",omitempty,multiline"`
+		Slice   []string          `toml:",omitempty,multiline"`
+		Ptr     *string           `toml:",omitempty,multiline"`
+		Iface   interface{}       `toml:",omitempty,multiline"`
+		Struct  struct{}          `toml:",omitempty,multiline"`
+	}
+
+	d := doc{}
+
+	b, err := toml.Marshal(d)
+	require.NoError(t, err)
+
+	expected := `[Struct]`
+
+	equalStringsIgnoreNewlines(t, expected, string(b))
+}
+
+func TestEncoderTagFieldName(t *testing.T) {
+	type doc struct {
+		String string `toml:"hello"`
+		OkSym  string `toml:"#"`
+		Bad    string `toml:"\"`
+	}
+
+	d := doc{String: "world"}
+
+	b, err := toml.Marshal(d)
+	require.NoError(t, err)
+
+	expected := `
+hello = 'world'
+'#' = ''
+Bad = ''
+`
+
+	equalStringsIgnoreNewlines(t, expected, string(b))
+}
+
 func TestIssue436(t *testing.T) {
 	data := []byte(`{"a": [ { "b": { "c": "d" } } ]}`)
 
@@ -273,3 +859,179 @@ c = 'd'
 `
 	equalStringsIgnoreNewlines(t, expected, buf.String())
 }
+
+func TestIssue424(t *testing.T) {
+	type Message1 struct {
+		Text string
+	}
+
+	type Message2 struct {
+		Text string `multiline:"true"`
+	}
+
+	msg1 := Message1{"Hello\\World"}
+	msg2 := Message2{"Hello\\World"}
+
+	toml1, err := toml.Marshal(msg1)
+	require.NoError(t, err)
+
+	toml2, err := toml.Marshal(msg2)
+	require.NoError(t, err)
+
+	msg1parsed := Message1{}
+	err = toml.Unmarshal(toml1, &msg1parsed)
+	require.NoError(t, err)
+	require.Equal(t, msg1, msg1parsed)
+
+	msg2parsed := Message2{}
+	err = toml.Unmarshal(toml2, &msg2parsed)
+	require.NoError(t, err)
+	require.Equal(t, msg2, msg2parsed)
+}
+
+func TestIssue567(t *testing.T) {
+	var m map[string]interface{}
+	err := toml.Unmarshal([]byte("A = 12:08:05"), &m)
+	require.NoError(t, err)
+	require.IsType(t, m["A"], toml.LocalTime{})
+}
+
+func TestIssue590(t *testing.T) {
+	type CustomType int
+	var cfg struct {
+		Option CustomType `toml:"option"`
+	}
+	err := toml.Unmarshal([]byte("option = 42"), &cfg)
+	require.NoError(t, err)
+}
+
+func TestIssue571(t *testing.T) {
+	type Foo struct {
+		Float32 float32
+		Float64 float64
+	}
+
+	const closeEnough = 1e-9
+
+	foo := Foo{
+		Float32: 42,
+		Float64: 43,
+	}
+	b, err := toml.Marshal(foo)
+	require.NoError(t, err)
+
+	var foo2 Foo
+	err = toml.Unmarshal(b, &foo2)
+	require.NoError(t, err)
+
+	assert.InDelta(t, 42, foo2.Float32, closeEnough)
+	assert.InDelta(t, 43, foo2.Float64, closeEnough)
+}
+
+func TestIssue678(t *testing.T) {
+	type Config struct {
+		BigInt big.Int
+	}
+
+	cfg := &Config{
+		BigInt: *big.NewInt(123),
+	}
+
+	out, err := toml.Marshal(cfg)
+	require.NoError(t, err)
+	equalStringsIgnoreNewlines(t, "BigInt = '123'", string(out))
+
+	cfg2 := &Config{}
+	err = toml.Unmarshal(out, cfg2)
+	require.NoError(t, err)
+	require.Equal(t, cfg, cfg2)
+}
+
+func TestMarshalNestedAnonymousStructs(t *testing.T) {
+	type Embedded struct {
+		Value string `toml:"value" json:"value"`
+		Top   struct {
+			Value string `toml:"value" json:"value"`
+		} `toml:"top" json:"top"`
+	}
+
+	type Named struct {
+		Value string `toml:"value" json:"value"`
+	}
+
+	var doc struct {
+		Embedded
+		Named     `toml:"named" json:"named"`
+		Anonymous struct {
+			Value string `toml:"value" json:"value"`
+		} `toml:"anonymous" json:"anonymous"`
+	}
+
+	expected := `value = ''
+[top]
+value = ''
+
+[named]
+value = ''
+
+[anonymous]
+value = ''
+
+`
+
+	result, err := toml.Marshal(doc)
+	require.NoError(t, err)
+	require.Equal(t, expected, string(result))
+}
+
+func TestMarshalNestedAnonymousStructs_DuplicateField(t *testing.T) {
+	type Embedded struct {
+		Value string `toml:"value" json:"value"`
+		Top   struct {
+			Value string `toml:"value" json:"value"`
+		} `toml:"top" json:"top"`
+	}
+
+	var doc struct {
+		Value string `toml:"value" json:"value"`
+		Embedded
+	}
+	doc.Embedded.Value = "shadowed"
+	doc.Value = "shadows"
+
+	expected := `value = 'shadows'
+[top]
+value = ''
+
+`
+
+	result, err := toml.Marshal(doc)
+	require.NoError(t, err)
+	require.NoError(t, err)
+	require.Equal(t, expected, string(result))
+}
+
+func ExampleMarshal() {
+	type MyConfig struct {
+		Version int
+		Name    string
+		Tags    []string
+	}
+
+	cfg := MyConfig{
+		Version: 2,
+		Name:    "go-toml",
+		Tags:    []string{"go", "toml"},
+	}
+
+	b, err := toml.Marshal(cfg)
+	if err != nil {
+		panic(err)
+	}
+	fmt.Println(string(b))
+
+	// Output:
+	// Version = 2
+	// Name = 'go-toml'
+	// Tags = ['go', 'toml']
+}

parser_test.go

@@ -1,12 +1,15 @@
 package toml
 
 import (
+	"strconv"
+	"strings"
 	"testing"
 
 	"github.com/pelletier/go-toml/v2/internal/ast"
 	"github.com/stretchr/testify/require"
 )
 
+//nolint:funlen
 func TestParser_AST_Numbers(t *testing.T) {
 	examples := []struct {
 		desc  string
@@ -132,6 +135,7 @@ func TestParser_AST_Numbers(t *testing.T) {
 	}
 
 	for _, e := range examples {
+		e := e
 		t.Run(e.desc, func(t *testing.T) {
 			p := parser{}
 			p.Reset([]byte(`A = ` + e.input))
@@ -155,19 +159,16 @@ func TestParser_AST_Numbers(t *testing.T) {
 	}
 }
 
-type astRoot []astNode
-type astNode struct {
-	Kind     ast.Kind
-	Data     []byte
-	Children []astNode
-}
+type (
+	astNode struct {
+		Kind     ast.Kind
+		Data     []byte
+		Children []astNode
+	}
+)
 
-func compareAST(t *testing.T, expected astRoot, actual *ast.Root) {
-	it := actual.Iterator()
-	compareIterator(t, expected, it)
-}
-
-func compareNode(t *testing.T, e astNode, n ast.Node) {
+func compareNode(t *testing.T, e astNode, n *ast.Node) {
+	t.Helper()
 	require.Equal(t, e.Kind, n.Kind)
 	require.Equal(t, e.Data, n.Data)
 
@@ -175,6 +176,7 @@ func compareNode(t *testing.T, e astNode, n ast.Node) {
 }
 
 func compareIterator(t *testing.T, expected []astNode, actual ast.Iterator) {
+	t.Helper()
 	idx := 0
 
 	for actual.Next() {
@@ -195,54 +197,7 @@ func compareIterator(t *testing.T, expected []astNode, actual ast.Iterator) {
 	}
 }
 
-func (r astRoot) toOrig() *ast.Root {
-	builder := &ast.Builder{}
-
-	var last ast.Reference
-
-	for i, n := range r {
-		ref := builder.Push(ast.Node{
-			Kind: n.Kind,
-			Data: n.Data,
-		})
-
-		if i > 0 {
-			builder.Chain(last, ref)
-		}
-		last = ref
-
-		if len(n.Children) > 0 {
-			c := childrenToOrig(builder, n.Children)
-			builder.AttachChild(ref, c)
-		}
-	}
-
-	return builder.Tree()
-}
-
-func childrenToOrig(b *ast.Builder, nodes []astNode) ast.Reference {
-	var first ast.Reference
-	var last ast.Reference
-	for i, n := range nodes {
-		ref := b.Push(ast.Node{
-			Kind: n.Kind,
-			Data: n.Data,
-		})
-		if i == 0 {
-			first = ref
-		} else {
-			b.Chain(last, ref)
-		}
-		last = ref
-
-		if len(n.Children) > 0 {
-			c := childrenToOrig(b, n.Children)
-			b.AttachChild(ref, c)
-		}
-	}
-	return first
-}
-
+//nolint:funlen
 func TestParser_AST(t *testing.T) {
 	examples := []struct {
 		desc  string
@@ -380,6 +335,7 @@ func TestParser_AST(t *testing.T) {
 	}
 
 	for _, e := range examples {
+		e := e
 		t.Run(e.desc, func(t *testing.T) {
 			p := parser{}
 			p.Reset([]byte(e.input))
@@ -394,3 +350,101 @@ func TestParser_AST(t *testing.T) {
 		})
 	}
 }
+
+func BenchmarkParseBasicStringWithUnicode(b *testing.B) {
+	p := &parser{}
+	b.Run("4", func(b *testing.B) {
+		input := []byte(`"\u1234\u5678\u9ABC\u1234\u5678\u9ABC"`)
+		b.ReportAllocs()
+		b.SetBytes(int64(len(input)))
+
+		for i := 0; i < b.N; i++ {
+			p.parseBasicString(input)
+		}
+	})
+	b.Run("8", func(b *testing.B) {
+		input := []byte(`"\u12345678\u9ABCDEF0\u12345678\u9ABCDEF0"`)
+		b.ReportAllocs()
+		b.SetBytes(int64(len(input)))
+
+		for i := 0; i < b.N; i++ {
+			p.parseBasicString(input)
+		}
+	})
+}
+
+func BenchmarkParseBasicStringsEasy(b *testing.B) {
+	p := &parser{}
+
+	for _, size := range []int{1, 4, 8, 16, 21} {
+		b.Run(strconv.Itoa(size), func(b *testing.B) {
+			input := []byte(`"` + strings.Repeat("A", size) + `"`)
+
+			b.ReportAllocs()
+			b.SetBytes(int64(len(input)))
+
+			for i := 0; i < b.N; i++ {
+				p.parseBasicString(input)
+			}
+		})
+	}
+}
+
+func TestParser_AST_DateTimes(t *testing.T) {
+	examples := []struct {
+		desc  string
+		input string
+		kind  ast.Kind
+		err   bool
+	}{
+		{
+			desc:  "offset-date-time with delim 'T' and UTC offset",
+			input: `2021-07-21T12:08:05Z`,
+			kind:  ast.DateTime,
+		},
+		{
+			desc:  "offset-date-time with space delim and +8hours offset",
+			input: `2021-07-21 12:08:05+08:00`,
+			kind:  ast.DateTime,
+		},
+		{
+			desc:  "local-date-time with nano second",
+			input: `2021-07-21T12:08:05.666666666`,
+			kind:  ast.LocalDateTime,
+		},
+		{
+			desc:  "local-date-time",
+			input: `2021-07-21T12:08:05`,
+			kind:  ast.LocalDateTime,
+		},
+		{
+			desc:  "local-date",
+			input: `2021-07-21`,
+			kind:  ast.LocalDate,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			p := parser{}
+			p.Reset([]byte(`A = ` + e.input))
+			p.NextExpression()
+			err := p.Error()
+			if e.err {
+				require.Error(t, err)
+			} else {
+				require.NoError(t, err)
+
+				expected := astNode{
+					Kind: ast.KeyValue,
+					Children: []astNode{
+						{Kind: e.kind, Data: []byte(e.input)},
+						{Kind: ast.Key, Data: []byte(`A`)},
+					},
+				}
+				compareNode(t, expected, p.Expression())
+			}
+		})
+	}
+}

scanner.go

@@ -1,36 +1,44 @@
 package toml
 
-import "fmt"
+func scanFollows(b []byte, pattern string) bool {
+	n := len(pattern)
 
-func scanFollows(pattern []byte) func(b []byte) bool {
-	return func(b []byte) bool {
-		if len(b) < len(pattern) {
-			return false
-		}
-		for i, c := range pattern {
-			if b[i] != c {
-				return false
-			}
-		}
-		return true
-	}
+	return len(b) >= n && string(b[:n]) == pattern
 }
 
-var scanFollowsMultilineBasicStringDelimiter = scanFollows([]byte{'"', '"', '"'})
-var scanFollowsMultilineLiteralStringDelimiter = scanFollows([]byte{'\'', '\'', '\''})
-var scanFollowsTrue = scanFollows([]byte{'t', 'r', 'u', 'e'})
-var scanFollowsFalse = scanFollows([]byte{'f', 'a', 'l', 's', 'e'})
-var scanFollowsInf = scanFollows([]byte{'i', 'n', 'f'})
-var scanFollowsNan = scanFollows([]byte{'n', 'a', 'n'})
+func scanFollowsMultilineBasicStringDelimiter(b []byte) bool {
+	return scanFollows(b, `"""`)
+}
 
-func scanUnquotedKey(b []byte) ([]byte, []byte, error) {
-	//unquoted-key = 1*( ALPHA / DIGIT / %x2D / %x5F ) ; A-Z / a-z / 0-9 / - / _
+func scanFollowsMultilineLiteralStringDelimiter(b []byte) bool {
+	return scanFollows(b, `'''`)
+}
+
+func scanFollowsTrue(b []byte) bool {
+	return scanFollows(b, `true`)
+}
+
+func scanFollowsFalse(b []byte) bool {
+	return scanFollows(b, `false`)
+}
+
+func scanFollowsInf(b []byte) bool {
+	return scanFollows(b, `inf`)
+}
+
+func scanFollowsNan(b []byte) bool {
+	return scanFollows(b, `nan`)
+}
+
+func scanUnquotedKey(b []byte) ([]byte, []byte) {
+	// unquoted-key = 1*( ALPHA / DIGIT / %x2D / %x5F ) ; A-Z / a-z / 0-9 / - / _
 	for i := 0; i < len(b); i++ {
 		if !isUnquotedKeyChar(b[i]) {
-			return b[:i], b[i:], nil
+			return b[:i], b[i:]
 		}
 	}
-	return b, b[len(b):], nil
+
+	return b, b[len(b):]
 }
 
 func isUnquotedKeyChar(r byte) bool {
@@ -38,49 +46,94 @@ func isUnquotedKeyChar(r byte) bool {
 }
 
 func scanLiteralString(b []byte) ([]byte, []byte, error) {
-	//literal-string = apostrophe *literal-char apostrophe
-	//apostrophe = %x27 ; ' apostrophe
-	//literal-char = %x09 / %x20-26 / %x28-7E / non-ascii
-	for i := 1; i < len(b); i++ {
+	// literal-string = apostrophe *literal-char apostrophe
+	// apostrophe = %x27 ; ' apostrophe
+	// literal-char = %x09 / %x20-26 / %x28-7E / non-ascii
+	for i := 1; i < len(b); {
 		switch b[i] {
 		case '\'':
 			return b[:i+1], b[i+1:], nil
-		case '\n':
+		case '\n', '\r':
 			return nil, nil, newDecodeError(b[i:i+1], "literal strings cannot have new lines")
 		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
 	}
+
 	return nil, nil, newDecodeError(b[len(b):], "unterminated literal string")
 }
 
 func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
-	//ml-literal-string = ml-literal-string-delim [ newline ] ml-literal-body
-	//ml-literal-string-delim
-	//ml-literal-string-delim = 3apostrophe
-	//ml-literal-body = *mll-content *( mll-quotes 1*mll-content ) [ mll-quotes ]
+	// ml-literal-string = ml-literal-string-delim [ newline ] ml-literal-body
+	// ml-literal-string-delim
+	// ml-literal-string-delim = 3apostrophe
+	// ml-literal-body = *mll-content *( mll-quotes 1*mll-content ) [ mll-quotes ]
 	//
-	//mll-content = mll-char / newline
-	//mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
-	//mll-quotes = 1*2apostrophe
-	for i := 3; i < len(b); i++ {
+	// mll-content = mll-char / newline
+	// mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
+	// mll-quotes = 1*2apostrophe
+	for i := 3; i < len(b); {
 		switch b[i] {
 		case '\'':
 			if scanFollowsMultilineLiteralStringDelimiter(b[i:]) {
-				return b[:i+3], b[i+3:], nil
+				i += 3
+
+				// At that point we found 3 apostrophe, and i is the
+				// index of the byte after the third one. The scanner
+				// needs to be eager, because there can be an extra 2
+				// apostrophe that can be accepted at the end of the
+				// string.
+
+				if i >= len(b) || b[i] != '\'' {
+					return b[:i], b[i:], nil
+				}
+				i++
+
+				if i >= len(b) || b[i] != '\'' {
+					return b[:i], b[i:], nil
+				}
+				i++
+
+				if i < len(b) && b[i] == '\'' {
+					return nil, nil, newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+				}
+
+				return b[:i], b[i:], nil
 			}
+		case '\r':
+			if len(b) < i+2 {
+				return nil, nil, newDecodeError(b[len(b):], `need a \n after \r`)
+			}
+			if b[i+1] != '\n' {
+				return nil, nil, newDecodeError(b[i:i+2], `need a \n after \r`)
+			}
+			i += 2 // skip the \n
+			continue
 		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
 	}
 
 	return nil, nil, newDecodeError(b[len(b):], `multiline literal string not terminated by '''`)
 }
 
 func scanWindowsNewline(b []byte) ([]byte, []byte, error) {
-	if len(b) < 2 {
-		return nil, nil, fmt.Errorf(`windows new line missing \n`)
+	const lenCRLF = 2
+	if len(b) < lenCRLF {
+		return nil, nil, newDecodeError(b, "windows new line expected")
 	}
+
 	if b[1] != '\n' {
-		return nil, nil, fmt.Errorf(`windows new line should be \r\n`)
+		return nil, nil, newDecodeError(b, `windows new line should be \r\n`)
 	}
-	return b[:2], b[2:], nil
+
+	return b[:lenCRLF], b[lenCRLF:], nil
 }
 
 func scanWhitespace(b []byte) ([]byte, []byte) {
@@ -92,77 +145,125 @@ func scanWhitespace(b []byte) ([]byte, []byte) {
 			return b[:i], b[i:]
 		}
 	}
+
 	return b, b[len(b):]
 }
 
+//nolint:unparam
 func scanComment(b []byte) ([]byte, []byte, error) {
-	//;; Comment
+	// comment-start-symbol = %x23 ; #
+	// non-ascii = %x80-D7FF / %xE000-10FFFF
+	// non-eol = %x09 / %x20-7F / non-ascii
 	//
-	//comment-start-symbol = %x23 ; #
-	//non-ascii = %x80-D7FF / %xE000-10FFFF
-	//non-eol = %x09 / %x20-7F / non-ascii
-	//
-	//comment = comment-start-symbol *non-eol
+	// comment = comment-start-symbol *non-eol
 
-	for i := 1; i < len(b); i++ {
-		switch b[i] {
-		case '\n':
+	for i := 1; i < len(b); {
+		if b[i] == '\n' {
 			return b[:i], b[i:], nil
 		}
+		if b[i] == '\r' {
+			if i+1 < len(b) && b[i+1] == '\n' {
+				return b[:i+1], b[i+1:], nil
+			}
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character in comment")
+		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character in comment")
+		}
+
+		i += size
 	}
-	return b, nil, nil
+
+	return b, b[len(b):], nil
 }
 
-// TODO perform validation on the string?
-func scanBasicString(b []byte) ([]byte, []byte, error) {
-	//basic-string = quotation-mark *basic-char quotation-mark
-	//quotation-mark = %x22            ; "
-	//basic-char = basic-unescaped / escaped
-	//basic-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
-	//escaped = escape escape-seq-char
-	for i := 1; i < len(b); i++ {
+func scanBasicString(b []byte) ([]byte, bool, []byte, error) {
+	// basic-string = quotation-mark *basic-char quotation-mark
+	// quotation-mark = %x22            ; "
+	// basic-char = basic-unescaped / escaped
+	// basic-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+	// escaped = escape escape-seq-char
+	escaped := false
+	i := 1
+
+	for ; i < len(b); i++ {
 		switch b[i] {
 		case '"':
-			return b[:i+1], b[i+1:], nil
-		case '\n':
-			return nil, nil, newDecodeError(b[i:i+1], "basic strings cannot have new lines")
+			return b[:i+1], escaped, b[i+1:], nil
+		case '\n', '\r':
+			return nil, escaped, nil, newDecodeError(b[i:i+1], "basic strings cannot have new lines")
 		case '\\':
 			if len(b) < i+2 {
-				return nil, nil, newDecodeError(b[i:i+1], "need a character after \\")
+				return nil, escaped, nil, newDecodeError(b[i:i+1], "need a character after \\")
 			}
+			escaped = true
 			i++ // skip the next character
 		}
 	}
 
-	return nil, nil, fmt.Errorf(`basic string not terminated by "`)
+	return nil, escaped, nil, newDecodeError(b[len(b):], `basic string not terminated by "`)
 }
 
-// TODO perform validation on the string?
-func scanMultilineBasicString(b []byte) ([]byte, []byte, error) {
-	//ml-basic-string = ml-basic-string-delim [ newline ] ml-basic-body
-	//ml-basic-string-delim
-	//ml-basic-string-delim = 3quotation-mark
-	//ml-basic-body = *mlb-content *( mlb-quotes 1*mlb-content ) [ mlb-quotes ]
+func scanMultilineBasicString(b []byte) ([]byte, bool, []byte, error) {
+	// ml-basic-string = ml-basic-string-delim [ newline ] ml-basic-body
+	// ml-basic-string-delim
+	// ml-basic-string-delim = 3quotation-mark
+	// ml-basic-body = *mlb-content *( mlb-quotes 1*mlb-content ) [ mlb-quotes ]
 	//
-	//mlb-content = mlb-char / newline / mlb-escaped-nl
-	//mlb-char = mlb-unescaped / escaped
-	//mlb-quotes = 1*2quotation-mark
-	//mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
-	//mlb-escaped-nl = escape ws newline *( wschar / newline )
+	// mlb-content = mlb-char / newline / mlb-escaped-nl
+	// mlb-char = mlb-unescaped / escaped
+	// mlb-quotes = 1*2quotation-mark
+	// mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+	// mlb-escaped-nl = escape ws newline *( wschar / newline )
 
-	for i := 3; i < len(b); i++ {
+	escaped := false
+	i := 3
+
+	for ; i < len(b); i++ {
 		switch b[i] {
 		case '"':
 			if scanFollowsMultilineBasicStringDelimiter(b[i:]) {
-				return b[:i+3], b[i+3:], nil
+				i += 3
+
+				// At that point we found 3 apostrophe, and i is the
+				// index of the byte after the third one. The scanner
+				// needs to be eager, because there can be an extra 2
+				// apostrophe that can be accepted at the end of the
+				// string.
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i < len(b) && b[i] == '"' {
+					return nil, escaped, nil, newDecodeError(b[i-3:i+1], `""" not allowed in multiline basic string`)
+				}
+
+				return b[:i], escaped, b[i:], nil
 			}
 		case '\\':
 			if len(b) < i+2 {
-				return nil, nil, newDecodeError(b[len(b):], "need a character after \\")
+				return nil, escaped, nil, newDecodeError(b[len(b):], "need a character after \\")
 			}
+			escaped = true
 			i++ // skip the next character
+		case '\r':
+			if len(b) < i+2 {
+				return nil, escaped, nil, newDecodeError(b[len(b):], `need a \n after \r`)
+			}
+			if b[i+1] != '\n' {
+				return nil, escaped, nil, newDecodeError(b[i:i+2], `need a \n after \r`)
+			}
+			i++ // skip the \n
 		}
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
+	return nil, escaped, nil, newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
 }

strict.go

@@ -0,0 +1,107 @@
+package toml
+
+import (
+	"github.com/pelletier/go-toml/v2/internal/ast"
+	"github.com/pelletier/go-toml/v2/internal/danger"
+	"github.com/pelletier/go-toml/v2/internal/tracker"
+)
+
+type strict struct {
+	Enabled bool
+
+	// Tracks the current key being processed.
+	key tracker.KeyTracker
+
+	missing []decodeError
+}
+
+func (s *strict) EnterTable(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.key.UpdateTable(node)
+}
+
+func (s *strict) EnterArrayTable(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.key.UpdateArrayTable(node)
+}
+
+func (s *strict) EnterKeyValue(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.key.Push(node)
+}
+
+func (s *strict) ExitKeyValue(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.key.Pop(node)
+}
+
+func (s *strict) MissingTable(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.missing = append(s.missing, decodeError{
+		highlight: keyLocation(node),
+		message:   "missing table",
+		key:       s.key.Key(),
+	})
+}
+
+func (s *strict) MissingField(node *ast.Node) {
+	if !s.Enabled {
+		return
+	}
+
+	s.missing = append(s.missing, decodeError{
+		highlight: keyLocation(node),
+		message:   "missing field",
+		key:       s.key.Key(),
+	})
+}
+
+func (s *strict) Error(doc []byte) error {
+	if !s.Enabled || len(s.missing) == 0 {
+		return nil
+	}
+
+	err := &StrictMissingError{
+		Errors: make([]DecodeError, 0, len(s.missing)),
+	}
+
+	for _, derr := range s.missing {
+		derr := derr
+		err.Errors = append(err.Errors, *wrapDecodeError(doc, &derr))
+	}
+
+	return err
+}
+
+func keyLocation(node *ast.Node) []byte {
+	k := node.Key()
+
+	hasOne := k.Next()
+	if !hasOne {
+		panic("should not be called with empty key")
+	}
+
+	start := k.Node().Data
+	end := k.Node().Data
+
+	for k.Next() {
+		end = k.Node().Data
+	}
+
+	return danger.BytesRange(start, end)
+}

targets.go

@@ -1,554 +0,0 @@
-package toml
-
-import (
-	"fmt"
-	"math"
-	"reflect"
-	"strings"
-	"sync"
-)
-
-type target interface {
-	// Dereferences the target.
-	get() reflect.Value
-
-	// Store a string at the target.
-	setString(v string) error
-
-	// Store a boolean at the target
-	setBool(v bool) error
-
-	// Store an int64 at the target
-	setInt64(v int64) error
-
-	// Store a float64 at the target
-	setFloat64(v float64) error
-
-	// Stores any  value at the target
-	set(v reflect.Value) error
-}
-
-// valueTarget just contains a reflect.Value that can be set.
-// It is used for struct fields.
-type valueTarget reflect.Value
-
-func (t valueTarget) get() reflect.Value {
-	return reflect.Value(t)
-}
-
-func (t valueTarget) set(v reflect.Value) error {
-	reflect.Value(t).Set(v)
-	return nil
-}
-
-func (t valueTarget) setString(v string) error {
-	t.get().SetString(v)
-	return nil
-}
-
-func (t valueTarget) setBool(v bool) error {
-	t.get().SetBool(v)
-	return nil
-}
-
-func (t valueTarget) setInt64(v int64) error {
-	t.get().SetInt(v)
-	return nil
-}
-
-func (t valueTarget) setFloat64(v float64) error {
-	t.get().SetFloat(v)
-	return nil
-}
-
-// interfaceTarget wraps an other target to dereference on get.
-type interfaceTarget struct {
-	x target
-}
-
-func (t interfaceTarget) get() reflect.Value {
-	return t.x.get().Elem()
-}
-
-func (t interfaceTarget) set(v reflect.Value) error {
-	return t.x.set(v)
-}
-
-func (t interfaceTarget) setString(v string) error {
-	return t.x.setString(v)
-}
-
-func (t interfaceTarget) setBool(v bool) error {
-	return t.x.setBool(v)
-}
-
-func (t interfaceTarget) setInt64(v int64) error {
-	return t.x.setInt64(v)
-}
-
-func (t interfaceTarget) setFloat64(v float64) error {
-	return t.x.setFloat64(v)
-}
-
-// mapTarget targets a specific key of a map.
-type mapTarget struct {
-	v reflect.Value
-	k reflect.Value
-}
-
-func (t mapTarget) get() reflect.Value {
-	return t.v.MapIndex(t.k)
-}
-
-func (t mapTarget) set(v reflect.Value) error {
-	t.v.SetMapIndex(t.k, v)
-	return nil
-}
-
-func (t mapTarget) setString(v string) error {
-	return t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setBool(v bool) error {
-	return t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setInt64(v int64) error {
-	return t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setFloat64(v float64) error {
-	return t.set(reflect.ValueOf(v))
-}
-
-// makes sure that the value pointed at by t is indexable (Slice, Array), or
-// dereferences to an indexable (Ptr, Interface).
-func ensureValueIndexable(t target) error {
-	f := t.get()
-
-	switch f.Type().Kind() {
-	case reflect.Slice:
-		if f.IsNil() {
-			return t.set(reflect.MakeSlice(f.Type(), 0, 0))
-		}
-	case reflect.Interface:
-		if f.IsNil() || f.Elem().Type() != sliceInterfaceType {
-			return t.set(reflect.MakeSlice(sliceInterfaceType, 0, 0))
-		}
-		if f.Elem().Type().Kind() != reflect.Slice {
-			return fmt.Errorf("interface is pointing to a %s, not a slice", f.Kind())
-		}
-	case reflect.Ptr:
-		if f.IsNil() {
-			ptr := reflect.New(f.Type().Elem())
-			err := t.set(ptr)
-			if err != nil {
-				return err
-			}
-			f = t.get()
-		}
-		return ensureValueIndexable(valueTarget(f.Elem()))
-	case reflect.Array:
-		// arrays are always initialized.
-	default:
-		return fmt.Errorf("cannot initialize a slice in %s", f.Kind())
-	}
-	return nil
-}
-
-var sliceInterfaceType = reflect.TypeOf([]interface{}{})
-var mapStringInterfaceType = reflect.TypeOf(map[string]interface{}{})
-
-func ensureMapIfInterface(x target) {
-	v := x.get()
-	if v.Kind() == reflect.Interface && v.IsNil() {
-		newElement := reflect.MakeMap(mapStringInterfaceType)
-		x.set(newElement)
-	}
-}
-
-func setString(t target, v string) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.String:
-		return t.setString(v)
-	case reflect.Interface:
-		return t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("cannot assign string to a %s", f.Kind())
-	}
-}
-
-func setBool(t target, v bool) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Bool:
-		return t.setBool(v)
-	case reflect.Interface:
-		return t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("cannot assign bool to a %s", f.String())
-	}
-}
-
-const maxInt = int64(^uint(0) >> 1)
-const minInt = -maxInt - 1
-
-func setInt64(t target, v int64) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Int64:
-		return t.setInt64(v)
-	case reflect.Int32:
-		if v < math.MinInt32 || v > math.MaxInt32 {
-			return fmt.Errorf("integer %d does not fit in an int32", v)
-		}
-		return t.set(reflect.ValueOf(int32(v)))
-	case reflect.Int16:
-		if v < math.MinInt16 || v > math.MaxInt16 {
-			return fmt.Errorf("integer %d does not fit in an int16", v)
-		}
-		return t.set(reflect.ValueOf(int16(v)))
-	case reflect.Int8:
-		if v < math.MinInt8 || v > math.MaxInt8 {
-			return fmt.Errorf("integer %d does not fit in an int8", v)
-		}
-		return t.set(reflect.ValueOf(int8(v)))
-	case reflect.Int:
-		if v < minInt || v > maxInt {
-			return fmt.Errorf("integer %d does not fit in an int", v)
-		}
-		return t.set(reflect.ValueOf(int(v)))
-
-	case reflect.Uint64:
-		if v < 0 {
-			return fmt.Errorf("negative integer %d cannot be stored in an uint64", v)
-		}
-		return t.set(reflect.ValueOf(uint64(v)))
-	case reflect.Uint32:
-		if v < 0 {
-			return fmt.Errorf("negative integer %d cannot be stored in an uint32", v)
-		}
-		if v > math.MaxUint32 {
-			return fmt.Errorf("integer %d cannot be stored in an uint32", v)
-		}
-		return t.set(reflect.ValueOf(uint32(v)))
-	case reflect.Uint16:
-		if v < 0 {
-			return fmt.Errorf("negative integer %d cannot be stored in an uint16", v)
-		}
-		if v > math.MaxUint16 {
-			return fmt.Errorf("integer %d cannot be stored in an uint16", v)
-		}
-		return t.set(reflect.ValueOf(uint16(v)))
-	case reflect.Uint8:
-		if v < 0 {
-			return fmt.Errorf("negative integer %d cannot be stored in an uint8", v)
-		}
-		if v > math.MaxUint8 {
-			return fmt.Errorf("integer %d cannot be stored in an uint8", v)
-		}
-		return t.set(reflect.ValueOf(uint8(v)))
-	case reflect.Uint:
-		if v < 0 {
-			return fmt.Errorf("negative integer %d cannot be stored in an uint", v)
-		}
-		return t.set(reflect.ValueOf(uint(v)))
-	case reflect.Interface:
-		return t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("cannot assign int64 to a %s", f.String())
-	}
-}
-
-func setFloat64(t target, v float64) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Float64:
-		return t.setFloat64(v)
-	case reflect.Float32:
-		if v > math.MaxFloat32 {
-			return fmt.Errorf("float %f cannot be stored in a float32", v)
-		}
-		return t.set(reflect.ValueOf(float32(v)))
-	case reflect.Interface:
-		return t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("cannot assign float64 to a %s", f.String())
-	}
-}
-
-// Returns the element at idx of the value pointed at by target, or an error if
-// t does not point to an indexable.
-// If the target points to an Array and idx is out of bounds, it returns
-// (nil, nil) as this is not a fatal error (the unmarshaler will skip).
-func elementAt(t target, idx int) (target, error) {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Slice:
-		// TODO: use the idx function argument and avoid alloc if possible.
-		idx := f.Len()
-		err := t.set(reflect.Append(f, reflect.New(f.Type().Elem()).Elem()))
-		if err != nil {
-			return nil, err
-		}
-		return valueTarget(t.get().Index(idx)), nil
-	case reflect.Array:
-		if idx >= f.Len() {
-			return nil, nil
-		}
-		return valueTarget(f.Index(idx)), nil
-	case reflect.Interface:
-		if f.IsNil() {
-			panic("interface should have been initialized")
-		}
-		ifaceElem := f.Elem()
-		if ifaceElem.Kind() != reflect.Slice {
-			return nil, fmt.Errorf("cannot elementAt on a %s", f.Kind())
-		}
-		idx := ifaceElem.Len()
-		newElem := reflect.New(ifaceElem.Type().Elem()).Elem()
-		newSlice := reflect.Append(ifaceElem, newElem)
-		err := t.set(newSlice)
-		if err != nil {
-			return nil, err
-		}
-		return valueTarget(t.get().Elem().Index(idx)), nil
-	case reflect.Ptr:
-		return elementAt(valueTarget(f.Elem()), idx)
-	default:
-		return nil, fmt.Errorf("cannot elementAt on a %s", f.Kind())
-	}
-}
-
-func (d *decoder) scopeTableTarget(append bool, t target, name string) (target, bool, error) {
-	x := t.get()
-
-	switch x.Kind() {
-	// Kinds that need to recurse
-
-	case reflect.Interface:
-		t, err := scopeInterface(append, t)
-		if err != nil {
-			return t, false, err
-		}
-		return d.scopeTableTarget(append, t, name)
-	case reflect.Ptr:
-		t, err := scopePtr(t)
-		if err != nil {
-			return t, false, err
-		}
-		return d.scopeTableTarget(append, t, name)
-	case reflect.Slice:
-		t, err := scopeSlice(append, t)
-		if err != nil {
-			return t, false, err
-		}
-		append = false
-		return d.scopeTableTarget(append, t, name)
-	case reflect.Array:
-		t, err := d.scopeArray(append, t)
-		if err != nil {
-			return t, false, err
-		}
-		append = false
-		return d.scopeTableTarget(append, t, name)
-
-	// Terminal kinds
-
-	case reflect.Struct:
-		return scopeStruct(x, name)
-	case reflect.Map:
-		if x.IsNil() {
-			t.set(reflect.MakeMap(x.Type()))
-			x = t.get()
-		}
-
-		return scopeMap(x, name)
-	default:
-		panic(fmt.Errorf("can't scope on a %s", x.Kind()))
-	}
-}
-
-func scopeInterface(append bool, t target) (target, error) {
-	err := initInterface(append, t)
-	if err != nil {
-		return t, err
-	}
-	return interfaceTarget{t}, nil
-}
-
-func scopePtr(t target) (target, error) {
-	err := initPtr(t)
-	if err != nil {
-		return t, err
-	}
-	return valueTarget(t.get().Elem()), nil
-}
-
-func initPtr(t target) error {
-	x := t.get()
-	if !x.IsNil() {
-		return nil
-	}
-	return t.set(reflect.New(x.Type().Elem()))
-}
-
-// initInterface makes sure that the interface pointed at by the target is not
-// nil.
-// Returns the target to the initialized value of the target.
-func initInterface(append bool, t target) error {
-	x := t.get()
-
-	if x.Kind() != reflect.Interface {
-		panic("this should only be called on interfaces")
-	}
-
-	if !x.IsNil() && (x.Elem().Type() == sliceInterfaceType || x.Elem().Type() == mapStringInterfaceType) {
-		return nil
-	}
-
-	var newElement reflect.Value
-	if append {
-		newElement = reflect.MakeSlice(sliceInterfaceType, 0, 0)
-	} else {
-		newElement = reflect.MakeMap(mapStringInterfaceType)
-	}
-	err := t.set(newElement)
-	if err != nil {
-		return err
-	}
-
-	return nil
-}
-
-func scopeSlice(append bool, t target) (target, error) {
-	v := t.get()
-
-	if append {
-		newElem := reflect.New(v.Type().Elem())
-		newSlice := reflect.Append(v, newElem.Elem())
-		err := t.set(newSlice)
-		if err != nil {
-			return t, err
-		}
-		v = t.get()
-	}
-	return valueTarget(v.Index(v.Len() - 1)), nil
-}
-
-func (d *decoder) scopeArray(append bool, t target) (target, error) {
-	v := t.get()
-
-	idx := d.arrayIndex(append, v)
-
-	if idx >= v.Len() {
-		return nil, fmt.Errorf("not enough space in the array")
-	}
-
-	return valueTarget(v.Index(idx)), nil
-}
-
-func scopeMap(v reflect.Value, name string) (target, bool, error) {
-	k := reflect.ValueOf(name)
-
-	keyType := v.Type().Key()
-	if !k.Type().AssignableTo(keyType) {
-		if !k.Type().ConvertibleTo(keyType) {
-			return nil, false, fmt.Errorf("cannot convert string into map key type %s", keyType)
-		}
-		k = k.Convert(keyType)
-	}
-
-	if !v.MapIndex(k).IsValid() {
-		newElem := reflect.New(v.Type().Elem())
-		v.SetMapIndex(k, newElem.Elem())
-	}
-
-	return mapTarget{
-		v: v,
-		k: k,
-	}, true, nil
-}
-
-type fieldPathsMap = map[string][]int
-
-type fieldPathsCache struct {
-	m map[reflect.Type]fieldPathsMap
-	l sync.RWMutex
-}
-
-func (c *fieldPathsCache) get(t reflect.Type) (fieldPathsMap, bool) {
-	c.l.RLock()
-	paths, ok := c.m[t]
-	c.l.RUnlock()
-	return paths, ok
-}
-
-func (c *fieldPathsCache) set(t reflect.Type, m fieldPathsMap) {
-	c.l.Lock()
-	c.m[t] = m
-	c.l.Unlock()
-}
-
-var globalFieldPathsCache = fieldPathsCache{
-	m: map[reflect.Type]fieldPathsMap{},
-	l: sync.RWMutex{},
-}
-
-func scopeStruct(v reflect.Value, name string) (target, bool, error) {
-	// TODO: cache this, and reduce allocations
-
-	fieldPaths, ok := globalFieldPathsCache.get(v.Type())
-	if !ok {
-		fieldPaths = map[string][]int{}
-
-		path := make([]int, 0, 16)
-		var walk func(reflect.Value)
-		walk = func(v reflect.Value) {
-			t := v.Type()
-			for i := 0; i < t.NumField(); i++ {
-				l := len(path)
-				path = append(path, i)
-				f := t.Field(i)
-				if f.PkgPath != "" {
-					// only consider exported fields
-				} else if f.Anonymous {
-					walk(v.Field(i))
-				} else {
-					fieldName, ok := f.Tag.Lookup("toml")
-					if !ok {
-						fieldName = f.Name
-					}
-
-					pathCopy := make([]int, len(path))
-					copy(pathCopy, path)
-
-					fieldPaths[fieldName] = pathCopy
-					// extra copy for the case-insensitive match
-					fieldPaths[strings.ToLower(fieldName)] = pathCopy
-				}
-				path = path[:l]
-			}
-		}
-
-		walk(v)
-
-		globalFieldPathsCache.set(v.Type(), fieldPaths)
-	}
-
-	path, ok := fieldPaths[name]
-	if !ok {
-		path, ok = fieldPaths[strings.ToLower(name)]
-	}
-	if !ok {
-		return nil, false, nil
-	}
-
-	return valueTarget(v.FieldByIndex(path)), true, nil
-}

targets_test.go

@@ -1,184 +0,0 @@
-package toml
-
-import (
-	"reflect"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestStructTarget_Ensure(t *testing.T) {
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		test  func(v reflect.Value, err error)
-	}{
-		{
-			desc:  "handle a nil slice of string",
-			input: reflect.ValueOf(&struct{ A []string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				assert.False(t, v.IsNil())
-			},
-		},
-		{
-			desc:  "handle an existing slice of string",
-			input: reflect.ValueOf(&struct{ A []string }{A: []string{"foo"}}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				require.False(t, v.IsNil())
-				s := v.Interface().([]string)
-				assert.Equal(t, []string{"foo"}, s)
-			},
-		},
-	}
-
-	for _, e := range examples {
-		t.Run(e.desc, func(t *testing.T) {
-			d := decoder{}
-			target, _, err := d.scopeTableTarget(false, valueTarget(e.input), e.name)
-			require.NoError(t, err)
-			err = ensureValueIndexable(target)
-			v := target.get()
-			e.test(v, err)
-		})
-	}
-}
-
-func TestStructTarget_SetString(t *testing.T) {
-	str := "value"
-
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		test  func(v reflect.Value, err error)
-	}{
-		{
-			desc:  "sets a string",
-			input: reflect.ValueOf(&struct{ A string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				assert.Equal(t, str, v.String())
-			},
-		},
-		{
-			desc:  "fails on a float",
-			input: reflect.ValueOf(&struct{ A float64 }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.Error(t, err)
-			},
-		},
-		{
-			desc:  "fails on a slice",
-			input: reflect.ValueOf(&struct{ A []string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.Error(t, err)
-			},
-		},
-	}
-
-	for _, e := range examples {
-		t.Run(e.desc, func(t *testing.T) {
-			d := decoder{}
-			target, _, err := d.scopeTableTarget(false, valueTarget(e.input), e.name)
-			require.NoError(t, err)
-			err = setString(target, str)
-			v := target.get()
-			e.test(v, err)
-		})
-	}
-}
-
-func TestPushNew(t *testing.T) {
-	t.Run("slice of strings", func(t *testing.T) {
-		type Doc struct {
-			A []string
-		}
-		d := Doc{}
-
-		dec := decoder{}
-		x, _, err := dec.scopeTableTarget(false, valueTarget(reflect.ValueOf(&d).Elem()), "A")
-		require.NoError(t, err)
-
-		n, err := elementAt(x, 0)
-		require.NoError(t, err)
-		require.NoError(t, n.setString("hello"))
-		require.Equal(t, []string{"hello"}, d.A)
-
-		n, err = elementAt(x, 1)
-		require.NoError(t, err)
-		require.NoError(t, n.setString("world"))
-		require.Equal(t, []string{"hello", "world"}, d.A)
-	})
-
-	t.Run("slice of interfaces", func(t *testing.T) {
-		type Doc struct {
-			A []interface{}
-		}
-		d := Doc{}
-
-		dec := decoder{}
-		x, _, err := dec.scopeTableTarget(false, valueTarget(reflect.ValueOf(&d).Elem()), "A")
-		require.NoError(t, err)
-
-		n, err := elementAt(x, 0)
-		require.NoError(t, err)
-		require.NoError(t, setString(n, "hello"))
-		require.Equal(t, []interface{}{"hello"}, d.A)
-
-		n, err = elementAt(x, 1)
-		require.NoError(t, err)
-		require.NoError(t, setString(n, "world"))
-		require.Equal(t, []interface{}{"hello", "world"}, d.A)
-	})
-}
-
-func TestScope_Struct(t *testing.T) {
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		err   bool
-		found bool
-		idx   []int
-	}{
-		{
-			desc:  "simple field",
-			input: reflect.ValueOf(&struct{ A string }{}).Elem(),
-			name:  "A",
-			idx:   []int{0},
-			found: true,
-		},
-		{
-			desc:  "fails not-exported field",
-			input: reflect.ValueOf(&struct{ a string }{}).Elem(),
-			name:  "a",
-			err:   false,
-			found: false,
-		},
-	}
-
-	for _, e := range examples {
-		t.Run(e.desc, func(t *testing.T) {
-			dec := decoder{}
-			x, found, err := dec.scopeTableTarget(false, valueTarget(e.input), e.name)
-			assert.Equal(t, e.found, found)
-			if e.err {
-				assert.Error(t, err)
-			}
-			if found {
-				x2, ok := x.(valueTarget)
-				require.True(t, ok)
-				x2.get()
-			}
-		})
-	}
-}

toml.abnf

@@ -59,6 +59,7 @@ val = string / boolean / array / inline-table / date-time / float / integer
 ;; String
 
 string = ml-basic-string / basic-string / ml-literal-string / literal-string
+
 ;; Basic String
 
 basic-string = quotation-mark *basic-char quotation-mark

toml_testgen_support_test.go

@@ -1,14 +1,14 @@
+//go:generate go run ./cmd/tomltestgen/main.go -o toml_testgen_test.go
+
 // This is a support file for toml_testgen_test.go
 package toml_test
 
 import (
 	"encoding/json"
-	"fmt"
-	"strconv"
 	"testing"
-	"time"
 
 	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/internal/testsuite"
 	"github.com/stretchr/testify/require"
 )
 
@@ -17,10 +17,14 @@ func testgenInvalid(t *testing.T, input string) {
 	t.Logf("Input TOML:\n%s", input)
 
 	doc := map[string]interface{}{}
-	err := toml.Unmarshal([]byte(input), &doc)
+	err := testsuite.Unmarshal([]byte(input), &doc)
 
 	if err == nil {
-		t.Log(json.Marshal(doc))
+		out, err := json.Marshal(doc)
+		if err != nil {
+			panic("could not marshal map to json")
+		}
+		t.Log("JSON output from unmarshal:", string(out))
 		t.Fatalf("test did not fail")
 	}
 }
@@ -29,110 +33,26 @@ func testgenValid(t *testing.T, input string, jsonRef string) {
 	t.Helper()
 	t.Logf("Input TOML:\n%s", input)
 
-	doc := map[string]interface{}{}
-	err := toml.Unmarshal([]byte(input), &doc)
+	// TODO: change this to interface{}
+	var doc map[string]interface{}
+
+	err := testsuite.Unmarshal([]byte(input), &doc)
 	if err != nil {
+		if de, ok := err.(*toml.DecodeError); ok {
+			t.Logf("%s\n%s", err, de)
+		}
 		t.Fatalf("failed parsing toml: %s", err)
 	}
-
-	refDoc := testgenBuildRefDoc(jsonRef)
-
-	require.Equal(t, refDoc, doc)
-
-	out, err := toml.Marshal(doc)
+	j, err := testsuite.ValueToTaggedJSON(doc)
 	require.NoError(t, err)
 
-	doc2 := map[string]interface{}{}
-	err = toml.Unmarshal(out, &doc2)
+	var ref interface{}
+	err = json.Unmarshal([]byte(jsonRef), &ref)
 	require.NoError(t, err)
 
-	require.Equal(t, refDoc, doc2)
-}
-
-type testGenDescNode struct {
-	Type  string
-	Value interface{}
-}
-
-func testgenBuildRefDoc(jsonRef string) map[string]interface{} {
-	descTree := map[string]interface{}{}
-	err := json.Unmarshal([]byte(jsonRef), &descTree)
-	if err != nil {
-		panic(fmt.Errorf("reference doc should be valid JSON: %s", err))
-	}
-
-	doc := testGenTranslateDesc(descTree)
-	if doc == nil {
-		return map[string]interface{}{}
-	}
-	return doc.(map[string]interface{})
-}
-
-func testGenTranslateDesc(input interface{}) interface{} {
-	a, ok := input.([]interface{})
-	if ok {
-		xs := make([]interface{}, len(a))
-		for i, v := range a {
-			xs[i] = testGenTranslateDesc(v)
-		}
-		return xs
-	}
-
-	d := input.(map[string]interface{})
-
-	var dtype string
-	var dvalue interface{}
-
-	if len(d) == 2 {
-		dtypeiface, ok := d["type"]
-		if ok {
-			dvalue, ok = d["value"]
-			if ok {
-				dtype = dtypeiface.(string)
-				switch dtype {
-				case "string":
-					return dvalue.(string)
-				case "float":
-					v, err := strconv.ParseFloat(dvalue.(string), 64)
-					if err != nil {
-						panic(fmt.Errorf("invalid float '%s': %s", dvalue, err))
-					}
-					return v
-				case "integer":
-					v, err := strconv.ParseInt(dvalue.(string), 10, 64)
-					if err != nil {
-						panic(fmt.Errorf("invalid int '%s': %s", dvalue, err))
-					}
-					return v
-				case "bool":
-					return dvalue.(string) == "true"
-				case "datetime":
-					dt, err := time.Parse("2006-01-02T15:04:05Z", dvalue.(string))
-					if err != nil {
-						panic(fmt.Errorf("invalid datetime '%s': %s", dvalue, err))
-					}
-					return dt
-				case "array":
-					if dvalue == nil {
-						return nil
-					}
-					a := dvalue.([]interface{})
-					xs := make([]interface{}, len(a))
-
-					for i, v := range a {
-						xs[i] = testGenTranslateDesc(v)
-					}
-
-					return xs
-				}
-				panic(fmt.Errorf("unknown type: %s", dtype))
-			}
-		}
-	}
-
-	dest := map[string]interface{}{}
-	for k, v := range d {
-		dest[k] = testGenTranslateDesc(v)
-	}
-	return dest
+	var actual interface{}
+	err = json.Unmarshal([]byte(j), &actual)
+	require.NoError(t, err)
+
+	testsuite.CmpJSON(t, "", ref, actual)
 }

types.go

@@ -0,0 +1,14 @@
+package toml
+
+import (
+	"encoding"
+	"reflect"
+	"time"
+)
+
+var timeType = reflect.TypeOf(time.Time{})
+var textMarshalerType = reflect.TypeOf(new(encoding.TextMarshaler)).Elem()
+var textUnmarshalerType = reflect.TypeOf(new(encoding.TextUnmarshaler)).Elem()
+var mapStringInterfaceType = reflect.TypeOf(map[string]interface{}{})
+var sliceInterfaceType = reflect.TypeOf([]interface{}{})
+var stringType = reflect.TypeOf("")

utf8.go

@@ -0,0 +1,240 @@
+package toml
+
+import (
+	"unicode/utf8"
+)
+
+type utf8Err struct {
+	Index int
+	Size  int
+}
+
+func (u utf8Err) Zero() bool {
+	return u.Size == 0
+}
+
+// Verified that a given string is only made of valid UTF-8 characters allowed
+// by the TOML spec:
+//
+// Any Unicode character may be used except those that must be escaped:
+// quotation mark, backslash, and the control characters other than tab (U+0000
+// to U+0008, U+000A to U+001F, U+007F).
+//
+// It is a copy of the Go 1.17 utf8.Valid implementation, tweaked to exit early
+// when a character is not allowed.
+//
+// The returned utf8Err is Zero() if the string is valid, or contains the byte
+// index and size of the invalid character.
+//
+// quotation mark => already checked
+// backslash => already checked
+// 0-0x8 => invalid
+// 0x9 => tab, ok
+// 0xA - 0x1F => invalid
+// 0x7F => invalid
+func utf8TomlValidAlreadyEscaped(p []byte) (err utf8Err) {
+	// Fast path. Check for and skip 8 bytes of ASCII characters per iteration.
+	offset := 0
+	for len(p) >= 8 {
+		// Combining two 32 bit loads allows the same code to be used
+		// for 32 and 64 bit platforms.
+		// The compiler can generate a 32bit load for first32 and second32
+		// on many platforms. See test/codegen/memcombine.go.
+		first32 := uint32(p[0]) | uint32(p[1])<<8 | uint32(p[2])<<16 | uint32(p[3])<<24
+		second32 := uint32(p[4]) | uint32(p[5])<<8 | uint32(p[6])<<16 | uint32(p[7])<<24
+		if (first32|second32)&0x80808080 != 0 {
+			// Found a non ASCII byte (>= RuneSelf).
+			break
+		}
+
+		for i, b := range p[:8] {
+			if invalidAscii(b) {
+				err.Index = offset + i
+				err.Size = 1
+				return
+			}
+		}
+
+		p = p[8:]
+		offset += 8
+	}
+	n := len(p)
+	for i := 0; i < n; {
+		pi := p[i]
+		if pi < utf8.RuneSelf {
+			if invalidAscii(pi) {
+				err.Index = offset + i
+				err.Size = 1
+				return
+			}
+			i++
+			continue
+		}
+		x := first[pi]
+		if x == xx {
+			// Illegal starter byte.
+			err.Index = offset + i
+			err.Size = 1
+			return
+		}
+		size := int(x & 7)
+		if i+size > n {
+			// Short or invalid.
+			err.Index = offset + i
+			err.Size = n - i
+			return
+		}
+		accept := acceptRanges[x>>4]
+		if c := p[i+1]; c < accept.lo || accept.hi < c {
+			err.Index = offset + i
+			err.Size = 2
+			return
+		} else if size == 2 {
+		} else if c := p[i+2]; c < locb || hicb < c {
+			err.Index = offset + i
+			err.Size = 3
+			return
+		} else if size == 3 {
+		} else if c := p[i+3]; c < locb || hicb < c {
+			err.Index = offset + i
+			err.Size = 4
+			return
+		}
+		i += size
+	}
+	return
+}
+
+// Return the size of the next rune if valid, 0 otherwise.
+func utf8ValidNext(p []byte) int {
+	c := p[0]
+
+	if c < utf8.RuneSelf {
+		if invalidAscii(c) {
+			return 0
+		}
+		return 1
+	}
+
+	x := first[c]
+	if x == xx {
+		// Illegal starter byte.
+		return 0
+	}
+	size := int(x & 7)
+	if size > len(p) {
+		// Short or invalid.
+		return 0
+	}
+	accept := acceptRanges[x>>4]
+	if c := p[1]; c < accept.lo || accept.hi < c {
+		return 0
+	} else if size == 2 {
+	} else if c := p[2]; c < locb || hicb < c {
+		return 0
+	} else if size == 3 {
+	} else if c := p[3]; c < locb || hicb < c {
+		return 0
+	}
+
+	return size
+}
+
+var invalidAsciiTable = [256]bool{
+	0x00: true,
+	0x01: true,
+	0x02: true,
+	0x03: true,
+	0x04: true,
+	0x05: true,
+	0x06: true,
+	0x07: true,
+	0x08: true,
+	// 0x09 TAB
+	// 0x0A LF
+	0x0B: true,
+	0x0C: true,
+	// 0x0D CR
+	0x0E: true,
+	0x0F: true,
+	0x10: true,
+	0x11: true,
+	0x12: true,
+	0x13: true,
+	0x14: true,
+	0x15: true,
+	0x16: true,
+	0x17: true,
+	0x18: true,
+	0x19: true,
+	0x1A: true,
+	0x1B: true,
+	0x1C: true,
+	0x1D: true,
+	0x1E: true,
+	0x1F: true,
+	// 0x20 - 0x7E Printable ASCII characters
+	0x7F: true,
+}
+
+func invalidAscii(b byte) bool {
+	return invalidAsciiTable[b]
+}
+
+// acceptRange gives the range of valid values for the second byte in a UTF-8
+// sequence.
+type acceptRange struct {
+	lo uint8 // lowest value for second byte.
+	hi uint8 // highest value for second byte.
+}
+
+// acceptRanges has size 16 to avoid bounds checks in the code that uses it.
+var acceptRanges = [16]acceptRange{
+	0: {locb, hicb},
+	1: {0xA0, hicb},
+	2: {locb, 0x9F},
+	3: {0x90, hicb},
+	4: {locb, 0x8F},
+}
+
+// first is information about the first byte in a UTF-8 sequence.
+var first = [256]uint8{
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x00-0x0F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x10-0x1F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x20-0x2F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x30-0x3F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x40-0x4F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x50-0x5F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x60-0x6F
+	as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, as, // 0x70-0x7F
+	//   1   2   3   4   5   6   7   8   9   A   B   C   D   E   F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x80-0x8F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0x90-0x9F
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xA0-0xAF
+	xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xB0-0xBF
+	xx, xx, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xC0-0xCF
+	s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, s1, // 0xD0-0xDF
+	s2, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s3, s4, s3, s3, // 0xE0-0xEF
+	s5, s6, s6, s6, s7, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, xx, // 0xF0-0xFF
+}
+
+const (
+	// The default lowest and highest continuation byte.
+	locb = 0b10000000
+	hicb = 0b10111111
+
+	// These names of these constants are chosen to give nice alignment in the
+	// table below. The first nibble is an index into acceptRanges or F for
+	// special one-byte cases. The second nibble is the Rune length or the
+	// Status for the special one-byte case.
+	xx = 0xF1 // invalid: size 1
+	as = 0xF0 // ASCII: size 1
+	s1 = 0x02 // accept 0, size 2
+	s2 = 0x13 // accept 1, size 3
+	s3 = 0x03 // accept 0, size 3
+	s4 = 0x23 // accept 2, size 3
+	s5 = 0x34 // accept 3, size 4
+	s6 = 0x04 // accept 0, size 4
+	s7 = 0x44 // accept 4, size 4
+)