encode: fix commented table with comment (#894)

And updated README.

.gitattributes

@@ -1,3 +1,4 @@
 * text=auto
 
 benchmark/benchmark.toml text eol=lf
+testdata/** text eol=lf

.github/workflows/cifuzz.yml

@@ -0,0 +1,26 @@
+name: CIFuzz
+on: [pull_request]
+jobs:
+  Fuzzing:
+    runs-on: ubuntu-latest
+    steps:
+    - name: Build Fuzzers
+      id: build
+      uses: google/oss-fuzz/infra/cifuzz/actions/build_fuzzers@master
+      with:
+        oss-fuzz-project-name: 'go-toml'
+        dry-run: false
+        language: go
+    - name: Run Fuzzers
+      uses: google/oss-fuzz/infra/cifuzz/actions/run_fuzzers@master
+      with:
+        oss-fuzz-project-name: 'go-toml'
+        fuzz-seconds: 300
+        dry-run: false
+        language: go
+    - name: Upload Crash
+      uses: actions/upload-artifact@v3
+      if: failure() && steps.build.outcome == 'success'
+      with:
+        name: artifacts
+        path: ./out/artifacts

.github/workflows/codeql-analysis.yml

@@ -35,11 +35,11 @@ jobs:
 
     steps:
     - name: Checkout repository
-      uses: actions/checkout@v2
+      uses: actions/checkout@v3
 
     # Initializes the CodeQL tools for scanning.
     - name: Initialize CodeQL
-      uses: github/codeql-action/init@v1
+      uses: github/codeql-action/init@v2
       with:
         languages: ${{ matrix.language }}
         # If you wish to specify custom queries, you can do so here or in a config file.
@@ -50,7 +50,7 @@ jobs:
     # Autobuild attempts to build any compiled languages  (C/C++, C#, or Java).
     # If this step fails, then you should remove it and run the build manually (see below)
     - name: Autobuild
-      uses: github/codeql-action/autobuild@v1
+      uses: github/codeql-action/autobuild@v2
 
     # ℹ️ Command-line programs to run using the OS shell.
     # 📚 https://git.io/JvXDl
@@ -64,4 +64,4 @@ jobs:
     #   make release
 
     - name: Perform CodeQL Analysis
-      uses: github/codeql-action/analyze@v1
+      uses: github/codeql-action/analyze@v2

.github/workflows/coverage.yml

@@ -9,12 +9,12 @@ jobs:
     runs-on: "ubuntu-latest"
     name: report
     steps:
-      - uses: actions/checkout@master
+      - uses: actions/checkout@v3
         with:
           fetch-depth: 0
       - name: Setup go
-        uses: actions/setup-go@master
+        uses: actions/setup-go@v4
         with:
-          go-version: 1.16
+          go-version: "1.21"
       - 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@v3
+        with:
+          fetch-depth: 0
+      - name: Set up Go
+        uses: actions/setup-go@v4
+        with:
+          go-version: "1.21"
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v2
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Run GoReleaser
+        uses: goreleaser/goreleaser-action@v3
+        with:
+          distribution: goreleaser
+          version: latest
+          args: release ${{ inputs.args }} --rm-dist
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/workflow.yml

@@ -11,15 +11,22 @@ jobs:
   build:
     strategy:
       matrix:
-        os: [ 'ubuntu-latest', 'windows-latest', 'macos-latest']
+        os: [ 'ubuntu-latest', 'windows-latest', 'macos-latest' ]
-        go: [ '1.16', '1.17' ]
+        go: [ '1.20', '1.21' ]
     runs-on: ${{ matrix.os }}
     name: ${{ matrix.go }}/${{ matrix.os }}
     steps:
-      - uses: actions/checkout@master
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
       - name: Setup go ${{ matrix.go }}
-        uses: actions/setup-go@master
+        uses: actions/setup-go@v4
         with:
           go-version: ${{ matrix.go }}
       - name: Run unit tests
         run: go test -race ./...
+  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

.goreleaser.yaml

@@ -0,0 +1,126 @@
+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
+      - linux_arm64
+      - linux_arm
+      - linux_riscv64
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - 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
+      - linux_arm64
+      - linux_arm
+      - linux_riscv64
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - 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
+      - linux_arm64
+      - linux_riscv64
+      - linux_arm
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - 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

@@ -155,6 +155,8 @@ Checklist:
 - 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
@@ -163,13 +165,25 @@ Checklist:
 
 ### New release
 
-1. Go to [releases][releases]. Click on "X commits to master since this
+1. Decide on the next version number. Use semver.
-   release".
+2. Generate release notes using [`gh`][gh]. Example:
-2. Make note of all the changes. Look for backward incompatible changes,
+```
-   new features, and bug fixes.
+$ gh api -X POST \
-3. Pick the new version using the above and semver.
+  -F tag_name='v2.0.0-beta.5' \
-4. Create a [new release][new-release].
+  -F target_commitish='v2' \
-5. Follow the same format as [1.1.0][release-110].
+  -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
@@ -177,6 +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

LICENSE

@@ -1,6 +1,7 @@
 The MIT License (MIT)
 
-Copyright (c) 2013 - 2021 Thomas Pelletier, Eric Anderton
+go-toml v2
+Copyright (c) 2021 - 2023 Thomas Pelletier
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -4,24 +4,14 @@ Go library for the [TOML](https://toml.io/en/) format.
 
 This library supports [TOML v1.0.0](https://toml.io/en/v1.0.0).
 
-## Development status
-
-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.
-
-If you do not need the advanced document editing features of v1, you are
-encouraged to try out this version.
-
-[👉 Roadmap for v2](https://github.com/pelletier/go-toml/discussions/506)
-
 [🐞 Bug Reports](https://github.com/pelletier/go-toml/issues)
 
 [💬 Anything else](https://github.com/pelletier/go-toml/discussions)
 
 ## Documentation
 
-Full API, examples, and implementation notes are available in the Go documentation.
+Full API, examples, and implementation notes are available in the Go
+documentation.
 
 [![Go Reference](https://pkg.go.dev/badge/github.com/pelletier/go-toml/v2.svg)](https://pkg.go.dev/github.com/pelletier/go-toml/v2)
 
@@ -48,23 +38,22 @@ 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
+the TOML document was not present 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.SetStrict
+[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]),
+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"
+1| [server]
-3| key2 = "missing2"
+2| path = 100
- | ~~~~ missing field
+ |        ~~~ cannot decode TOML integer into struct field toml_test.Server.Path of type string
-4| key3 = "missing3"
+3| port = 50
-5| key4 = "value4"
 ```
 
 [decode-err]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#DecodeError
@@ -83,6 +72,26 @@ representation.
 [tlt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalTime
 [tldt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalDateTime
 
+### Commented config
+
+Since TOML is often used for configuration files, go-toml can emit documents
+annotated with [comments and commented-out values][comments-example]. For
+example, it can generate the following file:
+
+```toml
+# Host IP to connect to.
+host = '127.0.0.1'
+# Port of the remote server.
+port = 4242
+
+# Encryption parameters (optional)
+# [TLS]
+# cipher = 'AEAD-AES128-GCM-SHA256'
+# version = 'TLS 1.3'
+```
+
+[comments-example]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#example-Marshal-Commented
+
 ## Getting started
 
 Given the following struct, let's see how to read it and write it as TOML:
@@ -150,6 +159,17 @@ fmt.Println(string(b))
 
 [marshal]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Marshal
 
+## Unstable API
+
+This API does not yet follow the backward compatibility guarantees of this
+library. They provide early access to features that may have rough edges or an
+API subject to change.
+
+### Parser
+
+Parser is the unstable API that allows iterative parsing of a TOML document at
+the AST level. See https://pkg.go.dev/github.com/pelletier/go-toml/v2/unstable.
+
 ## Benchmarks
 
 Execution time speedup compared to other Go TOML libraries:
@@ -160,11 +180,11 @@ Execution time speedup compared to other Go TOML libraries:
     </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/map-2</td><td>1.7x</td><td>1.8x</td></tr>
-        <tr><td>Marshal/ReferenceFile/struct-2</td><td>2.4x</td><td>2.6x</td></tr>
+        <tr><td>Marshal/ReferenceFile/struct-2</td><td>2.2x</td><td>2.5x</td></tr>
-        <tr><td>Unmarshal/HugoFrontMatter-2</td><td>2.9x</td><td>2.5x</td></tr>
+        <tr><td>Unmarshal/HugoFrontMatter-2</td><td>2.9x</td><td>2.9x</td></tr>
-        <tr><td>Unmarshal/ReferenceFile/map-2</td><td>2.7x</td><td>2.6x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/map-2</td><td>2.6x</td><td>2.9x</td></tr>
-        <tr><td>Unmarshal/ReferenceFile/struct-2</td><td>4.8x</td><td>5.1x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/struct-2</td><td>4.4x</td><td>5.3x</td></tr>
      </tbody>
 </table>
 <details><summary>See more</summary>
@@ -177,17 +197,17 @@ provided for completeness.</p>
         <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/map-2</td><td>1.8x</td><td>2.9x</td></tr>
-        <tr><td>Marshal/SimpleDocument/struct-2</td><td>2.5x</td><td>2.8x</td></tr>
+        <tr><td>Marshal/SimpleDocument/struct-2</td><td>2.7x</td><td>4.2x</td></tr>
-        <tr><td>Unmarshal/SimpleDocument/map-2</td><td>4.1x</td><td>3.1x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/map-2</td><td>4.5x</td><td>3.1x</td></tr>
-        <tr><td>Unmarshal/SimpleDocument/struct-2</td><td>6.4x</td><td>4.3x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/struct-2</td><td>6.2x</td><td>3.9x</td></tr>
-        <tr><td>UnmarshalDataset/example-2</td><td>3.4x</td><td>3.2x</td></tr>
+        <tr><td>UnmarshalDataset/example-2</td><td>3.1x</td><td>3.5x</td></tr>
-        <tr><td>UnmarshalDataset/code-2</td><td>2.2x</td><td>2.5x</td></tr>
+        <tr><td>UnmarshalDataset/code-2</td><td>2.3x</td><td>3.1x</td></tr>
-        <tr><td>UnmarshalDataset/twitter-2</td><td>2.8x</td><td>2.7x</td></tr>
+        <tr><td>UnmarshalDataset/twitter-2</td><td>2.5x</td><td>2.6x</td></tr>
-        <tr><td>UnmarshalDataset/citm_catalog-2</td><td>2.2x</td><td>2.0x</td></tr>
+        <tr><td>UnmarshalDataset/citm_catalog-2</td><td>2.1x</td><td>2.2x</td></tr>
-        <tr><td>UnmarshalDataset/canada-2</td><td>1.8x</td><td>1.4x</td></tr>
+        <tr><td>UnmarshalDataset/canada-2</td><td>1.6x</td><td>1.3x</td></tr>
-        <tr><td>UnmarshalDataset/config-2</td><td>4.4x</td><td>2.9x</td></tr>
+        <tr><td>UnmarshalDataset/config-2</td><td>4.3x</td><td>3.2x</td></tr>
-        <tr><td>[Geo mean]</td><td>2.8x</td><td>2.6x</td></tr>
+        <tr><td>[Geo mean]</td><td>2.7x</td><td>2.8x</td></tr>
      </tbody>
 </table>
 <p>This table can be generated with <code>./ci.sh benchmark -a -html</code>.</p>
@@ -207,6 +227,44 @@ 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
@@ -458,27 +516,20 @@ is not necessary anymore.
 
 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:
+`toml`, `multiline`, `commented`, and `omitempty`. For example:
 
 ```go
 type doc struct {
 	// v1
-	F string `toml:"field" multiline:"true" omitempty:"true"`
+	F string `toml:"field" multiline:"true" omitempty:"true" commented:"true"`
 	// v2
-	F string `toml:"field,multiline,omitempty"`
+	F string `toml:"field,multiline,omitempty,commented"`
 }
 ```
 
 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.
@@ -488,45 +539,37 @@ The new name is `Encoder.SetArraysMultiline`. The behavior should be the same.
 The new name is `Encoder.SetIndentSymbol`. The behavior should be the same.
 
 
-#### Embedded structs are tables
+#### 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. There is no way to revert
+a problem anymore, v2 does the right thing by default: it follows the behavior
-to the old behavior, and `Encoder.PromoteAnonymous` has been removed.
+of `encoding/json`. `Encoder.PromoteAnonymous` has been removed.
-
-```go
-type Embedded struct {
-	Value string `toml:"value"`
-}
-
-type Doc struct {
-	Embedded
-}
-
-d := Doc{}
-
-fmt.Println("v1:")
-b, err := v1.Marshal(d)
-fmt.Println(string(b))
-
-fmt.Println("v2:")
-b, err = v2.Marshal(d)
-fmt.Println(string(b))
-
-// Output:
-// v1:
-// value = ""
-//
-// v2:
-// [Embedded]
-// value = ''
-```
 
 [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
+
+## Versioning
+
+Go-toml follows [Semantic Versioning](https://semver.org). The supported version
+of [TOML](https://github.com/toml-lang/toml) is indicated at the beginning of
+this document. The last two major versions of Go are supported
+(see [Go Release Policy](https://golang.org/doc/devel/release.html#policy)).
+
 ## License
 
 The MIT License (MIT). Read [LICENSE](LICENSE).

ci.sh

@@ -76,8 +76,10 @@ cover() {
     fi
 
     pushd "$dir"
-    go test -covermode=atomic -coverprofile=coverage.out ./...
+    go test -covermode=atomic  -coverpkg=./... -coverprofile=coverage.out.tmp ./...
+    cat coverage.out.tmp | grep -v fuzz | grep -v testsuite | grep -v tomltestgen | grep -v gotoml-test-decoder > coverage.out
     go tool cover -func=coverage.out
+    echo "Coverage profile for ${branch}: ${dir}/coverage.out" >&2
     popd
 
     if [ "${branch}" != "HEAD" ]; then
@@ -103,16 +105,23 @@ coverage() {
 
 	    echo ""
 
-	    target_pct="$(cat ${target_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    target_pct="$(tail -n2 ${target_out} | head -n1 | sed -E 's/.*total.*\t([0-9.]+)%.*/\1/')"
-	    head_pct="$(cat ${head_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    head_pct="$(tail -n2 ${head_out} | head -n1 | sed -E 's/.*total.*\t([0-9.]+)%/\1/')"
 	    echo "Results: ${target} ${target_pct}% HEAD ${head_pct}%"
 
 	    delta_pct=$(echo "$head_pct - $target_pct" | bc -l)
 	    echo "Delta: ${delta_pct}"
 
 	    if [[ $delta_pct = \-* ]]; then
-		echo "Regression!";
+		    echo "Regression!";
-		return 1
+
+            target_diff="${output_dir}/target.diff.txt"
+            head_diff="${output_dir}/head.diff.txt"
+            cat "${target_out}" | grep -E '^github.com/pelletier/go-toml' | tr -s "\t " | cut -f 2,3 | sort > "${target_diff}"
+            cat "${head_out}" | grep -E '^github.com/pelletier/go-toml' | tr -s "\t " | cut -f 2,3 | sort > "${head_diff}"
+
+            diff --side-by-side --suppress-common-lines "${target_diff}" "${head_diff}"
+		    return 1
 	    fi
 	    return 0
 	    ;;

cmd/gotoml-test-decoder/main.go

@@ -6,7 +6,7 @@ import (
 	"os"
 	"path"
 
-	"github.com/pelletier/go-toml/v2/testsuite"
+	"github.com/pelletier/go-toml/v2/internal/testsuite"
 )
 
 func main() {

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 +1,48 @@
+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 +1,45 @@
+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

@@ -3,7 +3,7 @@
 //
 // Within the go-toml package, run `go generate`.  Otherwise, use:
 //
-//     go run github.com/pelletier/go-toml/cmd/tomltestgen -o toml_testgen_test.go
+//	go run github.com/pelletier/go-toml/cmd/tomltestgen -o toml_testgen_test.go
 package main
 
 import (

decode.go

@@ -5,6 +5,8 @@ import (
 	"math"
 	"strconv"
 	"time"
+
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 func parseInteger(b []byte) (int64, error) {
@@ -32,7 +34,7 @@ func parseLocalDate(b []byte) (LocalDate, error) {
 	var date LocalDate
 
 	if len(b) != 10 || b[4] != '-' || b[7] != '-' {
-		return date, newDecodeError(b, "dates are expected to have the format YYYY-MM-DD")
+		return date, unstable.NewParserError(b, "dates are expected to have the format YYYY-MM-DD")
 	}
 
 	var err error
@@ -53,7 +55,7 @@ func parseLocalDate(b []byte) (LocalDate, error) {
 	}
 
 	if !isValidDate(date.Year, date.Month, date.Day) {
-		return LocalDate{}, newDecodeError(b, "impossible date")
+		return LocalDate{}, unstable.NewParserError(b, "impossible date")
 	}
 
 	return date, nil
@@ -64,7 +66,7 @@ func parseDecimalDigits(b []byte) (int, error) {
 
 	for i, c := range b {
 		if c < '0' || c > '9' {
-			return 0, newDecodeError(b[i:i+1], "expected digit (0-9)")
+			return 0, unstable.NewParserError(b[i:i+1], "expected digit (0-9)")
 		}
 		v *= 10
 		v += int(c - '0')
@@ -97,7 +99,7 @@ func parseDateTime(b []byte) (time.Time, error) {
 	} else {
 		const dateTimeByteLen = 6
 		if len(b) != dateTimeByteLen {
-			return time.Time{}, newDecodeError(b, "invalid date-time timezone")
+			return time.Time{}, unstable.NewParserError(b, "invalid date-time timezone")
 		}
 		var direction int
 		switch b[0] {
@@ -106,11 +108,11 @@ func parseDateTime(b []byte) (time.Time, error) {
 		case '+':
 			direction = +1
 		default:
-			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset character")
+			return time.Time{}, unstable.NewParserError(b[:1], "invalid timezone offset character")
 		}
 
 		if b[3] != ':' {
-			return time.Time{}, newDecodeError(b[3:4], "expected a : separator")
+			return time.Time{}, unstable.NewParserError(b[3:4], "expected a : separator")
 		}
 
 		hours, err := parseDecimalDigits(b[1:3])
@@ -118,7 +120,7 @@ func parseDateTime(b []byte) (time.Time, error) {
 			return time.Time{}, err
 		}
 		if hours > 23 {
-			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset hours")
+			return time.Time{}, unstable.NewParserError(b[:1], "invalid timezone offset hours")
 		}
 
 		minutes, err := parseDecimalDigits(b[4:6])
@@ -126,16 +128,20 @@ func parseDateTime(b []byte) (time.Time, error) {
 			return time.Time{}, err
 		}
 		if minutes > 59 {
-			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset minutes")
+			return time.Time{}, unstable.NewParserError(b[:1], "invalid timezone offset minutes")
 		}
 
 		seconds := direction * (hours*3600 + minutes*60)
-		zone = time.FixedZone("", seconds)
+		if seconds == 0 {
+			zone = time.UTC
+		} else {
+			zone = time.FixedZone("", seconds)
+		}
 		b = b[dateTimeByteLen:]
 	}
 
 	if len(b) > 0 {
-		return time.Time{}, newDecodeError(b, "extra bytes at the end of the timezone")
+		return time.Time{}, unstable.NewParserError(b, "extra bytes at the end of the timezone")
 	}
 
 	t := time.Date(
@@ -156,7 +162,7 @@ func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
 
 	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]")
+		return dt, nil, unstable.NewParserError(b, "local datetimes are expected to have the format YYYY-MM-DDTHH:MM:SS[.NNNNNNNNN]")
 	}
 
 	date, err := parseLocalDate(b[:10])
@@ -167,7 +173,7 @@ func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
 
 	sep := b[10]
 	if sep != 'T' && sep != ' ' && sep != 't' {
-		return dt, nil, newDecodeError(b[10:11], "datetime separator is expected to be T or a space")
+		return dt, nil, unstable.NewParserError(b[10:11], "datetime separator is expected to be T or a space")
 	}
 
 	t, rest, err := parseLocalTime(b[11:])
@@ -191,7 +197,7 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 	// check if b matches to have expected format HH:MM:SS[.NNNNNN]
 	const localTimeByteLen = 8
 	if len(b) < localTimeByteLen {
-		return t, nil, newDecodeError(b, "times are expected to have the format HH:MM:SS[.NNNNNN]")
+		return t, nil, unstable.NewParserError(b, "times are expected to have the format HH:MM:SS[.NNNNNN]")
 	}
 
 	var err error
@@ -202,10 +208,10 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 	}
 
 	if t.Hour > 23 {
-		return t, nil, newDecodeError(b[0:2], "hour cannot be greater 23")
+		return t, nil, unstable.NewParserError(b[0:2], "hour cannot be greater 23")
 	}
 	if b[2] != ':' {
-		return t, nil, newDecodeError(b[2:3], "expecting colon between hours and minutes")
+		return t, nil, unstable.NewParserError(b[2:3], "expecting colon between hours and minutes")
 	}
 
 	t.Minute, err = parseDecimalDigits(b[3:5])
@@ -213,10 +219,10 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		return t, nil, err
 	}
 	if t.Minute > 59 {
-		return t, nil, newDecodeError(b[3:5], "minutes cannot be greater 59")
+		return t, nil, unstable.NewParserError(b[3:5], "minutes cannot be greater 59")
 	}
 	if b[5] != ':' {
-		return t, nil, newDecodeError(b[5:6], "expecting colon between minutes and seconds")
+		return t, nil, unstable.NewParserError(b[5:6], "expecting colon between minutes and seconds")
 	}
 
 	t.Second, err = parseDecimalDigits(b[6:8])
@@ -225,7 +231,7 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 	}
 
 	if t.Second > 60 {
-		return t, nil, newDecodeError(b[6:8], "seconds cannot be greater 60")
+		return t, nil, unstable.NewParserError(b[6:8], "seconds cannot be greater 60")
 	}
 
 	b = b[8:]
@@ -238,7 +244,7 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		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")
+					return t, nil, unstable.NewParserError(b[0:1], "need at least one digit after fraction point")
 				}
 				break
 			}
@@ -262,7 +268,7 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		}
 
 		if precision == 0 {
-			return t, nil, newDecodeError(b[:1], "nanoseconds need at least one digit")
+			return t, nil, unstable.NewParserError(b[:1], "nanoseconds need at least one digit")
 		}
 
 		t.Nanosecond = frac * nspow[precision]
@@ -285,40 +291,40 @@ func parseFloat(b []byte) (float64, error) {
 	}
 
 	if cleaned[0] == '.' {
-		return 0, newDecodeError(b, "float cannot start with a dot")
+		return 0, unstable.NewParserError(b, "float cannot start with a dot")
 	}
 
 	if cleaned[len(cleaned)-1] == '.' {
-		return 0, newDecodeError(b, "float cannot end with a dot")
+		return 0, unstable.NewParserError(b, "float cannot end with a dot")
 	}
 
 	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")
+				return 0, unstable.NewParserError(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")
+				return 0, unstable.NewParserError(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")
+				return 0, unstable.NewParserError(b[i:i+2], "float decimal point must be followed by a digit")
 			}
 			dotAlreadySeen = true
 		}
 	}
 
 	start := 0
-	if b[0] == '+' || b[0] == '-' {
+	if cleaned[0] == '+' || cleaned[0] == '-' {
 		start = 1
 	}
-	if b[start] == '0' && isDigit(b[start+1]) {
+	if cleaned[start] == '0' && len(cleaned) > start+1 && isDigit(cleaned[start+1]) {
-		return 0, newDecodeError(b, "float integer part cannot have leading zeroes")
+		return 0, unstable.NewParserError(b, "float integer part cannot have leading zeroes")
 	}
 
 	f, err := strconv.ParseFloat(string(cleaned), 64)
 	if err != nil {
-		return 0, newDecodeError(b, "unable to parse float: %w", err)
+		return 0, unstable.NewParserError(b, "unable to parse float: %w", err)
 	}
 
 	return f, nil
@@ -332,7 +338,7 @@ func parseIntHex(b []byte) (int64, error) {
 
 	i, err := strconv.ParseInt(string(cleaned), 16, 64)
 	if err != nil {
-		return 0, newDecodeError(b, "couldn't parse hexadecimal number: %w", err)
+		return 0, unstable.NewParserError(b, "couldn't parse hexadecimal number: %w", err)
 	}
 
 	return i, nil
@@ -346,7 +352,7 @@ func parseIntOct(b []byte) (int64, error) {
 
 	i, err := strconv.ParseInt(string(cleaned), 8, 64)
 	if err != nil {
-		return 0, newDecodeError(b, "couldn't parse octal number: %w", err)
+		return 0, unstable.NewParserError(b, "couldn't parse octal number: %w", err)
 	}
 
 	return i, nil
@@ -360,7 +366,7 @@ func parseIntBin(b []byte) (int64, error) {
 
 	i, err := strconv.ParseInt(string(cleaned), 2, 64)
 	if err != nil {
-		return 0, newDecodeError(b, "couldn't parse binary number: %w", err)
+		return 0, unstable.NewParserError(b, "couldn't parse binary number: %w", err)
 	}
 
 	return i, nil
@@ -383,12 +389,12 @@ func parseIntDec(b []byte) (int64, error) {
 	}
 
 	if len(cleaned) > startIdx+1 && cleaned[startIdx] == '0' {
-		return 0, newDecodeError(b, "leading zero not allowed on decimal number")
+		return 0, unstable.NewParserError(b, "leading zero not allowed on decimal number")
 	}
 
 	i, err := strconv.ParseInt(string(cleaned), 10, 64)
 	if err != nil {
-		return 0, newDecodeError(b, "couldn't parse decimal number: %w", err)
+		return 0, unstable.NewParserError(b, "couldn't parse decimal number: %w", err)
 	}
 
 	return i, nil
@@ -405,11 +411,11 @@ func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
 	}
 
 	if b[start] == '_' {
-		return nil, newDecodeError(b[start:start+1], "number cannot start with underscore")
+		return nil, unstable.NewParserError(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")
+		return nil, unstable.NewParserError(b[len(b)-1:], "number cannot end with underscore")
 	}
 
 	// fast path
@@ -431,7 +437,7 @@ func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
 		c := b[i]
 		if c == '_' {
 			if !before {
-				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+				return nil, unstable.NewParserError(b[i-1:i+1], "number must have at least one digit between underscores")
 			}
 			before = false
 		} else {
@@ -445,11 +451,11 @@ func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
 
 func checkAndRemoveUnderscoresFloats(b []byte) ([]byte, error) {
 	if b[0] == '_' {
-		return nil, newDecodeError(b[0:1], "number cannot start with underscore")
+		return nil, unstable.NewParserError(b[0:1], "number cannot start with underscore")
 	}
 
 	if b[len(b)-1] == '_' {
-		return nil, newDecodeError(b[len(b)-1:], "number cannot end with underscore")
+		return nil, unstable.NewParserError(b[len(b)-1:], "number cannot end with underscore")
 	}
 
 	// fast path
@@ -472,10 +478,10 @@ func checkAndRemoveUnderscoresFloats(b []byte) ([]byte, error) {
 		switch c {
 		case '_':
 			if !before {
-				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+				return nil, unstable.NewParserError(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")
+				return nil, unstable.NewParserError(b[i+1:i+2], "cannot have underscore before exponent")
 			}
 			before = false
 		case '+', '-':
@@ -484,15 +490,15 @@ func checkAndRemoveUnderscoresFloats(b []byte) ([]byte, error) {
 			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")
+				return nil, unstable.NewParserError(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")
+				return nil, unstable.NewParserError(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")
+				return nil, unstable.NewParserError(b[i-1:i], "cannot have underscore before decimal point")
 			}
 			cleaned = append(cleaned, c)
 		default:
@@ -538,3 +544,7 @@ func daysIn(m int, year int) int {
 func isLeap(year int) bool {
 	return year%4 == 0 && (year%100 != 0 || year%400 == 0)
 }
+
+func isDigit(r byte) bool {
+	return r >= '0' && r <= '9'
+}

errors.go

@@ -6,6 +6,7 @@ import (
 	"strings"
 
 	"github.com/pelletier/go-toml/v2/internal/danger"
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 // DecodeError represents an error encountered during the parsing or decoding
@@ -27,7 +28,7 @@ type DecodeError struct {
 // corresponding field in the target value. It contains all the missing fields
 // in Errors.
 //
-// Emitted by Decoder when SetStrict(true) was called.
+// Emitted by Decoder when DisallowUnknownFields() was called.
 type StrictMissingError struct {
 	// One error per field that could not be found.
 	Errors []DecodeError
@@ -55,25 +56,6 @@ func (s *StrictMissingError) String() 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 {
-	return de.message
-}
-
-func newDecodeError(highlight []byte, format string, args ...interface{}) error {
-	return &decodeError{
-		highlight: highlight,
-		message:   fmt.Errorf(format, args...).Error(),
-	}
-}
-
 // Error returns the error message contained in the DecodeError.
 func (e *DecodeError) Error() string {
 	return "toml: " + e.message
@@ -103,13 +85,14 @@ func (e *DecodeError) Key() Key {
 //
 // 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) *DecodeError {
+func wrapDecodeError(document []byte, de *unstable.ParserError) *DecodeError {
-	offset := danger.SubsliceOffset(document, de.highlight)
+	offset := danger.SubsliceOffset(document, de.Highlight)
 
 	errMessage := de.Error()
 	errLine, errColumn := positionAtEnd(document[:offset])
-	before, after := linesOfContext(document, de.highlight, offset, 3)
+	before, after := linesOfContext(document, de.Highlight, offset, 3)
 
 	var buf strings.Builder
 
@@ -139,7 +122,7 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 		buf.Write(before[0])
 	}
 
-	buf.Write(de.highlight)
+	buf.Write(de.Highlight)
 
 	if len(after) > 0 {
 		buf.Write(after[0])
@@ -157,7 +140,7 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 		buf.WriteString(strings.Repeat(" ", len(before[0])))
 	}
 
-	buf.WriteString(strings.Repeat("~", len(de.highlight)))
+	buf.WriteString(strings.Repeat("~", len(de.Highlight)))
 
 	if len(errMessage) > 0 {
 		buf.WriteString(" ")
@@ -182,7 +165,7 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 		message: errMessage,
 		line:    errLine,
 		column:  errColumn,
-		key:     de.key,
+		key:     de.Key,
 		human:   buf.String(),
 	}
 }

errors_test.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 	"testing"
 
+	"github.com/pelletier/go-toml/v2/unstable"
 	"github.com/stretchr/testify/assert"
 )
 
@@ -170,9 +171,9 @@ line 5`,
 			doc := b.Bytes()
 			hl := doc[start:end]
 
-			err := wrapDecodeError(doc, &decodeError{
+			err := wrapDecodeError(doc, &unstable.ParserError{
-				highlight: hl,
+				Highlight: hl,
-				message:   e.msg,
+				Message:   e.msg,
 			})
 
 			var derr *DecodeError

example_text_marshaling_test.go

@@ -0,0 +1,37 @@
+package toml_test
+
+import (
+	"fmt"
+	"log"
+	"strconv"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+type customInt int
+
+func (i *customInt) UnmarshalText(b []byte) error {
+	x, err := strconv.ParseInt(string(b), 10, 32)
+	if err != nil {
+		return err
+	}
+	*i = customInt(x * 100)
+	return nil
+}
+
+type doc struct {
+	Value customInt
+}
+
+func ExampleUnmarshal_textUnmarshal() {
+	var x doc
+
+	data := []byte(`value  = "42"`)
+	err := toml.Unmarshal(data, &x)
+	if err != nil {
+		log.Fatal(err)
+	}
+	fmt.Println(x)
+	// Output:
+	// {4200}
+}

fast_test.go

@@ -7,13 +7,20 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-func TestFastSimple(t *testing.T) {
+func TestFastSimpleInt(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 TestFastSimpleFloat(t *testing.T) {
+	m := map[string]float64{}
+	err := toml.Unmarshal([]byte("a = 42\nb = 1.1\nc = 12341234123412341234123412341234"), &m)
+	require.NoError(t, err)
+	require.Equal(t, map[string]float64{"a": 42, "b": 1.1, "c": 1.2341234123412342e+31}, m)
+}
+
 func TestFastSimpleString(t *testing.T) {
 	m := map[string]string{}
 	err := toml.Unmarshal([]byte(`a = "hello"`), &m)

fuzz_test.go

@@ -0,0 +1,56 @@
+//go:build go1.18 || go1.19 || go1.20 || go1.21
+// +build go1.18 go1.19 go1.20 go1.21
+
+package toml_test
+
+import (
+	"io/ioutil"
+	"strings"
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+func FuzzUnmarshal(f *testing.F) {
+	file, err := ioutil.ReadFile("benchmark/benchmark.toml")
+	if err != nil {
+		panic(err)
+	}
+	f.Add(file)
+
+	f.Fuzz(func(t *testing.T, b []byte) {
+		if strings.Contains(string(b), "nan") {
+			// Current limitation of testify.
+			// https://github.com/stretchr/testify/issues/624
+			t.Skip("can't compare NaNs")
+		}
+
+		t.Log("INITIAL DOCUMENT ===========================")
+		t.Log(string(b))
+
+		var v interface{}
+		err := toml.Unmarshal(b, &v)
+		if err != nil {
+			return
+		}
+
+		t.Log("DECODED VALUE ===========================")
+		t.Logf("%#+v", v)
+
+		encoded, err := toml.Marshal(v)
+		if err != nil {
+			t.Fatalf("cannot marshal unmarshaled document: %s", err)
+		}
+
+		t.Log("ENCODED DOCUMENT ===========================")
+		t.Log(string(encoded))
+
+		var v2 interface{}
+		err = toml.Unmarshal(encoded, &v2)
+		if err != nil {
+			t.Fatalf("failed round trip: %s", err)
+		}
+		require.Equal(t, v, v2)
+	})
+}

go.mod

@@ -2,5 +2,4 @@ module github.com/pelletier/go-toml/v2
 
 go 1.16
 
-// latest (v1.7.0) doesn't have the fix for time.Time
+require github.com/stretchr/testify v1.8.4
-require github.com/stretchr/testify v1.7.1-0.20210427113832-6241f9ab9942

go.sum

@@ -1,11 +1,17 @@
-github.com/davecgh/go-spew v1.1.0 h1:ZDRjVQ15GmhC3fiQ8ni8+OwkZQO4DARzQgrnXU1Liz8=
 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/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 v1.7.1-0.20210427113832-6241f9ab9942 h1:t0lM6y/M5IiUZyvbBTcngso8SZEZICH7is9B6g/obVU=
+github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
-github.com/stretchr/testify v1.7.1-0.20210427113832-6241f9ab9942/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
+github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO+kdMU+MU=
+github.com/stretchr/testify v1.8.4 h1:CcVxjf3Q8PM0mHUKJCdn+eZZtm5yQwehR5yeSVQQcUk=
+github.com/stretchr/testify v1.8.4/go.mod h1:sz/lmYIOXD/1dqDmKjjqLyZ2RngseejIcXlSw2iwfAo=
 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=
+gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
+gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

internal/ast/builder.go

@@ -1,51 +0,0 @@
-package ast
-
-type Reference int
-
-const InvalidReference Reference = -1
-
-func (r Reference) Valid() bool {
-	return r != InvalidReference
-}
-
-type Builder struct {
-	tree    Root
-	lastIdx int
-}
-
-func (b *Builder) Tree() *Root {
-	return &b.tree
-}
-
-func (b *Builder) NodeAt(ref Reference) *Node {
-	return b.tree.at(ref)
-}
-
-func (b *Builder) Reset() {
-	b.tree.nodes = b.tree.nodes[:0]
-	b.lastIdx = 0
-}
-
-func (b *Builder) Push(n Node) Reference {
-	b.lastIdx = len(b.tree.nodes)
-	b.tree.nodes = append(b.tree.nodes, n)
-	return Reference(b.lastIdx)
-}
-
-func (b *Builder) PushAndChain(n Node) Reference {
-	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.lastIdx
-	}
-	b.lastIdx = newIdx
-	return Reference(b.lastIdx)
-}
-
-func (b *Builder) AttachChild(parent Reference, child Reference) {
-	b.tree.nodes[parent].child = int(child) - int(parent)
-}
-
-func (b *Builder) Chain(from Reference, to Reference) {
-	b.tree.nodes[from].next = int(to) - int(from)
-}

internal/characters/ascii.go

@@ -0,0 +1,42 @@
+package characters
+
+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]
+}

internal/characters/utf8.go

@@ -1,4 +1,4 @@
-package toml
+package characters
 
 import (
 	"unicode/utf8"
@@ -32,7 +32,7 @@ func (u utf8Err) Zero() bool {
 // 0x9 => tab, ok
 // 0xA - 0x1F => invalid
 // 0x7F => invalid
-func utf8TomlValidAlreadyEscaped(p []byte) (err utf8Err) {
+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 {
@@ -48,7 +48,7 @@ func utf8TomlValidAlreadyEscaped(p []byte) (err utf8Err) {
 		}
 
 		for i, b := range p[:8] {
-			if invalidAscii(b) {
+			if InvalidAscii(b) {
 				err.Index = offset + i
 				err.Size = 1
 				return
@@ -62,7 +62,7 @@ func utf8TomlValidAlreadyEscaped(p []byte) (err utf8Err) {
 	for i := 0; i < n; {
 		pi := p[i]
 		if pi < utf8.RuneSelf {
-			if invalidAscii(pi) {
+			if InvalidAscii(pi) {
 				err.Index = offset + i
 				err.Size = 1
 				return
@@ -106,11 +106,11 @@ func utf8TomlValidAlreadyEscaped(p []byte) (err utf8Err) {
 }
 
 // Return the size of the next rune if valid, 0 otherwise.
-func utf8ValidNext(p []byte) int {
+func Utf8ValidNext(p []byte) int {
 	c := p[0]
 
 	if c < utf8.RuneSelf {
-		if invalidAscii(c) {
+		if InvalidAscii(c) {
 			return 0
 		}
 		return 1
@@ -140,47 +140,6 @@ func utf8ValidNext(p []byte) int {
 	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 {

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 conversion 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/imported_tests/marshal_imported_test.go

@@ -67,6 +67,7 @@ func TestDocMarshal(t *testing.T) {
 	}
 
 	marshalTestToml := `title = 'TOML Marshal Testing'
+
 [basic_lists]
 floats = [12.3, 45.6, 78.9]
 bools = [true, false, true]
@@ -89,7 +90,6 @@ name = 'Second'
 [subdoc.first]
 name = 'First'
 
-
 [basic]
 uint = 5001
 bool = true
@@ -101,9 +101,9 @@ date = 1979-05-27T07:32:00Z
 
 [[subdoclist]]
 name = 'List.First'
+
 [[subdoclist]]
 name = 'List.Second'
-
 `
 
 	result, err := toml.Marshal(docData)
@@ -117,14 +117,15 @@ func TestBasicMarshalQuotedKey(t *testing.T) {
 
 	expected := `'Z.string-àéù' = 'Hello'
 'Yfloat-𝟘' = 3.5
+
 ['Xsubdoc-àéù']
 String2 = 'One'
 
 [['W.sublist-𝟘']]
 String2 = 'Two'
+
 [['W.sublist-𝟘']]
 String2 = 'Three'
-
 `
 
 	require.Equal(t, string(expected), string(result))
@@ -159,8 +160,8 @@ bool = false
 int = 0
 string = ''
 stringlist = []
-[map]
 
+[map]
 `
 
 	require.Equal(t, string(expected), string(result))

internal/imported_tests/unmarshal_imported_test.go

@@ -151,6 +151,7 @@ type quotedKeyMarshalTestStruct struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var quotedKeyMarshalTestData = quotedKeyMarshalTestStruct{
 	String:  "Hello",
@@ -160,6 +161,7 @@ var quotedKeyMarshalTestData = quotedKeyMarshalTestStruct{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var quotedKeyMarshalTestToml = []byte(`"Yfloat-𝟘" = 3.5
 "Z.string-àéù" = "Hello"
@@ -272,6 +274,7 @@ var docData = testDoc{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapTestDoc = testMapDoc{
 	Title: "TOML Marshal Testing",
@@ -559,10 +562,12 @@ func (c customMarshaler) MarshalTOML() ([]byte, error) {
 var customMarshalerData = customMarshaler{FirstName: "Sally", LastName: "Fields"}
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customMarshalerToml = []byte(`Sally Fields`)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var nestedCustomMarshalerData = customMarshalerParent{
 	Self:    customMarshaler{FirstName: "Maiku", LastName: "Suteda"},
@@ -570,6 +575,7 @@ var nestedCustomMarshalerData = customMarshalerParent{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var nestedCustomMarshalerToml = []byte(`friends = ["Sally Fields"]
 me = "Maiku Suteda"
@@ -611,6 +617,7 @@ func TestUnmarshalTextMarshaler(t *testing.T) {
 }
 
 // TODO: Remove nolint once type and methods are used by a test
+//
 //nolint:unused
 type precedentMarshaler struct {
 	FirstName string
@@ -629,6 +636,7 @@ func (m precedentMarshaler) MarshalTOML() ([]byte, error) {
 }
 
 // TODO: Remove nolint once type and method are used by a test
+//
 //nolint:unused
 type customPointerMarshaler struct {
 	FirstName string
@@ -641,6 +649,7 @@ func (m *customPointerMarshaler) MarshalTOML() ([]byte, error) {
 }
 
 // TODO: Remove nolint once type and method are used by a test
+//
 //nolint:unused
 type textPointerMarshaler struct {
 	FirstName string
@@ -653,6 +662,7 @@ func (m *textPointerMarshaler) MarshalText() ([]byte, error) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var commentTestToml = []byte(`
 # it's a comment on type
@@ -690,6 +700,7 @@ type mapsTestStruct struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapsTestData = mapsTestStruct{
 	Simple: map[string]string{
@@ -713,6 +724,7 @@ var mapsTestData = mapsTestStruct{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapsTestToml = []byte(`
 [Other]
@@ -735,6 +747,7 @@ var mapsTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode,unused
 type structArrayNoTag struct {
 	A struct {
@@ -744,6 +757,7 @@ type structArrayNoTag struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customTagTestToml = []byte(`
 [postgres]
@@ -758,6 +772,7 @@ var customTagTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customCommentTagTestToml = []byte(`
 # db connection
@@ -771,6 +786,7 @@ var customCommentTagTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customCommentedTagTestToml = []byte(`
 [postgres]
@@ -825,6 +841,7 @@ func TestUnmarshalTabInStringAndQuotedKey(t *testing.T) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customMultilineTagTestToml = []byte(`int_slice = [
   1,
@@ -834,6 +851,7 @@ var customMultilineTagTestToml = []byte(`int_slice = [
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDocBasicToml = []byte(`
 [document]
@@ -846,12 +864,14 @@ var testDocBasicToml = []byte(`
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode
 type testDocCustomTag struct {
 	Doc testDocBasicsCustomTag `file:"document"`
 }
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode
 type testDocBasicsCustomTag struct {
 	Bool       bool      `file:"bool_val"`
@@ -864,6 +884,7 @@ type testDocBasicsCustomTag struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,varcheck
 var testDocCustomTagData = testDocCustomTag{
 	Doc: testDocBasicsCustomTag{
@@ -927,6 +948,29 @@ 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)
@@ -943,6 +987,7 @@ func TestUnmarshalInvalidPointerKind(t *testing.T) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused
 type testDuration struct {
 	Nanosec   time.Duration  `toml:"nanosec"`
@@ -957,6 +1002,7 @@ type testDuration struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDurationToml = []byte(`
 nanosec = "1ns"
@@ -971,6 +1017,7 @@ a_string = "15s"
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDurationToml2 = []byte(`a_string = "15s"
 hour = "1h0m0s"
@@ -984,6 +1031,7 @@ sec = "1s"
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode,unused
 type testBadDuration struct {
 	Val time.Duration `toml:"val"`
@@ -1037,10 +1085,6 @@ func TestUnmarshalCheckConversionFloatInt(t *testing.T) {
 			desc:  "int",
 			input: `I = 1e300`,
 		},
-		{
-			desc:  "float",
-			input: `F = 9223372036854775806`,
-		},
 	}
 
 	for _, test := range testCases {
@@ -1925,7 +1969,7 @@ func decoder(doc string) *toml.Decoder {
 
 func strictDecoder(doc string) *toml.Decoder {
 	d := decoder(doc)
-	d.SetStrict(true)
+	d.DisallowUnknownFields()
 	return d
 }
 

internal/tracker/key.go

@@ -1,8 +1,6 @@
 package tracker
 
-import (
+import "github.com/pelletier/go-toml/v2/unstable"
-	"github.com/pelletier/go-toml/v2/internal/ast"
-)
 
 // KeyTracker is a tracker that keeps track of the current Key as the AST is
 // walked.
@@ -11,19 +9,19 @@ type KeyTracker struct {
 }
 
 // UpdateTable sets the state of the tracker with the AST table node.
-func (t *KeyTracker) UpdateTable(node *ast.Node) {
+func (t *KeyTracker) UpdateTable(node *unstable.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) {
+func (t *KeyTracker) UpdateArrayTable(node *unstable.Node) {
 	t.reset()
 	t.Push(node)
 }
 
 // Push the given key on the stack.
-func (t *KeyTracker) Push(node *ast.Node) {
+func (t *KeyTracker) Push(node *unstable.Node) {
 	it := node.Key()
 	for it.Next() {
 		t.k = append(t.k, string(it.Node().Data))
@@ -31,7 +29,7 @@ func (t *KeyTracker) Push(node *ast.Node) {
 }
 
 // Pop key from stack.
-func (t *KeyTracker) Pop(node *ast.Node) {
+func (t *KeyTracker) Pop(node *unstable.Node) {
 	it := node.Key()
 	for it.Next() {
 		t.k = t.k[:len(t.k)-1]

internal/tracker/seen.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"sync"
 
-	"github.com/pelletier/go-toml/v2/internal/ast"
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 type keyKind uint8
@@ -79,6 +79,7 @@ type entry struct {
 	name     []byte
 	kind     keyKind
 	explicit bool
+	kv       bool
 }
 
 // Find the index of the child of parentIdx with key k. Returns -1 if
@@ -111,7 +112,7 @@ func (s *SeenTracker) clear(idx int) {
 	s.entries[idx].child = -1
 }
 
-func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit bool) int {
+func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit bool, kv bool) int {
 	e := entry{
 		child: -1,
 		next:  s.entries[parentIdx].child,
@@ -119,6 +120,7 @@ func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit
 		name:     name,
 		kind:     kind,
 		explicit: explicit,
+		kv:       kv,
 	}
 	var idx int
 	if s.entries[0].next >= 0 {
@@ -137,7 +139,10 @@ func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit
 
 func (s *SeenTracker) setExplicitFlag(parentIdx int) {
 	for i := s.entries[parentIdx].child; i >= 0; i = s.entries[i].next {
-		s.entries[i].explicit = true
+		if s.entries[i].kv {
+			s.entries[i].explicit = true
+			s.entries[i].kv = false
+		}
 		s.setExplicitFlag(i)
 	}
 }
@@ -145,23 +150,23 @@ func (s *SeenTracker) setExplicitFlag(parentIdx int) {
 // 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 {
+func (s *SeenTracker) CheckExpression(node *unstable.Node) error {
 	if s.entries == nil {
 		s.reset()
 	}
 	switch node.Kind {
-	case ast.KeyValue:
+	case unstable.KeyValue:
 		return s.checkKeyValue(node)
-	case ast.Table:
+	case unstable.Table:
 		return s.checkTable(node)
-	case ast.ArrayTable:
+	case unstable.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 {
+func (s *SeenTracker) checkTable(node *unstable.Node) error {
 	if s.currentIdx >= 0 {
 		s.setExplicitFlag(s.currentIdx)
 	}
@@ -183,7 +188,7 @@ func (s *SeenTracker) checkTable(node *ast.Node) error {
 		idx := s.find(parentIdx, k)
 
 		if idx < 0 {
-			idx = s.create(parentIdx, k, tableKind, false)
+			idx = s.create(parentIdx, k, tableKind, false, false)
 		} else {
 			entry := s.entries[idx]
 			if entry.kind == valueKind {
@@ -206,7 +211,7 @@ func (s *SeenTracker) checkTable(node *ast.Node) error {
 		}
 		s.entries[idx].explicit = true
 	} else {
-		idx = s.create(parentIdx, k, tableKind, true)
+		idx = s.create(parentIdx, k, tableKind, true, false)
 	}
 
 	s.currentIdx = idx
@@ -214,7 +219,7 @@ func (s *SeenTracker) checkTable(node *ast.Node) error {
 	return nil
 }
 
-func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
+func (s *SeenTracker) checkArrayTable(node *unstable.Node) error {
 	if s.currentIdx >= 0 {
 		s.setExplicitFlag(s.currentIdx)
 	}
@@ -233,7 +238,7 @@ func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
 		idx := s.find(parentIdx, k)
 
 		if idx < 0 {
-			idx = s.create(parentIdx, k, tableKind, false)
+			idx = s.create(parentIdx, k, tableKind, false, false)
 		} else {
 			entry := s.entries[idx]
 			if entry.kind == valueKind {
@@ -254,7 +259,7 @@ func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
 		}
 		s.clear(idx)
 	} else {
-		idx = s.create(parentIdx, k, arrayTableKind, true)
+		idx = s.create(parentIdx, k, arrayTableKind, true, false)
 	}
 
 	s.currentIdx = idx
@@ -262,7 +267,7 @@ func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
 	return nil
 }
 
-func (s *SeenTracker) checkKeyValue(node *ast.Node) error {
+func (s *SeenTracker) checkKeyValue(node *unstable.Node) error {
 	parentIdx := s.currentIdx
 	it := node.Key()
 
@@ -272,7 +277,7 @@ func (s *SeenTracker) checkKeyValue(node *ast.Node) error {
 		idx := s.find(parentIdx, k)
 
 		if idx < 0 {
-			idx = s.create(parentIdx, k, tableKind, false)
+			idx = s.create(parentIdx, k, tableKind, false, true)
 		} else {
 			entry := s.entries[idx]
 			if it.IsLast() {
@@ -292,26 +297,26 @@ func (s *SeenTracker) checkKeyValue(node *ast.Node) error {
 	value := node.Value()
 
 	switch value.Kind {
-	case ast.InlineTable:
+	case unstable.InlineTable:
 		return s.checkInlineTable(value)
-	case ast.Array:
+	case unstable.Array:
 		return s.checkArray(value)
 	}
 
 	return nil
 }
 
-func (s *SeenTracker) checkArray(node *ast.Node) error {
+func (s *SeenTracker) checkArray(node *unstable.Node) error {
 	it := node.Children()
 	for it.Next() {
 		n := it.Node()
 		switch n.Kind {
-		case ast.InlineTable:
+		case unstable.InlineTable:
 			err := s.checkInlineTable(n)
 			if err != nil {
 				return err
 			}
-		case ast.Array:
+		case unstable.Array:
 			err := s.checkArray(n)
 			if err != nil {
 				return err
@@ -321,7 +326,7 @@ func (s *SeenTracker) checkArray(node *ast.Node) error {
 	return nil
 }
 
-func (s *SeenTracker) checkInlineTable(node *ast.Node) error {
+func (s *SeenTracker) checkInlineTable(node *unstable.Node) error {
 	if pool.New == nil {
 		pool.New = func() interface{} {
 			return &SeenTracker{}

internal/tracker/seen_test.go

@@ -0,0 +1,16 @@
+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.
+	maxExpectedEntrySize := 48
+	require.LessOrEqual(t, int(unsafe.Sizeof(entry{})), maxExpectedEntrySize)
+}

localtime.go

@@ -4,6 +4,8 @@ import (
 	"fmt"
 	"strings"
 	"time"
+
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 // LocalDate represents a calendar day in no specific timezone.
@@ -75,7 +77,7 @@ func (d LocalTime) MarshalText() ([]byte, error) {
 func (d *LocalTime) UnmarshalText(b []byte) error {
 	res, left, err := parseLocalTime(b)
 	if err == nil && len(left) != 0 {
-		err = newDecodeError(left, "extra characters")
+		err = unstable.NewParserError(left, "extra characters")
 	}
 	if err != nil {
 		return err
@@ -109,7 +111,7 @@ func (d LocalDateTime) MarshalText() ([]byte, error) {
 func (d *LocalDateTime) UnmarshalText(data []byte) error {
 	res, left, err := parseLocalDateTime(data)
 	if err == nil && len(left) != 0 {
-		err = newDecodeError(left, "extra characters")
+		err = unstable.NewParserError(left, "extra characters")
 	}
 	if err != nil {
 		return err

marshaler.go

@@ -12,6 +12,8 @@ import (
 	"strings"
 	"time"
 	"unicode"
+
+	"github.com/pelletier/go-toml/v2/internal/characters"
 )
 
 // Marshal serializes a Go value as a TOML document.
@@ -54,7 +56,7 @@ func NewEncoder(w io.Writer) *Encoder {
 // This behavior can be controlled on an individual struct field basis with the
 // inline tag:
 //
-//   MyField `inline:"true"`
+//	MyField `toml:",inline"`
 func (enc *Encoder) SetTablesInline(inline bool) *Encoder {
 	enc.tablesInline = inline
 	return enc
@@ -65,7 +67,7 @@ func (enc *Encoder) SetTablesInline(inline bool) *Encoder {
 //
 // This behavior can be controlled on an individual struct field basis with the multiline tag:
 //
-//   MyField `multiline:"true"`
+//	MyField `multiline:"true"`
 func (enc *Encoder) SetArraysMultiline(multiline bool) *Encoder {
 	enc.arraysMultiline = multiline
 	return enc
@@ -89,7 +91,7 @@ func (enc *Encoder) SetIndentTables(indent bool) *Encoder {
 //
 // If v cannot be represented to TOML it returns an error.
 //
-// Encoding rules
+// # Encoding rules
 //
 // A top level slice containing only maps or structs is encoded as [[table
 // array]].
@@ -107,10 +109,30 @@ func (enc *Encoder) SetIndentTables(indent bool) *Encoder {
 // a newline character or a single quote. In that case they are emitted as
 // quoted strings.
 //
+// Unsigned integers larger than math.MaxInt64 cannot be encoded. Doing so
+// results in an error. This rule exists because the TOML specification only
+// requires parsers to support at least the 64 bits integer range. Allowing
+// larger numbers would create non-standard TOML documents, which may not be
+// readable (at best) by other implementations. To encode such numbers, a
+// solution is a custom type that implements encoding.TextMarshaler.
+//
 // When encoding structs, fields are encoded in order of definition, with their
 // exact name.
 //
-// Struct tags
+// Tables and array tables are separated by empty lines. However, consecutive
+// subtables definitions are not. For example:
+//
+//	[top1]
+//
+//	[top2]
+//	[top2.child1]
+//
+//	[[array]]
+//
+//	[[array]]
+//	[array.child2]
+//
+// # Struct tags
 //
 // The encoding of each public struct field can be customized by the format
 // string in the "toml" key of the struct field's tag. This follows
@@ -126,9 +148,13 @@ func (enc *Encoder) SetIndentTables(indent bool) *Encoder {
 //
 // The "omitempty" option prevents empty values or groups from being emitted.
 //
+// The "commented" option prefixes the value and all its children with a comment
+// symbol.
+//
 // In addition to the "toml" tag struct tag, a "comment" tag can be used to emit
 // a TOML comment before the value being annotated. Comments are ignored inside
-// inline tables.
+// inline tables. For array tables, the comment is only present before the first
+// element of the array.
 func (enc *Encoder) Encode(v interface{}) error {
 	var (
 		b   []byte
@@ -157,6 +183,7 @@ func (enc *Encoder) Encode(v interface{}) error {
 type valueOptions struct {
 	multiline bool
 	omitempty bool
+	commented bool
 	comment   string
 }
 
@@ -182,6 +209,9 @@ type encoderCtx struct {
 	// Indentation level
 	indent int
 
+	// Prefix the current value with a comment.
+	commented bool
+
 	// Options coming from struct tags
 	options valueOptions
 }
@@ -208,11 +238,20 @@ func (ctx *encoderCtx) isRoot() bool {
 }
 
 func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
-	if !v.IsZero() {
+	i := v.Interface()
-		i, ok := v.Interface().(time.Time)
+
-		if ok {
+	switch x := i.(type) {
-			return i.AppendFormat(b, time.RFC3339), nil
+	case time.Time:
+		if x.Nanosecond() > 0 {
+			return x.AppendFormat(b, time.RFC3339Nano), nil
 		}
+		return x.AppendFormat(b, time.RFC3339), nil
+	case LocalTime:
+		return append(b, x.String()...), nil
+	case LocalDate:
+		return append(b, x.String()...), nil
+	case LocalDateTime:
+		return append(b, x.String()...), nil
 	}
 
 	hasTextMarshaler := v.Type().Implements(textMarshalerType)
@@ -241,7 +280,7 @@ func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, e
 		return enc.encodeMap(b, ctx, v)
 	case reflect.Struct:
 		return enc.encodeStruct(b, ctx, v)
-	case reflect.Slice:
+	case reflect.Slice, reflect.Array:
 		return enc.encodeSlice(b, ctx, v)
 	case reflect.Interface:
 		if v.IsNil() {
@@ -260,16 +299,31 @@ func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, e
 	case reflect.String:
 		b = enc.encodeString(b, v.String(), ctx.options)
 	case reflect.Float32:
-		if math.Trunc(v.Float()) == v.Float() {
+		f := v.Float()
-			b = strconv.AppendFloat(b, v.Float(), 'f', 1, 32)
+
+		if math.IsNaN(f) {
+			b = append(b, "nan"...)
+		} else if f > math.MaxFloat32 {
+			b = append(b, "inf"...)
+		} else if f < -math.MaxFloat32 {
+			b = append(b, "-inf"...)
+		} else if math.Trunc(f) == f {
+			b = strconv.AppendFloat(b, f, 'f', 1, 32)
 		} else {
-			b = strconv.AppendFloat(b, v.Float(), 'f', -1, 32)
+			b = strconv.AppendFloat(b, f, 'f', -1, 32)
 		}
 	case reflect.Float64:
-		if math.Trunc(v.Float()) == v.Float() {
+		f := v.Float()
-			b = strconv.AppendFloat(b, v.Float(), 'f', 1, 64)
+		if math.IsNaN(f) {
+			b = append(b, "nan"...)
+		} else if f > math.MaxFloat64 {
+			b = append(b, "inf"...)
+		} else if f < -math.MaxFloat64 {
+			b = append(b, "-inf"...)
+		} else if math.Trunc(f) == f {
+			b = strconv.AppendFloat(b, f, 'f', 1, 64)
 		} else {
-			b = strconv.AppendFloat(b, v.Float(), 'f', -1, 64)
+			b = strconv.AppendFloat(b, f, 'f', -1, 64)
 		}
 	case reflect.Bool:
 		if v.Bool() {
@@ -278,7 +332,11 @@ func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, e
 			b = append(b, "false"...)
 		}
 	case reflect.Uint64, reflect.Uint32, reflect.Uint16, reflect.Uint8, reflect.Uint:
-		b = strconv.AppendUint(b, v.Uint(), 10)
+		x := v.Uint()
+		if x > uint64(math.MaxInt64) {
+			return nil, fmt.Errorf("toml: not encoding uint (%d) greater than max int64 (%d)", x, int64(math.MaxInt64))
+		}
+		b = strconv.AppendUint(b, x, 10)
 	case reflect.Int64, reflect.Int32, reflect.Int16, reflect.Int8, reflect.Int:
 		b = strconv.AppendInt(b, v.Int(), 10)
 	default:
@@ -297,28 +355,20 @@ func isNil(v reflect.Value) bool {
 	}
 }
 
+func shouldOmitEmpty(options valueOptions, v reflect.Value) bool {
+	return options.omitempty && isEmptyValue(v)
+}
+
 func (enc *Encoder) encodeKv(b []byte, ctx encoderCtx, options valueOptions, v reflect.Value) ([]byte, error) {
 	var err error
 
-	if !ctx.hasKey {
-		panic("caller of encodeKv should have set the key in the context")
-	}
-
-	if (ctx.options.omitempty || options.omitempty) && isEmptyValue(v) {
-		return b, nil
-	}
-
 	if !ctx.inline {
 		b = enc.encodeComment(ctx.indent, options.comment, b)
+		b = enc.commented(ctx.commented, b)
+		b = enc.indent(ctx.indent, b)
 	}
 
-	b = enc.indent(ctx.indent, b)
+	b = enc.encodeKey(b, ctx.key)
-
-	b, err = enc.encodeKey(b, ctx.key)
-	if err != nil {
-		return nil, err
-	}
-
 	b = append(b, " = "...)
 
 	// create a copy of the context because the value of a KV shouldn't
@@ -336,8 +386,17 @@ func (enc *Encoder) encodeKv(b []byte, ctx encoderCtx, options valueOptions, v r
 	return b, nil
 }
 
+func (enc *Encoder) commented(commented bool, b []byte) []byte {
+	if commented {
+		return append(b, "# "...)
+	}
+	return b
+}
+
 func isEmptyValue(v reflect.Value) bool {
 	switch v.Kind() {
+	case reflect.Struct:
+		return isEmptyStruct(v)
 	case reflect.Array, reflect.Map, reflect.Slice, reflect.String:
 		return v.Len() == 0
 	case reflect.Bool:
@@ -354,6 +413,34 @@ func isEmptyValue(v reflect.Value) bool {
 	return false
 }
 
+func isEmptyStruct(v reflect.Value) bool {
+	// TODO: merge with walkStruct and cache.
+	typ := v.Type()
+	for i := 0; i < typ.NumField(); i++ {
+		fieldType := typ.Field(i)
+
+		// only consider exported fields
+		if fieldType.PkgPath != "" {
+			continue
+		}
+
+		tag := fieldType.Tag.Get("toml")
+
+		// special field name to skip field
+		if tag == "-" {
+			continue
+		}
+
+		f := v.Field(i)
+
+		if !isEmptyValue(f) {
+			return false
+		}
+	}
+
+	return true
+}
+
 const literalQuote = '\''
 
 func (enc *Encoder) encodeString(b []byte, v string, options valueOptions) []byte {
@@ -365,7 +452,13 @@ func (enc *Encoder) encodeString(b []byte, v string, options valueOptions) []byt
 }
 
 func needsQuoting(v string) bool {
-	return strings.ContainsAny(v, "'\b\f\n\r\t")
+	// TODO: vectorize
+	for _, b := range []byte(v) {
+		if b == '\'' || b == '\r' || b == '\n' || characters.InvalidAscii(b) {
+			return true
+		}
+	}
+	return false
 }
 
 // caller should have checked that the string does not contain new lines or ' .
@@ -377,7 +470,6 @@ func (enc *Encoder) encodeLiteralString(b []byte, v string) []byte {
 	return b
 }
 
-//nolint:cyclop
 func (enc *Encoder) encodeQuotedString(multiline bool, b []byte, v string) []byte {
 	stringQuote := `"`
 
@@ -437,7 +529,7 @@ func (enc *Encoder) encodeQuotedString(multiline bool, b []byte, v string) []byt
 	return b
 }
 
-// called should have checked that the string is in A-Z / a-z / 0-9 / - / _ .
+// caller should have checked that the string is in A-Z / a-z / 0-9 / - / _ .
 func (enc *Encoder) encodeUnquotedKey(b []byte, v string) []byte {
 	return append(b, v...)
 }
@@ -449,24 +541,17 @@ func (enc *Encoder) encodeTableHeader(ctx encoderCtx, b []byte) ([]byte, error)
 
 	b = enc.encodeComment(ctx.indent, ctx.options.comment, b)
 
+	b = enc.commented(ctx.commented, b)
+
 	b = enc.indent(ctx.indent, b)
 
 	b = append(b, '[')
 
-	var err error
+	b = enc.encodeKey(b, ctx.parentKey[0])
-
-	b, err = enc.encodeKey(b, ctx.parentKey[0])
-	if err != nil {
-		return nil, err
-	}
 
 	for _, k := range ctx.parentKey[1:] {
 		b = append(b, '.')
-
+		b = enc.encodeKey(b, k)
-		b, err = enc.encodeKey(b, k)
-		if err != nil {
-			return nil, err
-		}
 	}
 
 	b = append(b, "]\n"...)
@@ -475,19 +560,19 @@ func (enc *Encoder) encodeTableHeader(ctx encoderCtx, b []byte) ([]byte, error)
 }
 
 //nolint:cyclop
-func (enc *Encoder) encodeKey(b []byte, k string) ([]byte, error) {
+func (enc *Encoder) encodeKey(b []byte, k string) []byte {
 	needsQuotation := false
 	cannotUseLiteral := false
 
+	if len(k) == 0 {
+		return append(b, "''"...)
+	}
+
 	for _, c := range k {
 		if (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') || c == '-' || c == '_' {
 			continue
 		}
 
-		if c == '\n' {
-			return nil, fmt.Errorf("toml: new line characters in keys are not supported")
-		}
-
 		if c == literalQuote {
 			cannotUseLiteral = true
 		}
@@ -495,21 +580,37 @@ func (enc *Encoder) encodeKey(b []byte, k string) ([]byte, error) {
 		needsQuotation = true
 	}
 
+	if needsQuotation && needsQuoting(k) {
+		cannotUseLiteral = true
+	}
+
 	switch {
 	case cannotUseLiteral:
-		return enc.encodeQuotedString(false, b, k), nil
+		return enc.encodeQuotedString(false, b, k)
 	case needsQuotation:
-		return enc.encodeLiteralString(b, k), nil
+		return enc.encodeLiteralString(b, k)
 	default:
-		return enc.encodeUnquotedKey(b, k), nil
+		return enc.encodeUnquotedKey(b, k)
 	}
 }
 
-func (enc *Encoder) encodeMap(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+func (enc *Encoder) keyToString(k reflect.Value) (string, error) {
-	if v.Type().Key().Kind() != reflect.String {
+	keyType := k.Type()
-		return nil, fmt.Errorf("toml: type %s is not supported as a map key", v.Type().Key().Kind())
+	switch {
-	}
+	case keyType.Kind() == reflect.String:
+		return k.String(), nil
 
+	case keyType.Implements(textMarshalerType):
+		keyB, err := k.Interface().(encoding.TextMarshaler).MarshalText()
+		if err != nil {
+			return "", fmt.Errorf("toml: error marshalling key %v from text: %w", k, err)
+		}
+		return string(keyB), nil
+	}
+	return "", fmt.Errorf("toml: type %s is not supported as a map key", keyType.Kind())
+}
+
+func (enc *Encoder) encodeMap(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
 	var (
 		t                 table
 		emptyValueOptions valueOptions
@@ -517,13 +618,17 @@ func (enc *Encoder) encodeMap(b []byte, ctx encoderCtx, v reflect.Value) ([]byte
 
 	iter := v.MapRange()
 	for iter.Next() {
-		k := iter.Key().String()
 		v := iter.Value()
 
 		if isNil(v) {
 			continue
 		}
 
+		k, err := enc.keyToString(iter.Key())
+		if err != nil {
+			return nil, err
+		}
+
 		if willConvertToTableOrArrayTable(ctx, v) {
 			t.pushTable(k, v, emptyValueOptions)
 		} else {
@@ -555,16 +660,25 @@ type table struct {
 }
 
 func (t *table) pushKV(k string, v reflect.Value, options valueOptions) {
+	for _, e := range t.kvs {
+		if e.Key == k {
+			return
+		}
+	}
+
 	t.kvs = append(t.kvs, entry{Key: k, Value: v, Options: options})
 }
 
 func (t *table) pushTable(k string, v reflect.Value, options valueOptions) {
+	for _, e := range t.tables {
+		if e.Key == k {
+			return
+		}
+	}
 	t.tables = append(t.tables, entry{Key: k, Value: v, Options: options})
 }
 
-func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+func walkStruct(ctx encoderCtx, t *table, v reflect.Value) {
-	var t table
-
 	// TODO: cache this
 	typ := v.Type()
 	for i := 0; i < typ.NumField(); i++ {
@@ -575,8 +689,6 @@ func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]b
 			continue
 		}
 
-		k := fieldType.Name
-
 		tag := fieldType.Tag.Get("toml")
 
 		// special field name to skip field
@@ -584,13 +696,24 @@ func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]b
 			continue
 		}
 
-		name, opts := parseTag(tag)
+		k, opts := parseTag(tag)
-		if isValidName(name) {
+		if !isValidName(k) {
-			k = name
+			k = ""
 		}
 
 		f := v.Field(i)
 
+		if k == "" {
+			if fieldType.Anonymous {
+				if fieldType.Type.Kind() == reflect.Struct {
+					walkStruct(ctx, t, f)
+				}
+				continue
+			} else {
+				k = fieldType.Name
+			}
+		}
+
 		if isNil(f) {
 			continue
 		}
@@ -598,6 +721,7 @@ func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]b
 		options := valueOptions{
 			multiline: opts.multiline,
 			omitempty: opts.omitempty,
+			commented: opts.commented,
 			comment:   fieldType.Tag.Get("comment"),
 		}
 
@@ -607,15 +731,30 @@ func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]b
 			t.pushTable(k, f, options)
 		}
 	}
+}
+
+func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	var t table
+
+	walkStruct(ctx, &t, v)
 
 	return enc.encodeTable(b, ctx, t)
 }
 
 func (enc *Encoder) encodeComment(indent int, comment string, b []byte) []byte {
-	if comment != "" {
+	for len(comment) > 0 {
+		var line string
+		idx := strings.IndexByte(comment, '\n')
+		if idx >= 0 {
+			line = comment[:idx]
+			comment = comment[idx+1:]
+		} else {
+			line = comment
+			comment = ""
+		}
 		b = enc.indent(indent, b)
 		b = append(b, "# "...)
-		b = append(b, comment...)
+		b = append(b, line...)
 		b = append(b, '\n')
 	}
 	return b
@@ -642,6 +781,7 @@ type tagOptions struct {
 	multiline bool
 	inline    bool
 	omitempty bool
+	commented bool
 }
 
 func parseTag(tag string) (string, tagOptions) {
@@ -669,6 +809,8 @@ func parseTag(tag string) (string, tagOptions) {
 			opts.inline = true
 		case "omitempty":
 			opts.omitempty = true
+		case "commented":
+			opts.commented = true
 		}
 	}
 
@@ -696,10 +838,18 @@ func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, erro
 	}
 	ctx.skipTableHeader = false
 
+	hasNonEmptyKV := false
 	for _, kv := range t.kvs {
-		ctx.setKey(kv.Key)
+		if shouldOmitEmpty(kv.Options, kv.Value) {
+			continue
+		}
+		hasNonEmptyKV = true
 
-		b, err = enc.encodeKv(b, ctx, kv.Options, kv.Value)
+		ctx.setKey(kv.Key)
+		ctx2 := ctx
+		ctx2.commented = kv.Options.commented || ctx2.commented
+
+		b, err = enc.encodeKv(b, ctx2, kv.Options, kv.Value)
 		if err != nil {
 			return nil, err
 		}
@@ -707,17 +857,30 @@ func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, erro
 		b = append(b, '\n')
 	}
 
+	first := true
 	for _, table := range t.tables {
+		if shouldOmitEmpty(table.Options, table.Value) {
+			continue
+		}
+		if first {
+			first = false
+			if hasNonEmptyKV {
+				b = append(b, '\n')
+			}
+		} else {
+			b = append(b, "\n"...)
+		}
+
 		ctx.setKey(table.Key)
 
 		ctx.options = table.Options
+		ctx2 := ctx
+		ctx2.commented = ctx2.commented || ctx.options.commented
 
-		b, err = enc.encode(b, ctx, table.Value)
+		b, err = enc.encode(b, ctx2, table.Value)
 		if err != nil {
 			return nil, err
 		}
-
-		b = append(b, '\n')
 	}
 
 	return b, nil
@@ -730,6 +893,10 @@ func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte
 
 	first := true
 	for _, kv := range t.kvs {
+		if shouldOmitEmpty(kv.Options, kv.Value) {
+			continue
+		}
+
 		if first {
 			first = false
 		} else {
@@ -745,7 +912,7 @@ func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte
 	}
 
 	if len(t.tables) > 0 {
-		panic("inline table cannot contain nested tables, online key-values")
+		panic("inline table cannot contain nested tables, only key-values")
 	}
 
 	b = append(b, "}"...)
@@ -779,13 +946,16 @@ func willConvertToTable(ctx encoderCtx, v reflect.Value) bool {
 }
 
 func willConvertToTableOrArrayTable(ctx encoderCtx, v reflect.Value) bool {
+	if ctx.insideKv {
+		return false
+	}
 	t := v.Type()
 
 	if t.Kind() == reflect.Interface {
 		return willConvertToTableOrArrayTable(ctx, v.Elem())
 	}
 
-	if t.Kind() == reflect.Slice {
+	if t.Kind() == reflect.Slice || t.Kind() == reflect.Array {
 		if v.Len() == 0 {
 			// An empty slice should be a kv = [].
 			return false
@@ -824,8 +994,10 @@ func (enc *Encoder) encodeSlice(b []byte, ctx encoderCtx, v reflect.Value) ([]by
 func (enc *Encoder) encodeSliceAsArrayTable(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
 	ctx.shiftKey()
 
-	var err error
 	scratch := make([]byte, 0, 64)
+
+	scratch = enc.commented(ctx.commented, scratch)
+
 	scratch = append(scratch, "[["...)
 
 	for i, k := range ctx.parentKey {
@@ -833,18 +1005,26 @@ func (enc *Encoder) encodeSliceAsArrayTable(b []byte, ctx encoderCtx, v reflect.
 			scratch = append(scratch, '.')
 		}
 
-		scratch, err = enc.encodeKey(scratch, k)
+		scratch = enc.encodeKey(scratch, k)
-		if err != nil {
-			return nil, err
-		}
 	}
 
 	scratch = append(scratch, "]]\n"...)
 	ctx.skipTableHeader = true
 
+	b = enc.encodeComment(ctx.indent, ctx.options.comment, b)
+
+	if enc.indentTables {
+		ctx.indent++
+	}
+
 	for i := 0; i < v.Len(); i++ {
+		if i != 0 {
+			b = append(b, "\n"...)
+		}
+
 		b = append(b, scratch...)
 
+		var err error
 		b, err = enc.encode(b, ctx, v.Index(i))
 		if err != nil {
 			return nil, err

ossfuzz/fuzz.go

@@ -0,0 +1,45 @@
+//go:build go1.18 || go1.19 || go1.20 || go1.21
+// +build go1.18 go1.19 go1.20 go1.21
+
+package ossfuzz
+
+import (
+	"fmt"
+	"reflect"
+	"strings"
+
+	"github.com/pelletier/go-toml/v2"
+)
+
+func FuzzToml(data []byte) int {
+	if len(data) >= 2048 {
+		return 0
+	}
+
+	if strings.Contains(string(data), "nan") {
+		return 0
+	}
+
+	var v interface{}
+	err := toml.Unmarshal(data, &v)
+	if err != nil {
+		return 0
+	}
+
+	encoded, err := toml.Marshal(v)
+	if err != nil {
+		panic(fmt.Sprintf("failed to marshal unmarshaled document: %s", err))
+	}
+
+	var v2 interface{}
+	err = toml.Unmarshal(encoded, &v2)
+	if err != nil {
+		panic(fmt.Sprintf("failed round trip: %s", err))
+	}
+
+	if !reflect.DeepEqual(v, v2) {
+		panic(fmt.Sprintf("not equal: %#+v %#+v", v, v2))
+	}
+
+	return 1
+}

parser_test.go

@@ -1,450 +0,0 @@
-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
-		input string
-		kind  ast.Kind
-		err   bool
-	}{
-		{
-			desc:  "integer just digits",
-			input: `1234`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer zero",
-			input: `0`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer sign",
-			input: `+99`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer hex uppercase",
-			input: `0xDEADBEEF`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer hex lowercase",
-			input: `0xdead_beef`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer octal",
-			input: `0o01234567`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "integer binary",
-			input: `0b11010110`,
-			kind:  ast.Integer,
-		},
-		{
-			desc:  "float zero",
-			input: `0.0`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float positive zero",
-			input: `+0.0`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float negative zero",
-			input: `-0.0`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float pi",
-			input: `3.1415`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float negative",
-			input: `-0.01`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float signed exponent",
-			input: `5e+22`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float exponent lowercase",
-			input: `1e06`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float exponent uppercase",
-			input: `-2E-2`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float fractional with exponent",
-			input: `6.626e-34`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "float underscores",
-			input: `224_617.445_991_228`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "inf",
-			input: `inf`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "inf negative",
-			input: `-inf`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "inf positive",
-			input: `+inf`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "nan",
-			input: `nan`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "nan negative",
-			input: `-nan`,
-			kind:  ast.Float,
-		},
-		{
-			desc:  "nan positive",
-			input: `+nan`,
-			kind:  ast.Float,
-		},
-	}
-
-	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())
-			}
-		})
-	}
-}
-
-type (
-	astNode struct {
-		Kind     ast.Kind
-		Data     []byte
-		Children []astNode
-	}
-)
-
-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)
-
-	compareIterator(t, e.Children, n.Children())
-}
-
-func compareIterator(t *testing.T, expected []astNode, actual ast.Iterator) {
-	t.Helper()
-	idx := 0
-
-	for actual.Next() {
-		n := actual.Node()
-
-		if idx >= len(expected) {
-			t.Fatal("extra child in actual tree")
-		}
-		e := expected[idx]
-
-		compareNode(t, e, n)
-
-		idx++
-	}
-
-	if idx < len(expected) {
-		t.Fatal("missing children in actual", "idx =", idx, "expected =", len(expected))
-	}
-}
-
-//nolint:funlen
-func TestParser_AST(t *testing.T) {
-	examples := []struct {
-		desc  string
-		input string
-		ast   astNode
-		err   bool
-	}{
-		{
-			desc:  "simple string assignment",
-			input: `A = "hello"`,
-			ast: astNode{
-				Kind: ast.KeyValue,
-				Children: []astNode{
-					{
-						Kind: ast.String,
-						Data: []byte(`hello`),
-					},
-					{
-						Kind: ast.Key,
-						Data: []byte(`A`),
-					},
-				},
-			},
-		},
-		{
-			desc:  "simple bool assignment",
-			input: `A = true`,
-			ast: astNode{
-				Kind: ast.KeyValue,
-				Children: []astNode{
-					{
-						Kind: ast.Bool,
-						Data: []byte(`true`),
-					},
-					{
-						Kind: ast.Key,
-						Data: []byte(`A`),
-					},
-				},
-			},
-		},
-		{
-			desc:  "array of strings",
-			input: `A = ["hello", ["world", "again"]]`,
-			ast: astNode{
-				Kind: ast.KeyValue,
-				Children: []astNode{
-					{
-						Kind: ast.Array,
-						Children: []astNode{
-							{
-								Kind: ast.String,
-								Data: []byte(`hello`),
-							},
-							{
-								Kind: ast.Array,
-								Children: []astNode{
-									{
-										Kind: ast.String,
-										Data: []byte(`world`),
-									},
-									{
-										Kind: ast.String,
-										Data: []byte(`again`),
-									},
-								},
-							},
-						},
-					},
-					{
-						Kind: ast.Key,
-						Data: []byte(`A`),
-					},
-				},
-			},
-		},
-		{
-			desc:  "array of arrays of strings",
-			input: `A = ["hello", "world"]`,
-			ast: astNode{
-				Kind: ast.KeyValue,
-				Children: []astNode{
-					{
-						Kind: ast.Array,
-						Children: []astNode{
-							{
-								Kind: ast.String,
-								Data: []byte(`hello`),
-							},
-							{
-								Kind: ast.String,
-								Data: []byte(`world`),
-							},
-						},
-					},
-					{
-						Kind: ast.Key,
-						Data: []byte(`A`),
-					},
-				},
-			},
-		},
-		{
-			desc:  "inline table",
-			input: `name = { first = "Tom", last = "Preston-Werner" }`,
-			ast: astNode{
-				Kind: ast.KeyValue,
-				Children: []astNode{
-					{
-						Kind: ast.InlineTable,
-						Children: []astNode{
-							{
-								Kind: ast.KeyValue,
-								Children: []astNode{
-									{Kind: ast.String, Data: []byte(`Tom`)},
-									{Kind: ast.Key, Data: []byte(`first`)},
-								},
-							},
-							{
-								Kind: ast.KeyValue,
-								Children: []astNode{
-									{Kind: ast.String, Data: []byte(`Preston-Werner`)},
-									{Kind: ast.Key, Data: []byte(`last`)},
-								},
-							},
-						},
-					},
-					{
-						Kind: ast.Key,
-						Data: []byte(`name`),
-					},
-				},
-			},
-		},
-	}
-
-	for _, e := range examples {
-		e := e
-		t.Run(e.desc, func(t *testing.T) {
-			p := parser{}
-			p.Reset([]byte(e.input))
-			p.NextExpression()
-			err := p.Error()
-			if e.err {
-				require.Error(t, err)
-			} else {
-				require.NoError(t, err)
-				compareNode(t, e.ast, p.Expression())
-			}
-		})
-	}
-}
-
-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())
-			}
-		})
-	}
-}

strict.go

@@ -1,9 +1,9 @@
 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"
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 type strict struct {
@@ -12,10 +12,10 @@ type strict struct {
 	// Tracks the current key being processed.
 	key tracker.KeyTracker
 
-	missing []decodeError
+	missing []unstable.ParserError
 }
 
-func (s *strict) EnterTable(node *ast.Node) {
+func (s *strict) EnterTable(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -23,7 +23,7 @@ func (s *strict) EnterTable(node *ast.Node) {
 	s.key.UpdateTable(node)
 }
 
-func (s *strict) EnterArrayTable(node *ast.Node) {
+func (s *strict) EnterArrayTable(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -31,7 +31,7 @@ func (s *strict) EnterArrayTable(node *ast.Node) {
 	s.key.UpdateArrayTable(node)
 }
 
-func (s *strict) EnterKeyValue(node *ast.Node) {
+func (s *strict) EnterKeyValue(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -39,7 +39,7 @@ func (s *strict) EnterKeyValue(node *ast.Node) {
 	s.key.Push(node)
 }
 
-func (s *strict) ExitKeyValue(node *ast.Node) {
+func (s *strict) ExitKeyValue(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -47,27 +47,27 @@ func (s *strict) ExitKeyValue(node *ast.Node) {
 	s.key.Pop(node)
 }
 
-func (s *strict) MissingTable(node *ast.Node) {
+func (s *strict) MissingTable(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
 
-	s.missing = append(s.missing, decodeError{
+	s.missing = append(s.missing, unstable.ParserError{
-		highlight: keyLocation(node),
+		Highlight: keyLocation(node),
-		message:   "missing table",
+		Message:   "missing table",
-		key:       s.key.Key(),
+		Key:       s.key.Key(),
 	})
 }
 
-func (s *strict) MissingField(node *ast.Node) {
+func (s *strict) MissingField(node *unstable.Node) {
 	if !s.Enabled {
 		return
 	}
 
-	s.missing = append(s.missing, decodeError{
+	s.missing = append(s.missing, unstable.ParserError{
-		highlight: keyLocation(node),
+		Highlight: keyLocation(node),
-		message:   "missing field",
+		Message:   "missing field",
-		key:       s.key.Key(),
+		Key:       s.key.Key(),
 	})
 }
 
@@ -88,7 +88,7 @@ func (s *strict) Error(doc []byte) error {
 	return err
 }
 
-func keyLocation(node *ast.Node) []byte {
+func keyLocation(node *unstable.Node) []byte {
 	k := node.Key()
 
 	hasOne := k.Next()

testdata/fuzz/FuzzUnmarshal/0c08b924aa26af23d16b77029b4f5567a96af93402e0472fef7b837b37580d7c

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=0000-01-01 00:00:00")

testdata/fuzz/FuzzUnmarshal/249a017d8ffa0d88d41c594ada5399f764833f64050180cb39f106d12666853f

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("\"\\n\"=\"\"")

testdata/fuzz/FuzzUnmarshal/391ce67b866d58464d067c5981e99951fd499896453a2fe2779a6bfe7df11bf5

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("''=0")

testdata/fuzz/FuzzUnmarshal/49ae83283c5cfbed874bcacbfb342de107b69c2a3d3c35c2d99f67345fe89946

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=0000-01-01")

testdata/fuzz/FuzzUnmarshal/69ea82e85bb4ad966ae86bb05eea3af69ed1a826c846f1bf2282f329b4d8fb36

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=\"\"\"\\U00000000\"\"\"")

testdata/fuzz/FuzzUnmarshal/746910fac6deb42e7e62dce60b333c81c39d844a122c9075894c023400e8fdbf

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=[[{}]]")

testdata/fuzz/FuzzUnmarshal/844d1171a3adc95e51627f0e9378bbaa193fe2d0284f4ba8ed19a5226a8cbc74

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("\"\\b\"=\"\"")

testdata/fuzz/FuzzUnmarshal/900c54b5071113ae2a5365b3156377ddd775b9d38add61629544903cb9c94b00

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=inf")

testdata/fuzz/FuzzUnmarshal/c87e8eb4e8fa40fd3c89d8ed16a049ce81592ded710965ee29a4e506eb5427b3

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=0000-01-01 00:00:00+00:00")

testdata/fuzz/FuzzUnmarshal/d0e5d6fc3b865cfae5480c01f301b87c932722ae0c1d692deea413349ea028be

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=[{}]")

testdata/fuzz/FuzzUnmarshal/d2cecea5b8ee5f148419671cef0644178cdba1ee91d3f295be38da576e89ef40

@@ -0,0 +1,2 @@
+go test fuzz v1
+[]byte("0=nan")

toml_testgen_support_test.go

@@ -8,7 +8,7 @@ import (
 	"testing"
 
 	"github.com/pelletier/go-toml/v2"
-	"github.com/pelletier/go-toml/v2/testsuite"
+	"github.com/pelletier/go-toml/v2/internal/testsuite"
 	"github.com/stretchr/testify/require"
 )
 

toml_testgen_test.go

@@ -1,4 +1,4 @@
-// Generated by tomltestgen for toml-test ref master on 2021-11-08T22:33:24-05:00
+// Generated by tomltestgen for toml-test ref master on 2022-04-07T20:09:42-04:00
 package toml_test
 
 import (
@@ -55,11 +55,51 @@ func TestTOMLTest_Invalid_Array_TextInArray(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Bool_AlmostFalseWithExtra(t *testing.T) {
+	input := "a = falsify\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_AlmostFalse(t *testing.T) {
+	input := "a = fals\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_AlmostTrueWithExtra(t *testing.T) {
+	input := "a = truthy\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_AlmostTrue(t *testing.T) {
+	input := "a = tru\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_JustF(t *testing.T) {
+	input := "a = f\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_JustT(t *testing.T) {
+	input := "a = t\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Bool_MixedCase(t *testing.T) {
 	input := "valid = False\n"
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Bool_StartingSameFalse(t *testing.T) {
+	input := "a = falsey\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Bool_StartingSameTrue(t *testing.T) {
+	input := "a = truer\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Bool_WrongCaseFalse(t *testing.T) {
 	input := "b = FALSE\n"
 	testgenInvalid(t, input)
@@ -70,6 +110,31 @@ func TestTOMLTest_Invalid_Bool_WrongCaseTrue(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Control_BareCr(t *testing.T) {
+	input := "# The following line contains a single carriage return control character\r\n\r"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Control_BareFormfeed(t *testing.T) {
+	input := "bare-formfeed = \f\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Control_BareNull(t *testing.T) {
+	input := "bare-null = \"some value\" \x00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Control_BareVerticalTab(t *testing.T) {
+	input := "bare-vertical-tab = \v\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Control_CommentCr(t *testing.T) {
+	input := "comment-cr = \"Carriage return in comment\" # \ra=1\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Control_CommentDel(t *testing.T) {
 	input := "comment-del = \"0x7f\" # \u007f\n"
 	testgenInvalid(t, input)
@@ -175,33 +240,73 @@ func TestTOMLTest_Invalid_Control_StringUs(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
-func TestTOMLTest_Invalid_Datetime_ImpossibleDate(t *testing.T) {
+func TestTOMLTest_Invalid_Datetime_HourOver(t *testing.T) {
-	input := "d = 2006-01-50T00:00:00Z\n"
+	input := "# time-hour       = 2DIGIT  ; 00-23\nd = 2006-01-01T24:00:00-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_MdayOver(t *testing.T) {
+	input := "# date-mday       = 2DIGIT  ; 01-28, 01-29, 01-30, 01-31 based on\n#                           ; month/year\nd = 2006-01-32T00:00:00-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_MdayUnder(t *testing.T) {
+	input := "# date-mday       = 2DIGIT  ; 01-28, 01-29, 01-30, 01-31 based on\n#                           ; month/year\nd = 2006-01-00T00:00:00-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_MinuteOver(t *testing.T) {
+	input := "# time-minute     = 2DIGIT  ; 00-59\nd = 2006-01-01T00:60:00-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_MonthOver(t *testing.T) {
+	input := "# date-month      = 2DIGIT  ; 01-12\nd = 2006-13-01T00:00:00-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_MonthUnder(t *testing.T) {
+	input := "# date-month      = 2DIGIT  ; 01-12\nd = 2007-00-01T00:00:00-00:00\n"
 	testgenInvalid(t, input)
 }
 
 func TestTOMLTest_Invalid_Datetime_NoLeadsWithMilli(t *testing.T) {
-	input := "with-milli = 1987-07-5T17:45:00.12Z\n"
+	input := "# Day \"5\" instead of \"05\"; the leading zero is required.\nwith-milli = 1987-07-5T17:45:00.12Z\n"
 	testgenInvalid(t, input)
 }
 
 func TestTOMLTest_Invalid_Datetime_NoLeads(t *testing.T) {
-	input := "no-leads = 1987-7-05T17:45:00Z\n"
+	input := "# Month \"7\" instead of \"07\"; the leading zero is required.\nno-leads = 1987-7-05T17:45:00Z\n"
 	testgenInvalid(t, input)
 }
 
 func TestTOMLTest_Invalid_Datetime_NoSecs(t *testing.T) {
-	input := "no-secs = 1987-07-05T17:45Z\n"
+	input := "# No seconds in time.\nno-secs = 1987-07-05T17:45Z\n"
 	testgenInvalid(t, input)
 }
 
 func TestTOMLTest_Invalid_Datetime_NoT(t *testing.T) {
-	input := "no-t = 1987-07-0517:45:00Z\n"
+	input := "# No \"t\" or \"T\" between the date and time.\nno-t = 1987-07-0517:45:00Z\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_SecondOver(t *testing.T) {
+	input := "# time-second     = 2DIGIT  ; 00-58, 00-59, 00-60 based on leap second\n#                           ; rules\nd = 2006-01-01T00:00:61-00:00\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_TimeNoLeads2(t *testing.T) {
+	input := "# Leading 0 is always required.\nd = 01:32:0\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Datetime_TimeNoLeads(t *testing.T) {
+	input := "# Leading 0 is always required.\nd = 1:32:00\n"
 	testgenInvalid(t, input)
 }
 
 func TestTOMLTest_Invalid_Datetime_TrailingT(t *testing.T) {
-	input := "d = 2006-01-30T\n"
+	input := "# Date cannot end with trailing T\nd = 2006-01-30T\n"
 	testgenInvalid(t, input)
 }
 
@@ -395,6 +500,11 @@ func TestTOMLTest_Invalid_Float_UsBeforePoint(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_InlineTable_Add(t *testing.T) {
+	input := "a={}\n# Inline tables are immutable and can't be extended\n[a.b]\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_InlineTable_DoubleComma(t *testing.T) {
 	input := "t = {x=3,,y=4}\n"
 	testgenInvalid(t, input)
@@ -435,6 +545,11 @@ func TestTOMLTest_Invalid_InlineTable_NoComma(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_InlineTable_Overwrite(t *testing.T) {
+	input := "a.b=0\n# Since table \"a\" is already defined, it can't be replaced by an inline table.\na={}\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_InlineTable_TrailingComma(t *testing.T) {
 	input := "# A terminating comma (also called trailing comma) is not permitted after the\n# last key/value pair in an inline table\nabc = { abc = 123, }\n"
 	testgenInvalid(t, input)
@@ -470,6 +585,21 @@ func TestTOMLTest_Invalid_Integer_DoubleUs(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Integer_IncompleteBin(t *testing.T) {
+	input := "incomplete-bin = 0b\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Integer_IncompleteHex(t *testing.T) {
+	input := "incomplete-hex = 0x\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Integer_IncompleteOct(t *testing.T) {
+	input := "incomplete-oct = 0o\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Integer_InvalidBin(t *testing.T) {
 	input := "invalid-bin = 0b0012\n"
 	testgenInvalid(t, input)
@@ -515,6 +645,11 @@ func TestTOMLTest_Invalid_Integer_LeadingZero2(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Integer_LeadingZero3(t *testing.T) {
+	input := "leading-zero-3 = 0_0\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Integer_LeadingZeroSign1(t *testing.T) {
 	input := "leading-zero-sign-1 = -01\n"
 	testgenInvalid(t, input)
@@ -525,6 +660,11 @@ func TestTOMLTest_Invalid_Integer_LeadingZeroSign2(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Integer_LeadingZeroSign3(t *testing.T) {
+	input := "leading-zero-sign-3 = +0_1\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Integer_NegativeBin(t *testing.T) {
 	input := "negative-bin = -0b11010110\n"
 	testgenInvalid(t, input)
@@ -730,11 +870,16 @@ func TestTOMLTest_Invalid_String_BadConcat(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
-func TestTOMLTest_Invalid_String_BadEscape(t *testing.T) {
+func TestTOMLTest_Invalid_String_BadEscape1(t *testing.T) {
 	input := "invalid-escape = \"This string has a bad \\a escape character.\"\n"
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_String_BadEscape2(t *testing.T) {
+	input := "invalid-escape = \"This string has a bad \\  escape character.\"\n\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_String_BadMultiline(t *testing.T) {
 	input := "multi = \"first line\nsecond line\"\n"
 	testgenInvalid(t, input)
@@ -805,6 +950,21 @@ func TestTOMLTest_Invalid_String_MissingQuotes(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_String_MultilineBadEscape1(t *testing.T) {
+	input := "k = \"\"\"t\\a\"\"\"\n\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_String_MultilineBadEscape2(t *testing.T) {
+	input := "# \\<Space> is not a valid escape.\nk = \"\"\"t\\ t\"\"\"\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_String_MultilineBadEscape3(t *testing.T) {
+	input := "# \\<Space> is not a valid escape.\nk = \"\"\"t\\ \"\"\"\n\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_String_MultilineEscapeSpace(t *testing.T) {
 	input := "a = \"\"\"\n  foo \\ \\n\n  bar\"\"\"\n"
 	testgenInvalid(t, input)
@@ -825,11 +985,6 @@ func TestTOMLTest_Invalid_String_MultilineQuotes1(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
-func TestTOMLTest_Invalid_String_MultilineQuotes2(t *testing.T) {
-	input := "a = \"\"\"6 quotes: \"\"\"\"\"\"\n"
-	testgenInvalid(t, input)
-}
-
 func TestTOMLTest_Invalid_String_NoClose(t *testing.T) {
 	input := "no-ending-quote = \"One time, at band camp\n"
 	testgenInvalid(t, input)
@@ -845,6 +1000,16 @@ func TestTOMLTest_Invalid_String_WrongClose(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Table_AppendWithDottedKeys1(t *testing.T) {
+	input := "# First a.b.c defines a table: a.b.c = {z=9}\n#\n# Then we define a.b.c.t = \"str\" to add a str to the above table, making it:\n#\n#   a.b.c = {z=9, t=\"...\"}\n#\n# While this makes sense, logically, it was decided this is not valid TOML as\n# it's too confusing/convoluted.\n# \n# See: https://github.com/toml-lang/toml/issues/846\n#      https://github.com/toml-lang/toml/pull/859\n\n[a.b.c]\n  z = 9\n\n[a]\n  b.c.t = \"Using dotted keys to add to [a.b.c] after explicitly defining it above is not allowed\"\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Table_AppendWithDottedKeys2(t *testing.T) {
+	input := "# This is the same issue as in injection-1.toml, except that nests one level\n# deeper. See that file for a more complete description.\n\n[a.b.c.d]\n  z = 9\n\n[a]\n  b.c.d.k.t = \"Using dotted keys to add to [a.b.c.d] after explicitly defining it above is not allowed\"\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Table_ArrayEmpty(t *testing.T) {
 	input := "[[]]\nname = \"Born to Run\"\n"
 	testgenInvalid(t, input)
@@ -860,6 +1025,16 @@ func TestTOMLTest_Invalid_Table_ArrayMissingBracket(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
+func TestTOMLTest_Invalid_Table_DuplicateKeyDottedTable(t *testing.T) {
+	input := "[fruit]\napple.color = \"red\"\n\n[fruit.apple] # INVALID\n"
+	testgenInvalid(t, input)
+}
+
+func TestTOMLTest_Invalid_Table_DuplicateKeyDottedTable2(t *testing.T) {
+	input := "[fruit]\napple.taste.sweet = true\n\n[fruit.apple.taste] # INVALID\n"
+	testgenInvalid(t, input)
+}
+
 func TestTOMLTest_Invalid_Table_DuplicateKeyTable(t *testing.T) {
 	input := "[fruit]\ntype = \"apple\"\n\n[fruit.type]\napple = \"yes\"\n"
 	testgenInvalid(t, input)
@@ -895,16 +1070,6 @@ func TestTOMLTest_Invalid_Table_EqualsSign(t *testing.T) {
 	testgenInvalid(t, input)
 }
 
-func TestTOMLTest_Invalid_Table_Injection1(t *testing.T) {
-	input := "[a.b.c]\n  z = 9\n[a]\n  b.c.t = \"Using dotted keys to add to [a.b.c] after explicitly defining it above is not allowed\"\n  \n# see https://github.com/toml-lang/toml/issues/846\n"
-	testgenInvalid(t, input)
-}
-
-func TestTOMLTest_Invalid_Table_Injection2(t *testing.T) {
-	input := "[a.b.c.d]\n  z = 9\n[a]\n  b.c.d.k.t = \"Using dotted keys to add to [a.b.c.d] after explicitly defining it above is not allowed\"\n  \n# see https://github.com/toml-lang/toml/issues/846\n"
-	testgenInvalid(t, input)
-}
-
 func TestTOMLTest_Invalid_Table_Llbrace(t *testing.T) {
 	input := "[ [table]]\n"
 	testgenInvalid(t, input)
@@ -1071,8 +1236,14 @@ func TestTOMLTest_Valid_Comment_AtEof2(t *testing.T) {
 }
 
 func TestTOMLTest_Valid_Comment_Everywhere(t *testing.T) {
-	input := "# Top comment.\n  # Top comment.\n# Top comment.\n\n# [no-extraneous-groups-please]\n\n[group] # Comment\nanswer = 42 # Comment\n# no-extraneous-keys-please = 999\n# Inbetween comment.\nmore = [ # Comment\n  # What about multiple # comments?\n  # Can you handle it?\n  #\n          # Evil.\n# Evil.\n  42, 42, # Comments within arrays are fun.\n  # What about multiple # comments?\n  # Can you handle it?\n  #\n          # Evil.\n# Evil.\n# ] Did I fool you?\n] # Hopefully not.\n\n# Make sure the space between the datetime and \"#\" isn't lexed.\nd = 1979-05-27T07:32:12-07:00  # c\n"
+	input := "# Top comment.\n  # Top comment.\n# Top comment.\n\n# [no-extraneous-groups-please]\n\n[group] # Comment\nanswer = 42 # Comment\n# no-extraneous-keys-please = 999\n# Inbetween comment.\nmore = [ # Comment\n  # What about multiple # comments?\n  # Can you handle it?\n  #\n          # Evil.\n# Evil.\n  42, 42, # Comments within arrays are fun.\n  # What about multiple # comments?\n  # Can you handle it?\n  #\n          # Evil.\n# Evil.\n# ] Did I fool you?\n] # Hopefully not.\n\n# Make sure the space between the datetime and \"#\" isn't lexed.\ndt = 1979-05-27T07:32:12-07:00  # c\nd = 1979-05-27 # Comment\n"
-	jsonRef := "{\n  \"group\": {\n    \"answer\": {\n      \"type\": \"integer\",\n      \"value\": \"42\"\n    },\n    \"d\": {\n      \"type\": \"datetime\",\n      \"value\": \"1979-05-27T07:32:12-07:00\"\n    },\n    \"more\": [\n      {\n        \"type\": \"integer\",\n        \"value\": \"42\"\n      },\n      {\n        \"type\": \"integer\",\n        \"value\": \"42\"\n      }\n    ]\n  }\n}\n"
+	jsonRef := "{\n  \"group\": {\n    \"answer\": {\n      \"type\": \"integer\",\n      \"value\": \"42\"\n    },\n    \"dt\": {\n      \"type\": \"datetime\",\n      \"value\": \"1979-05-27T07:32:12-07:00\"\n    },\n    \"d\": {\n      \"type\": \"date-local\",\n      \"value\": \"1979-05-27\"\n    },\n    \"more\": [\n      {\n        \"type\": \"integer\",\n        \"value\": \"42\"\n      },\n      {\n        \"type\": \"integer\",\n        \"value\": \"42\"\n      }\n    ]\n  }\n}\n"
+	testgenValid(t, input, jsonRef)
+}
+
+func TestTOMLTest_Valid_Comment_Noeol(t *testing.T) {
+	input := "# single comment without any eol characters"
+	jsonRef := "{}\n"
 	testgenValid(t, input, jsonRef)
 }
 
@@ -1107,8 +1278,8 @@ func TestTOMLTest_Valid_Datetime_Local(t *testing.T) {
 }
 
 func TestTOMLTest_Valid_Datetime_Milliseconds(t *testing.T) {
-	input := "utc1  = 1987-07-05T17:45:56.123456Z\nutc2  = 1987-07-05T17:45:56.6Z\nwita1 = 1987-07-05T17:45:56.123456+08:00\nwita2 = 1987-07-05T17:45:56.6+08:00\n"
+	input := "utc1  = 1987-07-05T17:45:56.1234Z\nutc2  = 1987-07-05T17:45:56.6Z\nwita1 = 1987-07-05T17:45:56.1234+08:00\nwita2 = 1987-07-05T17:45:56.6+08:00\n"
-	jsonRef := "{\n  \"utc1\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.123456Z\"\n  },\n  \"utc2\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.600000Z\"\n  },\n  \"wita1\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.123456+08:00\"\n  },\n  \"wita2\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.600000+08:00\"\n  }\n}\n"
+	jsonRef := "{\n  \"utc1\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.1234Z\"\n  },\n  \"utc2\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.6000Z\"\n  },\n  \"wita1\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.1234+08:00\"\n  },\n  \"wita2\": {\n    \"type\": \"datetime\",\n    \"value\": \"1987-07-05T17:45:56.6000+08:00\"\n  }\n}\n"
 	testgenValid(t, input, jsonRef)
 }
 
@@ -1370,6 +1541,12 @@ func TestTOMLTest_Valid_String_Empty(t *testing.T) {
 	testgenValid(t, input, jsonRef)
 }
 
+func TestTOMLTest_Valid_String_EscapeEsc(t *testing.T) {
+	input := "esc = \"\\e There is no escape! \\e\"\n"
+	jsonRef := "{\n  \"esc\": {\n    \"type\": \"string\",\n    \"value\": \"\\u001b There is no escape! \\u001b\"\n  }\n}\n"
+	testgenValid(t, input, jsonRef)
+}
+
 func TestTOMLTest_Valid_String_EscapeTricky(t *testing.T) {
 	input := "end_esc = \"String does not end here\\\" but ends here\\\\\"\nlit_end_esc = 'String ends here\\'\n\nmultiline_unicode = \"\"\"\n\\u00a0\"\"\"\n\nmultiline_not_unicode = \"\"\"\n\\\\u0041\"\"\"\n\nmultiline_end_esc = \"\"\"When will it end? \\\"\"\"...\"\"\\\" should be here\\\"\"\"\"\n\nlit_multiline_not_unicode = '''\n\\u007f'''\n\nlit_multiline_end = '''There is no escape\\'''\n"
 	jsonRef := "{\n  \"end_esc\": {\n    \"type\": \"string\",\n    \"value\": \"String does not end here\\\" but ends here\\\\\"\n  },\n  \"lit_end_esc\": {\n    \"type\": \"string\",\n    \"value\": \"String ends here\\\\\"\n  },\n  \"lit_multiline_end\": {\n    \"type\": \"string\",\n    \"value\": \"There is no escape\\\\\"\n  },\n  \"lit_multiline_not_unicode\": {\n    \"type\": \"string\",\n    \"value\": \"\\\\u007f\"\n  },\n  \"multiline_end_esc\": {\n    \"type\": \"string\",\n    \"value\": \"When will it end? \\\"\\\"\\\"...\\\"\\\"\\\" should be here\\\"\"\n  },\n  \"multiline_not_unicode\": {\n    \"type\": \"string\",\n    \"value\": \"\\\\u0041\"\n  },\n  \"multiline_unicode\": {\n    \"type\": \"string\",\n    \"value\": \"\u00a0\"\n  }\n}\n"
@@ -1388,14 +1565,20 @@ func TestTOMLTest_Valid_String_Escapes(t *testing.T) {
 	testgenValid(t, input, jsonRef)
 }
 
+func TestTOMLTest_Valid_String_MultilineEscapedCrlf(t *testing.T) {
+	input := "# The following line should be an unescaped backslash followed by a Windows\r\n# newline sequence (\"\\r\\n\")\r\n0=\"\"\"\\\r\n\"\"\"\r\n"
+	jsonRef := "{\n  \"0\": {\n    \"type\": \"string\",\n    \"value\": \"\"\n  }\n}\n"
+	testgenValid(t, input, jsonRef)
+}
+
 func TestTOMLTest_Valid_String_MultilineQuotes(t *testing.T) {
-	input := "# Make sure that quotes inside multiline strings are allowed, including right\n# after the opening '''/\"\"\" and before the closing '''/\"\"\"\n\nlit_one = ''''one quote''''\nlit_two = '''''two quotes'''''\nlit_one_space = ''' 'one quote' '''\nlit_two_space = ''' ''two quotes'' '''\n\none = \"\"\"\"one quote\"\"\"\"\ntwo = \"\"\"\"\"two quotes\"\"\"\"\"\none_space = \"\"\" \"one quote\" \"\"\"\ntwo_space = \"\"\" \"\"two quotes\"\" \"\"\"\n\nmismatch1 = \"\"\"aaa'''bbb\"\"\"\nmismatch2 = '''aaa\"\"\"bbb'''\n"
+	input := "# Make sure that quotes inside multiline strings are allowed, including right\n# after the opening '''/\"\"\" and before the closing '''/\"\"\"\n\nlit_one = ''''one quote''''\nlit_two = '''''two quotes'''''\nlit_one_space = ''' 'one quote' '''\nlit_two_space = ''' ''two quotes'' '''\n\none = \"\"\"\"one quote\"\"\"\"\ntwo = \"\"\"\"\"two quotes\"\"\"\"\"\none_space = \"\"\" \"one quote\" \"\"\"\ntwo_space = \"\"\" \"\"two quotes\"\" \"\"\"\n\nmismatch1 = \"\"\"aaa'''bbb\"\"\"\nmismatch2 = '''aaa\"\"\"bbb'''\n\n# Three opening \"\"\", then one escaped \", then two \"\" (allowed), and then three\n# closing \"\"\"\nescaped = \"\"\"lol\\\"\"\"\"\"\"\n"
-	jsonRef := "{\n  \"lit_one\": {\n    \"type\": \"string\",\n    \"value\": \"'one quote'\"\n  },\n  \"lit_one_space\": {\n    \"type\": \"string\",\n    \"value\": \" 'one quote' \"\n  },\n  \"lit_two\": {\n    \"type\": \"string\",\n    \"value\": \"''two quotes''\"\n  },\n  \"lit_two_space\": {\n    \"type\": \"string\",\n    \"value\": \" ''two quotes'' \"\n  },\n  \"mismatch1\": {\n    \"type\": \"string\",\n    \"value\": \"aaa'''bbb\"\n  },\n  \"mismatch2\": {\n    \"type\": \"string\",\n    \"value\": \"aaa\\\"\\\"\\\"bbb\"\n  },\n  \"one\": {\n    \"type\": \"string\",\n    \"value\": \"\\\"one quote\\\"\"\n  },\n  \"one_space\": {\n    \"type\": \"string\",\n    \"value\": \" \\\"one quote\\\" \"\n  },\n  \"two\": {\n    \"type\": \"string\",\n    \"value\": \"\\\"\\\"two quotes\\\"\\\"\"\n  },\n  \"two_space\": {\n    \"type\": \"string\",\n    \"value\": \" \\\"\\\"two quotes\\\"\\\" \"\n  }\n}\n"
+	jsonRef := "{\n  \"escaped\": {\n    \"type\": \"string\",\n    \"value\": \"lol\\\"\\\"\\\"\"\n  },\n  \"lit_one\": {\n    \"type\": \"string\",\n    \"value\": \"'one quote'\"\n  },\n  \"lit_one_space\": {\n    \"type\": \"string\",\n    \"value\": \" 'one quote' \"\n  },\n  \"lit_two\": {\n    \"type\": \"string\",\n    \"value\": \"''two quotes''\"\n  },\n  \"lit_two_space\": {\n    \"type\": \"string\",\n    \"value\": \" ''two quotes'' \"\n  },\n  \"mismatch1\": {\n    \"type\": \"string\",\n    \"value\": \"aaa'''bbb\"\n  },\n  \"mismatch2\": {\n    \"type\": \"string\",\n    \"value\": \"aaa\\\"\\\"\\\"bbb\"\n  },\n  \"one\": {\n    \"type\": \"string\",\n    \"value\": \"\\\"one quote\\\"\"\n  },\n  \"one_space\": {\n    \"type\": \"string\",\n    \"value\": \" \\\"one quote\\\" \"\n  },\n  \"two\": {\n    \"type\": \"string\",\n    \"value\": \"\\\"\\\"two quotes\\\"\\\"\"\n  },\n  \"two_space\": {\n    \"type\": \"string\",\n    \"value\": \" \\\"\\\"two quotes\\\"\\\" \"\n  }\n}\n"
 	testgenValid(t, input, jsonRef)
 }
 
 func TestTOMLTest_Valid_String_Multiline(t *testing.T) {
-	input := "# NOTE: this file includes some literal tab characters.\n\nmultiline_empty_one = \"\"\"\"\"\"\nmultiline_empty_two = \"\"\"\n\"\"\"\nmultiline_empty_three = \"\"\"\\\n    \"\"\"\nmultiline_empty_four = \"\"\"\\\n   \\\n   \\  \n   \"\"\"\n\nequivalent_one = \"The quick brown fox jumps over the lazy dog.\"\nequivalent_two = \"\"\"\nThe quick brown \\\n\n\n  fox jumps over \\\n    the lazy dog.\"\"\"\n\nequivalent_three = \"\"\"\\\n       The quick brown \\\n       fox jumps over \\\n       the lazy dog.\\\n       \"\"\"\n\nwhitespace-after-bs = \"\"\"\\\n       The quick brown \\\n       fox jumps over \\   \n       the lazy dog.\\\t\n       \"\"\"\n\nno-space = \"\"\"a\\\n    b\"\"\"\n\nkeep-ws-before = \"\"\"a   \t\\\n   b\"\"\"\n\nescape-bs-1 = \"\"\"a \\\\\nb\"\"\"\n\nescape-bs-2 = \"\"\"a \\\\\\\nb\"\"\"\n\nescape-bs-3 = \"\"\"a \\\\\\\\\n  b\"\"\"\n"
+	input := "# NOTE: this file includes some literal tab characters.\n\nmultiline_empty_one = \"\"\"\"\"\"\n\n# A newline immediately following the opening delimiter will be trimmed.\nmultiline_empty_two = \"\"\"\n\"\"\"\n\n# \\ at the end of line trims newlines as well; note that last \\ is followed by\n# two spaces, which are ignored.\nmultiline_empty_three = \"\"\"\\\n    \"\"\"\nmultiline_empty_four = \"\"\"\\\n   \\\n   \\  \n   \"\"\"\n\nequivalent_one = \"The quick brown fox jumps over the lazy dog.\"\nequivalent_two = \"\"\"\nThe quick brown \\\n\n\n  fox jumps over \\\n    the lazy dog.\"\"\"\n\nequivalent_three = \"\"\"\\\n       The quick brown \\\n       fox jumps over \\\n       the lazy dog.\\\n       \"\"\"\n\nwhitespace-after-bs = \"\"\"\\\n       The quick brown \\\n       fox jumps over \\   \n       the lazy dog.\\\t\n       \"\"\"\n\nno-space = \"\"\"a\\\n    b\"\"\"\n\n# Has tab character.\nkeep-ws-before = \"\"\"a   \t\\\n   b\"\"\"\n\nescape-bs-1 = \"\"\"a \\\\\nb\"\"\"\n\nescape-bs-2 = \"\"\"a \\\\\\\nb\"\"\"\n\nescape-bs-3 = \"\"\"a \\\\\\\\\n  b\"\"\"\n"
 	jsonRef := "{\n  \"equivalent_one\": {\n    \"type\": \"string\",\n    \"value\": \"The quick brown fox jumps over the lazy dog.\"\n  },\n  \"equivalent_three\": {\n    \"type\": \"string\",\n    \"value\": \"The quick brown fox jumps over the lazy dog.\"\n  },\n  \"equivalent_two\": {\n    \"type\": \"string\",\n    \"value\": \"The quick brown fox jumps over the lazy dog.\"\n  },\n  \"escape-bs-1\": {\n    \"type\": \"string\",\n    \"value\": \"a \\\\\\nb\"\n  },\n  \"escape-bs-2\": {\n    \"type\": \"string\",\n    \"value\": \"a \\\\b\"\n  },\n  \"escape-bs-3\": {\n    \"type\": \"string\",\n    \"value\": \"a \\\\\\\\\\n  b\"\n  },\n  \"keep-ws-before\": {\n    \"type\": \"string\",\n    \"value\": \"a   \\tb\"\n  },\n  \"multiline_empty_four\": {\n    \"type\": \"string\",\n    \"value\": \"\"\n  },\n  \"multiline_empty_one\": {\n    \"type\": \"string\",\n    \"value\": \"\"\n  },\n  \"multiline_empty_three\": {\n    \"type\": \"string\",\n    \"value\": \"\"\n  },\n  \"multiline_empty_two\": {\n    \"type\": \"string\",\n    \"value\": \"\"\n  },\n  \"no-space\": {\n    \"type\": \"string\",\n    \"value\": \"ab\"\n  },\n  \"whitespace-after-bs\": {\n    \"type\": \"string\",\n    \"value\": \"The quick brown fox jumps over the lazy dog.\"\n  }\n}\n"
 	testgenValid(t, input, jsonRef)
 }
@@ -1407,7 +1590,7 @@ func TestTOMLTest_Valid_String_Nl(t *testing.T) {
 }
 
 func TestTOMLTest_Valid_String_RawMultiline(t *testing.T) {
-	input := "oneline = '''This string has a ' quote character.'''\nfirstnl = '''\nThis string has a ' quote character.'''\nmultiline = '''\nThis string\nhas ' a quote character\nand more than\none newline\nin it.'''\n"
+	input := "# Single ' should be allowed.\noneline = '''This string has a ' quote character.'''\n\n# A newline immediately following the opening delimiter will be trimmed.\nfirstnl = '''\nThis string has a ' quote character.'''\n\n# All other whitespace and newline characters remain intact.\nmultiline = '''\nThis string\nhas ' a quote character\nand more than\none newline\nin it.'''\n"
 	jsonRef := "{\n  \"firstnl\": {\n    \"type\": \"string\",\n    \"value\": \"This string has a ' quote character.\"\n  },\n  \"multiline\": {\n    \"type\": \"string\",\n    \"value\": \"This string\\nhas ' a quote character\\nand more than\\none newline\\nin it.\"\n  },\n  \"oneline\": {\n    \"type\": \"string\",\n    \"value\": \"This string has a ' quote character.\"\n  }\n}\n"
 	testgenValid(t, input, jsonRef)
 }

types.go

@@ -6,9 +6,9 @@ import (
 	"time"
 )
 
-var timeType = reflect.TypeOf(time.Time{})
+var timeType = reflect.TypeOf((*time.Time)(nil)).Elem()
-var textMarshalerType = reflect.TypeOf(new(encoding.TextMarshaler)).Elem()
+var textMarshalerType = reflect.TypeOf((*encoding.TextMarshaler)(nil)).Elem()
-var textUnmarshalerType = reflect.TypeOf(new(encoding.TextUnmarshaler)).Elem()
+var textUnmarshalerType = reflect.TypeOf((*encoding.TextUnmarshaler)(nil)).Elem()
-var mapStringInterfaceType = reflect.TypeOf(map[string]interface{}{})
+var mapStringInterfaceType = reflect.TypeOf(map[string]interface{}(nil))
-var sliceInterfaceType = reflect.TypeOf([]interface{}{})
+var sliceInterfaceType = reflect.TypeOf([]interface{}(nil))
 var stringType = reflect.TypeOf("")

unmarshaler.go

@@ -12,16 +12,16 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/pelletier/go-toml/v2/internal/ast"
 	"github.com/pelletier/go-toml/v2/internal/danger"
 	"github.com/pelletier/go-toml/v2/internal/tracker"
+	"github.com/pelletier/go-toml/v2/unstable"
 )
 
 // Unmarshal deserializes a TOML document into a Go value.
 //
 // It is a shortcut for Decoder.Decode() with the default options.
 func Unmarshal(data []byte, v interface{}) error {
-	p := parser{}
+	p := unstable.Parser{}
 	p.Reset(data)
 	d := decoder{p: &p}
 
@@ -42,24 +42,25 @@ func NewDecoder(r io.Reader) *Decoder {
 	return &Decoder{r: r}
 }
 
-// SetStrict toggles decoding in stict mode.
+// DisallowUnknownFields causes the Decoder to return an error when the
+// destination is a struct and the input contains a key that does not match a
+// non-ignored field.
 //
-// When the decoder is in strict mode, it will record fields from the document
+// In that case, the Decoder returns a StrictMissingError that can be used to
-// that could not be set on the target value. In that case, the decoder returns
+// retrieve the individual errors as well as generate a human readable
-// a StrictMissingError that can be used to retrieve the individual errors as
+// description of the missing fields.
-// well as generate a human readable description of the missing fields.
+func (d *Decoder) DisallowUnknownFields() *Decoder {
-func (d *Decoder) SetStrict(strict bool) *Decoder {
+	d.strict = true
-	d.strict = strict
 	return d
 }
 
 // Decode the whole content of r into v.
 //
 // By default, values in the document that don't exist in the target Go value
-// are ignored. See Decoder.SetStrict() to change this behavior.
+// are ignored. See Decoder.DisallowUnknownFields() to change this behavior.
 //
 // When a TOML local date, time, or date-time is decoded into a time.Time, its
-// value is represented in time.Local timezone. Otherwise the approriate Local*
+// value is represented in time.Local timezone. Otherwise the appropriate Local*
 // structure is used. For time values, precision up to the nanosecond is
 // supported by truncating extra digits.
 //
@@ -78,29 +79,29 @@ func (d *Decoder) SetStrict(strict bool) *Decoder {
 // strict mode and a field is missing, a `toml.StrictMissingError` is
 // returned. In any other case, this function returns a standard Go error.
 //
-// Type mapping
+// # Type mapping
 //
 // List of supported TOML types and their associated accepted Go types:
 //
-//   String           -> string
+//	String           -> string
-//   Integer          -> uint*, int*, depending on size
+//	Integer          -> uint*, int*, depending on size
-//   Float            -> float*, depending on size
+//	Float            -> float*, depending on size
-//   Boolean          -> bool
+//	Boolean          -> bool
-//   Offset Date-Time -> time.Time
+//	Offset Date-Time -> time.Time
-//   Local Date-time  -> LocalDateTime, time.Time
+//	Local Date-time  -> LocalDateTime, time.Time
-//   Local Date       -> LocalDate, time.Time
+//	Local Date       -> LocalDate, time.Time
-//   Local Time       -> LocalTime, time.Time
+//	Local Time       -> LocalTime, time.Time
-//   Array            -> slice and array, depending on elements types
+//	Array            -> slice and array, depending on elements types
-//   Table            -> map and struct
+//	Table            -> map and struct
-//   Inline Table     -> same as Table
+//	Inline Table     -> same as Table
-//   Array of Tables  -> same as Array and Table
+//	Array of Tables  -> same as Array and Table
 func (d *Decoder) Decode(v interface{}) error {
 	b, err := ioutil.ReadAll(d.r)
 	if err != nil {
 		return fmt.Errorf("toml: %w", err)
 	}
 
-	p := parser{}
+	p := unstable.Parser{}
 	p.Reset(b)
 	dec := decoder{
 		p: &p,
@@ -114,7 +115,7 @@ func (d *Decoder) Decode(v interface{}) error {
 
 type decoder struct {
 	// Which parser instance in use for this decoding session.
-	p *parser
+	p *unstable.Parser
 
 	// Flag indicating that the current expression is stashed.
 	// If set to true, calling nextExpr will not actually pull a new expression
@@ -122,7 +123,7 @@ type decoder struct {
 	stashedExpr bool
 
 	// Skip expressions until a table is found. This is set to true when a
-	// table could not be create (missing field in map), so all KV expressions
+	// table could not be created (missing field in map), so all KV expressions
 	// need to be skipped.
 	skipUntilTable bool
 
@@ -148,15 +149,19 @@ type errorContext struct {
 }
 
 func (d *decoder) typeMismatchError(toml string, target reflect.Type) error {
+	return fmt.Errorf("toml: %s", d.typeMismatchString(toml, target))
+}
+
+func (d *decoder) typeMismatchString(toml string, target reflect.Type) string {
 	if d.errorContext != nil && d.errorContext.Struct != nil {
 		ctx := d.errorContext
 		f := ctx.Struct.FieldByIndex(ctx.Field)
-		return fmt.Errorf("toml: cannot decode TOML %s into struct field %s.%s of type %s", toml, ctx.Struct, f.Name, f.Type)
+		return fmt.Sprintf("cannot decode TOML %s into struct field %s.%s of type %s", toml, ctx.Struct, f.Name, f.Type)
 	}
-	return fmt.Errorf("toml: cannot decode TOML %s into a Go value of type %s", toml, target)
+	return fmt.Sprintf("cannot decode TOML %s into a Go value of type %s", toml, target)
 }
 
-func (d *decoder) expr() *ast.Node {
+func (d *decoder) expr() *unstable.Node {
 	return d.p.Expression()
 }
 
@@ -207,12 +212,12 @@ func (d *decoder) FromParser(v interface{}) error {
 
 	err := d.fromParser(r)
 	if err == nil {
-		return d.strict.Error(d.p.data)
+		return d.strict.Error(d.p.Data())
 	}
 
-	var e *decodeError
+	var e *unstable.ParserError
 	if errors.As(err, &e) {
-		return wrapDecodeError(d.p.data, e)
+		return wrapDecodeError(d.p.Data(), e)
 	}
 
 	return err
@@ -233,16 +238,16 @@ func (d *decoder) fromParser(root reflect.Value) error {
 Rules for the unmarshal code:
 
 - The stack is used to keep track of which values need to be set where.
-- handle* functions <=> switch on a given ast.Kind.
+- handle* functions <=> switch on a given unstable.Kind.
 - unmarshalX* functions need to unmarshal a node of kind X.
 - An "object" is either a struct or a map.
 */
 
-func (d *decoder) handleRootExpression(expr *ast.Node, v reflect.Value) error {
+func (d *decoder) handleRootExpression(expr *unstable.Node, v reflect.Value) error {
 	var x reflect.Value
 	var err error
 
-	if !(d.skipUntilTable && expr.Kind == ast.KeyValue) {
+	if !(d.skipUntilTable && expr.Kind == unstable.KeyValue) {
 		err = d.seen.CheckExpression(expr)
 		if err != nil {
 			return err
@@ -250,16 +255,16 @@ func (d *decoder) handleRootExpression(expr *ast.Node, v reflect.Value) error {
 	}
 
 	switch expr.Kind {
-	case ast.KeyValue:
+	case unstable.KeyValue:
 		if d.skipUntilTable {
 			return nil
 		}
 		x, err = d.handleKeyValue(expr, v)
-	case ast.Table:
+	case unstable.Table:
 		d.skipUntilTable = false
 		d.strict.EnterTable(expr)
 		x, err = d.handleTable(expr.Key(), v)
-	case ast.ArrayTable:
+	case unstable.ArrayTable:
 		d.skipUntilTable = false
 		d.strict.EnterArrayTable(expr)
 		x, err = d.handleArrayTable(expr.Key(), v)
@@ -268,7 +273,7 @@ func (d *decoder) handleRootExpression(expr *ast.Node, v reflect.Value) error {
 	}
 
 	if d.skipUntilTable {
-		if expr.Kind == ast.Table || expr.Kind == ast.ArrayTable {
+		if expr.Kind == unstable.Table || expr.Kind == unstable.ArrayTable {
 			d.strict.MissingTable(expr)
 		}
 	} else if err == nil && x.IsValid() {
@@ -278,14 +283,14 @@ func (d *decoder) handleRootExpression(expr *ast.Node, v reflect.Value) error {
 	return err
 }
 
-func (d *decoder) handleArrayTable(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleArrayTable(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	if key.Next() {
 		return d.handleArrayTablePart(key, v)
 	}
 	return d.handleKeyValues(v)
 }
 
-func (d *decoder) handleArrayTableCollectionLast(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleArrayTableCollectionLast(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	switch v.Kind() {
 	case reflect.Interface:
 		elem := v.Elem()
@@ -321,10 +326,12 @@ func (d *decoder) handleArrayTableCollectionLast(key ast.Iterator, v reflect.Val
 		return v, nil
 	case reflect.Slice:
 		elemType := v.Type().Elem()
+		var elem reflect.Value
 		if elemType.Kind() == reflect.Interface {
-			elemType = mapStringInterfaceType
+			elem = makeMapStringInterface()
+		} else {
+			elem = reflect.New(elemType).Elem()
 		}
-		elem := reflect.New(elemType).Elem()
 		elem2, err := d.handleArrayTable(key, elem)
 		if err != nil {
 			return reflect.Value{}, err
@@ -336,21 +343,21 @@ func (d *decoder) handleArrayTableCollectionLast(key ast.Iterator, v reflect.Val
 	case reflect.Array:
 		idx := d.arrayIndex(true, v)
 		if idx >= v.Len() {
-			return v, fmt.Errorf("toml: cannot decode array table into %s at position %d", v.Type(), idx)
+			return v, fmt.Errorf("%s at position %d", d.typeMismatchError("array table", v.Type()), idx)
 		}
 		elem := v.Index(idx)
 		_, err := d.handleArrayTable(key, elem)
 		return v, err
+	default:
+		return reflect.Value{}, d.typeMismatchError("array table", v.Type())
 	}
-
-	return d.handleArrayTable(key, v)
 }
 
 // When parsing an array table expression, each part of the key needs to be
 // evaluated like a normal key, but if it returns a collection, it also needs to
 // point to the last element of the collection. Unless it is the last part of
 // the key, then it needs to create a new element at the end.
-func (d *decoder) handleArrayTableCollection(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleArrayTableCollection(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	if key.IsLast() {
 		return d.handleArrayTableCollectionLast(key, v)
 	}
@@ -387,7 +394,7 @@ func (d *decoder) handleArrayTableCollection(key ast.Iterator, v reflect.Value)
 	case reflect.Array:
 		idx := d.arrayIndex(false, v)
 		if idx >= v.Len() {
-			return v, fmt.Errorf("toml: cannot decode array table into %s at position %d", v.Type(), idx)
+			return v, fmt.Errorf("%s at position %d", d.typeMismatchError("array table", v.Type()), idx)
 		}
 		elem := v.Index(idx)
 		_, err := d.handleArrayTable(key, elem)
@@ -397,7 +404,7 @@ func (d *decoder) handleArrayTableCollection(key ast.Iterator, v reflect.Value)
 	return d.handleArrayTable(key, v)
 }
 
-func (d *decoder) handleKeyPart(key ast.Iterator, v reflect.Value, nextFn handlerFn, makeFn valueMakerFn) (reflect.Value, error) {
+func (d *decoder) handleKeyPart(key unstable.Iterator, v reflect.Value, nextFn handlerFn, makeFn valueMakerFn) (reflect.Value, error) {
 	var rv reflect.Value
 
 	// First, dispatch over v to make sure it is a valid object.
@@ -411,9 +418,13 @@ func (d *decoder) handleKeyPart(key ast.Iterator, v reflect.Value, nextFn handle
 		elem = v.Elem()
 		return d.handleKeyPart(key, elem, nextFn, makeFn)
 	case reflect.Map:
+		vt := v.Type()
 
-		// Create the key for the map element. For now assume it's a string.
+		// Create the key for the map element. Convert to key type.
-		mk := reflect.ValueOf(string(key.Node().Data))
+		mk, err := d.keyFromData(vt.Key(), key.Node().Data)
+		if err != nil {
+			return reflect.Value{}, err
+		}
 
 		// If the map does not exist, create it.
 		if v.IsNil() {
@@ -430,7 +441,6 @@ func (d *decoder) handleKeyPart(key ast.Iterator, v reflect.Value, nextFn handle
 			// map[string]interface{} or a []interface{} depending on whether
 			// this is the last part of the array table key.
 
-			vt := v.Type()
 			t := vt.Elem()
 			if t.Kind() == reflect.Interface {
 				mv = makeFn()
@@ -480,7 +490,7 @@ func (d *decoder) handleKeyPart(key ast.Iterator, v reflect.Value, nextFn handle
 		d.errorContext.Struct = t
 		d.errorContext.Field = path
 
-		f := v.FieldByIndex(path)
+		f := fieldByIndex(v, path)
 		x, err := nextFn(key, f)
 		if err != nil || d.skipUntilTable {
 			return reflect.Value{}, err
@@ -515,7 +525,7 @@ func (d *decoder) handleKeyPart(key ast.Iterator, v reflect.Value, nextFn handle
 // HandleArrayTablePart navigates the Go structure v using the key v. It is
 // only used for the prefix (non-last) parts of an array-table. When
 // encountering a collection, it should go to the last element.
-func (d *decoder) handleArrayTablePart(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleArrayTablePart(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	var makeFn valueMakerFn
 	if key.IsLast() {
 		makeFn = makeSliceInterface
@@ -527,10 +537,10 @@ func (d *decoder) handleArrayTablePart(key ast.Iterator, v reflect.Value) (refle
 
 // HandleTable returns a reference when it has checked the next expression but
 // cannot handle it.
-func (d *decoder) handleTable(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleTable(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	if v.Kind() == reflect.Slice {
 		if v.Len() == 0 {
-			return reflect.Value{}, newDecodeError(key.Node().Data, "cannot store a table in a slice")
+			return reflect.Value{}, unstable.NewParserError(key.Node().Data, "cannot store a table in a slice")
 		}
 		elem := v.Index(v.Len() - 1)
 		x, err := d.handleTable(key, elem)
@@ -557,7 +567,7 @@ func (d *decoder) handleKeyValues(v reflect.Value) (reflect.Value, error) {
 	var rv reflect.Value
 	for d.nextExpr() {
 		expr := d.expr()
-		if expr.Kind != ast.KeyValue {
+		if expr.Kind != unstable.KeyValue {
 			// Stash the expression so that fromParser can just loop and use
 			// the right handler.
 			// We could just recurse ourselves here, but at least this gives a
@@ -584,7 +594,7 @@ func (d *decoder) handleKeyValues(v reflect.Value) (reflect.Value, error) {
 }
 
 type (
-	handlerFn    func(key ast.Iterator, v reflect.Value) (reflect.Value, error)
+	handlerFn    func(key unstable.Iterator, v reflect.Value) (reflect.Value, error)
 	valueMakerFn func() reflect.Value
 )
 
@@ -596,11 +606,11 @@ func makeSliceInterface() reflect.Value {
 	return reflect.MakeSlice(sliceInterfaceType, 0, 16)
 }
 
-func (d *decoder) handleTablePart(key ast.Iterator, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleTablePart(key unstable.Iterator, v reflect.Value) (reflect.Value, error) {
 	return d.handleKeyPart(key, v, d.handleTable, makeMapStringInterface)
 }
 
-func (d *decoder) tryTextUnmarshaler(node *ast.Node, v reflect.Value) (bool, error) {
+func (d *decoder) tryTextUnmarshaler(node *unstable.Node, v reflect.Value) (bool, error) {
 	// Special case for time, because we allow to unmarshal to it from
 	// different kind of AST nodes.
 	if v.Type() == timeType {
@@ -610,7 +620,7 @@ func (d *decoder) tryTextUnmarshaler(node *ast.Node, v reflect.Value) (bool, err
 	if v.CanAddr() && v.Addr().Type().Implements(textUnmarshalerType) {
 		err := v.Addr().Interface().(encoding.TextUnmarshaler).UnmarshalText(node.Data)
 		if err != nil {
-			return false, newDecodeError(d.p.Raw(node.Raw), "error calling UnmarshalText: %w", err)
+			return false, unstable.NewParserError(d.p.Raw(node.Raw), "%w", err)
 		}
 
 		return true, nil
@@ -619,7 +629,7 @@ func (d *decoder) tryTextUnmarshaler(node *ast.Node, v reflect.Value) (bool, err
 	return false, nil
 }
 
-func (d *decoder) handleValue(value *ast.Node, v reflect.Value) error {
+func (d *decoder) handleValue(value *unstable.Node, v reflect.Value) error {
 	for v.Kind() == reflect.Ptr {
 		v = initAndDereferencePointer(v)
 	}
@@ -630,32 +640,32 @@ func (d *decoder) handleValue(value *ast.Node, v reflect.Value) error {
 	}
 
 	switch value.Kind {
-	case ast.String:
+	case unstable.String:
 		return d.unmarshalString(value, v)
-	case ast.Integer:
+	case unstable.Integer:
 		return d.unmarshalInteger(value, v)
-	case ast.Float:
+	case unstable.Float:
 		return d.unmarshalFloat(value, v)
-	case ast.Bool:
+	case unstable.Bool:
 		return d.unmarshalBool(value, v)
-	case ast.DateTime:
+	case unstable.DateTime:
 		return d.unmarshalDateTime(value, v)
-	case ast.LocalDate:
+	case unstable.LocalDate:
 		return d.unmarshalLocalDate(value, v)
-	case ast.LocalTime:
+	case unstable.LocalTime:
 		return d.unmarshalLocalTime(value, v)
-	case ast.LocalDateTime:
+	case unstable.LocalDateTime:
 		return d.unmarshalLocalDateTime(value, v)
-	case ast.InlineTable:
+	case unstable.InlineTable:
 		return d.unmarshalInlineTable(value, v)
-	case ast.Array:
+	case unstable.Array:
 		return d.unmarshalArray(value, v)
 	default:
 		panic(fmt.Errorf("handleValue not implemented for %s", value.Kind))
 	}
 }
 
-func (d *decoder) unmarshalArray(array *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalArray(array *unstable.Node, v reflect.Value) error {
 	switch v.Kind() {
 	case reflect.Slice:
 		if v.IsNil() {
@@ -726,7 +736,7 @@ func (d *decoder) unmarshalArray(array *ast.Node, v reflect.Value) error {
 	return nil
 }
 
-func (d *decoder) unmarshalInlineTable(itable *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalInlineTable(itable *unstable.Node, v reflect.Value) error {
 	// Make sure v is an initialized object.
 	switch v.Kind() {
 	case reflect.Map:
@@ -743,7 +753,7 @@ func (d *decoder) unmarshalInlineTable(itable *ast.Node, v reflect.Value) error
 		}
 		return d.unmarshalInlineTable(itable, elem)
 	default:
-		return newDecodeError(itable.Data, "cannot store inline table in Go type %s", v.Kind())
+		return unstable.NewParserError(d.p.Raw(itable.Raw), "cannot store inline table in Go type %s", v.Kind())
 	}
 
 	it := itable.Children()
@@ -762,7 +772,7 @@ func (d *decoder) unmarshalInlineTable(itable *ast.Node, v reflect.Value) error
 	return nil
 }
 
-func (d *decoder) unmarshalDateTime(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalDateTime(value *unstable.Node, v reflect.Value) error {
 	dt, err := parseDateTime(value.Data)
 	if err != nil {
 		return err
@@ -772,7 +782,7 @@ func (d *decoder) unmarshalDateTime(value *ast.Node, v reflect.Value) error {
 	return nil
 }
 
-func (d *decoder) unmarshalLocalDate(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalLocalDate(value *unstable.Node, v reflect.Value) error {
 	ld, err := parseLocalDate(value.Data)
 	if err != nil {
 		return err
@@ -789,28 +799,28 @@ func (d *decoder) unmarshalLocalDate(value *ast.Node, v reflect.Value) error {
 	return nil
 }
 
-func (d *decoder) unmarshalLocalTime(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalLocalTime(value *unstable.Node, v reflect.Value) error {
 	lt, rest, err := parseLocalTime(value.Data)
 	if err != nil {
 		return err
 	}
 
 	if len(rest) > 0 {
-		return newDecodeError(rest, "extra characters at the end of a local time")
+		return unstable.NewParserError(rest, "extra characters at the end of a local time")
 	}
 
 	v.Set(reflect.ValueOf(lt))
 	return nil
 }
 
-func (d *decoder) unmarshalLocalDateTime(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalLocalDateTime(value *unstable.Node, v reflect.Value) error {
 	ldt, rest, err := parseLocalDateTime(value.Data)
 	if err != nil {
 		return err
 	}
 
 	if len(rest) > 0 {
-		return newDecodeError(rest, "extra characters at the end of a local date time")
+		return unstable.NewParserError(rest, "extra characters at the end of a local date time")
 	}
 
 	if v.Type() == timeType {
@@ -825,7 +835,7 @@ func (d *decoder) unmarshalLocalDateTime(value *ast.Node, v reflect.Value) error
 	return nil
 }
 
-func (d *decoder) unmarshalBool(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalBool(value *unstable.Node, v reflect.Value) error {
 	b := value.Data[0] == 't'
 
 	switch v.Kind() {
@@ -834,13 +844,13 @@ func (d *decoder) unmarshalBool(value *ast.Node, v reflect.Value) error {
 	case reflect.Interface:
 		v.Set(reflect.ValueOf(b))
 	default:
-		return newDecodeError(value.Data, "cannot assign boolean to a %t", b)
+		return unstable.NewParserError(value.Data, "cannot assign boolean to a %t", b)
 	}
 
 	return nil
 }
 
-func (d *decoder) unmarshalFloat(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalFloat(value *unstable.Node, v reflect.Value) error {
 	f, err := parseFloat(value.Data)
 	if err != nil {
 		return err
@@ -851,23 +861,43 @@ func (d *decoder) unmarshalFloat(value *ast.Node, v reflect.Value) error {
 		v.SetFloat(f)
 	case reflect.Float32:
 		if f > math.MaxFloat32 {
-			return newDecodeError(value.Data, "number %f does not fit in a float32", f)
+			return unstable.NewParserError(value.Data, "number %f does not fit in a float32", f)
 		}
 		v.SetFloat(f)
 	case reflect.Interface:
 		v.Set(reflect.ValueOf(f))
 	default:
-		return newDecodeError(value.Data, "float cannot be assigned to %s", v.Kind())
+		return unstable.NewParserError(value.Data, "float cannot be assigned to %s", v.Kind())
 	}
 
 	return nil
 }
 
-func (d *decoder) unmarshalInteger(value *ast.Node, v reflect.Value) error {
+const (
-	const (
+	maxInt = int64(^uint(0) >> 1)
-		maxInt = int64(^uint(0) >> 1)
+	minInt = -maxInt - 1
-		minInt = -maxInt - 1
+)
-	)
+
+// Maximum value of uint for decoding. Currently the decoder parses the integer
+// into an int64. As a result, on architectures where uint is 64 bits, the
+// effective maximum uint we can decode is the maximum of int64. On
+// architectures where uint is 32 bits, the maximum value we can decode is
+// lower: the maximum of uint32. I didn't find a way to figure out this value at
+// compile time, so it is computed during initialization.
+var maxUint int64 = math.MaxInt64
+
+func init() {
+	m := uint64(^uint(0))
+	if m < uint64(maxUint) {
+		maxUint = int64(m)
+	}
+}
+
+func (d *decoder) unmarshalInteger(value *unstable.Node, v reflect.Value) error {
+	kind := v.Kind()
+	if kind == reflect.Float32 || kind == reflect.Float64 {
+		return d.unmarshalFloat(value, v)
+	}
 
 	i, err := parseInteger(value.Data)
 	if err != nil {
@@ -876,7 +906,7 @@ func (d *decoder) unmarshalInteger(value *ast.Node, v reflect.Value) error {
 
 	var r reflect.Value
 
-	switch v.Kind() {
+	switch kind {
 	case reflect.Int64:
 		v.SetInt(i)
 		return nil
@@ -929,7 +959,7 @@ func (d *decoder) unmarshalInteger(value *ast.Node, v reflect.Value) error {
 
 		r = reflect.ValueOf(uint8(i))
 	case reflect.Uint:
-		if i < 0 {
+		if i < 0 || i > maxUint {
 			return fmt.Errorf("toml: negative number %d does not fit in an uint", i)
 		}
 
@@ -937,7 +967,7 @@ func (d *decoder) unmarshalInteger(value *ast.Node, v reflect.Value) error {
 	case reflect.Interface:
 		r = reflect.ValueOf(i)
 	default:
-		return d.typeMismatchError("integer", v.Type())
+		return unstable.NewParserError(d.p.Raw(value.Raw), d.typeMismatchString("integer", v.Type()))
 	}
 
 	if !r.Type().AssignableTo(v.Type()) {
@@ -949,20 +979,20 @@ func (d *decoder) unmarshalInteger(value *ast.Node, v reflect.Value) error {
 	return nil
 }
 
-func (d *decoder) unmarshalString(value *ast.Node, v reflect.Value) error {
+func (d *decoder) unmarshalString(value *unstable.Node, v reflect.Value) error {
 	switch v.Kind() {
 	case reflect.String:
 		v.SetString(string(value.Data))
 	case reflect.Interface:
 		v.Set(reflect.ValueOf(string(value.Data)))
 	default:
-		return newDecodeError(d.p.Raw(value.Raw), "cannot store TOML string into a Go %s", v.Kind())
+		return unstable.NewParserError(d.p.Raw(value.Raw), d.typeMismatchString("string", v.Type()))
 	}
 
 	return nil
 }
 
-func (d *decoder) handleKeyValue(expr *ast.Node, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleKeyValue(expr *unstable.Node, v reflect.Value) (reflect.Value, error) {
 	d.strict.EnterKeyValue(expr)
 
 	v, err := d.handleKeyValueInner(expr.Key(), expr.Value(), v)
@@ -976,7 +1006,7 @@ func (d *decoder) handleKeyValue(expr *ast.Node, v reflect.Value) (reflect.Value
 	return v, err
 }
 
-func (d *decoder) handleKeyValueInner(key ast.Iterator, value *ast.Node, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) handleKeyValueInner(key unstable.Iterator, value *unstable.Node, v reflect.Value) (reflect.Value, error) {
 	if key.Next() {
 		// Still scoping the key
 		return d.handleKeyValuePart(key, value, v)
@@ -986,7 +1016,32 @@ func (d *decoder) handleKeyValueInner(key ast.Iterator, value *ast.Node, v refle
 	return reflect.Value{}, d.handleValue(value, v)
 }
 
-func (d *decoder) handleKeyValuePart(key ast.Iterator, value *ast.Node, v reflect.Value) (reflect.Value, error) {
+func (d *decoder) keyFromData(keyType reflect.Type, data []byte) (reflect.Value, error) {
+	switch {
+	case stringType.AssignableTo(keyType):
+		return reflect.ValueOf(string(data)), nil
+
+	case stringType.ConvertibleTo(keyType):
+		return reflect.ValueOf(string(data)).Convert(keyType), nil
+
+	case keyType.Implements(textUnmarshalerType):
+		mk := reflect.New(keyType.Elem())
+		if err := mk.Interface().(encoding.TextUnmarshaler).UnmarshalText(data); err != nil {
+			return reflect.Value{}, fmt.Errorf("toml: error unmarshalling key type %s from text: %w", stringType, err)
+		}
+		return mk, nil
+
+	case reflect.PtrTo(keyType).Implements(textUnmarshalerType):
+		mk := reflect.New(keyType)
+		if err := mk.Interface().(encoding.TextUnmarshaler).UnmarshalText(data); err != nil {
+			return reflect.Value{}, fmt.Errorf("toml: error unmarshalling key type %s from text: %w", stringType, err)
+		}
+		return mk.Elem(), nil
+	}
+	return reflect.Value{}, fmt.Errorf("toml: cannot convert map key of type %s to expected type %s", stringType, keyType)
+}
+
+func (d *decoder) handleKeyValuePart(key unstable.Iterator, value *unstable.Node, v reflect.Value) (reflect.Value, error) {
 	// contains the replacement for v
 	var rv reflect.Value
 
@@ -996,16 +1051,9 @@ func (d *decoder) handleKeyValuePart(key ast.Iterator, value *ast.Node, v reflec
 	case reflect.Map:
 		vt := v.Type()
 
-		mk := reflect.ValueOf(string(key.Node().Data))
+		mk, err := d.keyFromData(vt.Key(), key.Node().Data)
-		mkt := stringType
+		if err != nil {
-
+			return reflect.Value{}, err
-		keyType := vt.Key()
-		if !mkt.AssignableTo(keyType) {
-			if !mkt.ConvertibleTo(keyType) {
-				return reflect.Value{}, fmt.Errorf("toml: cannot convert map key of type %s to expected type %s", mkt, keyType)
-			}
-
-			mk = mk.Convert(keyType)
 		}
 
 		// If the map does not exist, create it.
@@ -1016,15 +1064,9 @@ func (d *decoder) handleKeyValuePart(key ast.Iterator, value *ast.Node, v reflec
 
 		mv := v.MapIndex(mk)
 		set := false
-		if !mv.IsValid() {
+		if !mv.IsValid() || key.IsLast() {
 			set = true
 			mv = reflect.New(v.Type().Elem()).Elem()
-		} else {
-			if key.IsLast() {
-				var x interface{}
-				mv = reflect.ValueOf(&x).Elem()
-				set = true
-			}
 		}
 
 		nv, err := d.handleKeyValueInner(key, value, mv)
@@ -1053,7 +1095,20 @@ func (d *decoder) handleKeyValuePart(key ast.Iterator, value *ast.Node, v reflec
 		d.errorContext.Struct = t
 		d.errorContext.Field = path
 
-		f := v.FieldByIndex(path)
+		f := fieldByIndex(v, path)
+
+		if !f.CanSet() {
+			// If the field is not settable, need to take a slower path and make a copy of
+			// the struct itself to a new location.
+			nvp := reflect.New(v.Type())
+			nvp.Elem().Set(v)
+			v = nvp.Elem()
+			_, err := d.handleKeyValuePart(key, value, v)
+			if err != nil {
+				return reflect.Value{}, err
+			}
+			return nvp.Elem(), nil
+		}
 		x, err := d.handleKeyValueInner(key, value, f)
 		if err != nil {
 			return reflect.Value{}, err
@@ -1117,6 +1172,21 @@ func initAndDereferencePointer(v reflect.Value) reflect.Value {
 	return elem
 }
 
+// Same as reflect.Value.FieldByIndex, but creates pointers if needed.
+func fieldByIndex(v reflect.Value, path []int) reflect.Value {
+	for _, x := range path {
+		v = v.Field(x)
+
+		if v.Kind() == reflect.Ptr {
+			if v.IsNil() {
+				v.Set(reflect.New(v.Type().Elem()))
+			}
+			v = v.Elem()
+		}
+	}
+	return v
+}
+
 type fieldPathsMap = map[string][]int
 
 var globalFieldPathsCache atomic.Value // map[danger.TypeID]fieldPathsMap
@@ -1164,11 +1234,6 @@ func forEachField(t reflect.Type, path []int, do func(name string, path []int))
 		fieldPath := append(path, i)
 		fieldPath = fieldPath[:len(fieldPath):len(fieldPath)]
 
-		if f.Anonymous {
-			forEachField(f.Type, fieldPath, do)
-			continue
-		}
-
 		name := f.Tag.Get("toml")
 		if name == "-" {
 			continue
@@ -1177,6 +1242,19 @@ func forEachField(t reflect.Type, path []int, do func(name string, path []int))
 		if i := strings.IndexByte(name, ','); i >= 0 {
 			name = name[:i]
 		}
+
+		if f.Anonymous && name == "" {
+			t2 := f.Type
+			if t2.Kind() == reflect.Ptr {
+				t2 = t2.Elem()
+			}
+
+			if t2.Kind() == reflect.Struct {
+				forEachField(t2, fieldPath, do)
+			}
+			continue
+		}
+
 		if name == "" {
 			name = f.Name
 		}

unmarshaler_test.go

@@ -16,7 +16,28 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-func ExampleDecoder_SetStrict() {
+type unmarshalTextKey struct {
+	A string
+	B string
+}
+
+func (k *unmarshalTextKey) UnmarshalText(text []byte) error {
+	parts := strings.Split(string(text), "-")
+	if len(parts) != 2 {
+		return fmt.Errorf("invalid text key: %s", text)
+	}
+	k.A = parts[0]
+	k.B = parts[1]
+	return nil
+}
+
+type unmarshalBadTextKey struct{}
+
+func (k *unmarshalBadTextKey) UnmarshalText(text []byte) error {
+	return fmt.Errorf("error")
+}
+
+func ExampleDecoder_DisallowUnknownFields() {
 	type S struct {
 		Key1 string
 		Key3 string
@@ -28,7 +49,7 @@ key3 = "value3"
 `
 	r := strings.NewReader(doc)
 	d := toml.NewDecoder(r)
-	d.SetStrict(true)
+	d.DisallowUnknownFields()
 	s := S{}
 	err := d.Decode(&s)
 
@@ -69,7 +90,6 @@ func ExampleUnmarshal() {
 	fmt.Println("version:", cfg.Version)
 	fmt.Println("name:", cfg.Name)
 	fmt.Println("tags:", cfg.Tags)
-
 	// Output:
 	// version: 2
 	// name: go-toml
@@ -220,6 +240,11 @@ func TestUnmarshal_Floats(t *testing.T) {
 			input:    `0E0`,
 			expected: 0.0,
 		},
+		{
+			desc:     "float zero without decimals",
+			input:    `0`,
+			expected: 0.0,
+		},
 		{
 			desc:     "float fractional with exponent",
 			input:    `6.626e-34`,
@@ -279,6 +304,11 @@ func TestUnmarshal_Floats(t *testing.T) {
 			input: `1.0_e2`,
 			err:   true,
 		},
+		{
+			desc:  "leading zero in positive float",
+			input: `+0_0.0`,
+			err:   true,
+		},
 	}
 
 	type doc struct {
@@ -310,6 +340,7 @@ func TestUnmarshal(t *testing.T) {
 		target   interface{}
 		expected interface{}
 		err      bool
+		assert   func(t *testing.T, test test)
 	}
 	examples := []struct {
 		skip  bool
@@ -345,6 +376,96 @@ func TestUnmarshal(t *testing.T) {
 				}
 			},
 		},
+		{
+			desc:  "kv text key",
+			input: `a-1 = "foo"`,
+			gen: func() test {
+				type doc = map[unmarshalTextKey]string
+
+				return test{
+					target:   &doc{},
+					expected: &doc{{A: "a", B: "1"}: "foo"},
+				}
+			},
+		},
+		{
+			desc: "table text key",
+			input: `["a-1"]
+foo = "bar"`,
+			gen: func() test {
+				type doc = map[unmarshalTextKey]map[string]string
+
+				return test{
+					target:   &doc{},
+					expected: &doc{{A: "a", B: "1"}: map[string]string{"foo": "bar"}},
+				}
+			},
+		},
+		{
+			desc:  "kv ptr text key",
+			input: `a-1 = "foo"`,
+			gen: func() test {
+				type doc = map[*unmarshalTextKey]string
+
+				return test{
+					target:   &doc{},
+					expected: &doc{{A: "a", B: "1"}: "foo"},
+					assert: func(t *testing.T, test test) {
+						// Despite the documentation:
+						//     Pointer variable equality is determined based on the equality of the
+						// 		 referenced values (as opposed to the memory addresses).
+						// assert.Equal does not work properly with maps with pointer keys
+						// https://github.com/stretchr/testify/issues/1143
+						expected := make(map[unmarshalTextKey]string)
+						for k, v := range *(test.expected.(*doc)) {
+							expected[*k] = v
+						}
+						got := make(map[unmarshalTextKey]string)
+						for k, v := range *(test.target.(*doc)) {
+							got[*k] = v
+						}
+						assert.Equal(t, expected, got)
+					},
+				}
+			},
+		},
+		{
+			desc:  "kv bad text key",
+			input: `a-1 = "foo"`,
+			gen: func() test {
+				type doc = map[unmarshalBadTextKey]string
+
+				return test{
+					target: &doc{},
+					err:    true,
+				}
+			},
+		},
+		{
+			desc:  "kv bad ptr text key",
+			input: `a-1 = "foo"`,
+			gen: func() test {
+				type doc = map[*unmarshalBadTextKey]string
+
+				return test{
+					target: &doc{},
+					err:    true,
+				}
+			},
+		},
+		{
+			desc: "table bad text key",
+			input: `["a-1"]
+foo = "bar"`,
+			gen: func() test {
+				type doc = map[unmarshalBadTextKey]map[string]string
+
+				return test{
+					target: &doc{},
+					err:    true,
+				}
+			},
+		},
 		{
 			desc:  "time.time with negative zone",
 			input: `a = 1979-05-27T00:32:00-07:00 `, // space intentional
@@ -540,6 +661,35 @@ func TestUnmarshal(t *testing.T) {
 				}
 			},
 		},
+		{
+			desc: "issue 739 - table redefinition",
+			input: `
+[foo.bar.baz]
+wibble = 'wobble'
+
+[foo]
+
+[foo.bar]
+huey = 'dewey'
+			`,
+			gen: func() test {
+				m := map[string]interface{}{}
+
+				return test{
+					target: &m,
+					expected: &map[string]interface{}{
+						`foo`: map[string]interface{}{
+							"bar": map[string]interface{}{
+								"huey": "dewey",
+								"baz": map[string]interface{}{
+									"wibble": "wobble",
+								},
+							},
+						},
+					},
+				}
+			},
+		},
 		{
 			desc: "multiline basic string",
 			input: `A = """\
@@ -937,7 +1087,7 @@ B = "data"`,
 								"Name": "Hammer",
 								"Sku":  int64(738594937),
 							},
-							map[string]interface{}(nil),
+							map[string]interface{}{},
 							map[string]interface{}{
 								"Name":  "Nail",
 								"Sku":   int64(284758393),
@@ -1132,6 +1282,64 @@ B = "data"`,
 				}
 			},
 		},
+		{
+			desc: "array table into maps with pointer on last key",
+			input: `[[foo]]
+			bar = "hello"`,
+			gen: func() test {
+				type doc struct {
+					Foo **[]interface{}
+				}
+				x := &[]interface{}{
+					map[string]interface{}{
+						"bar": "hello",
+					},
+				}
+				return test{
+					target: &doc{},
+					expected: &doc{
+						Foo: &x,
+					},
+				}
+			},
+		},
+		{
+			desc: "array table into maps with pointer on intermediate key",
+			input: `[[foo.foo2]]
+			bar = "hello"`,
+			gen: func() test {
+				type doc struct {
+					Foo **map[string]interface{}
+				}
+				x := &map[string]interface{}{
+					"foo2": []interface{}{
+						map[string]interface{}{
+							"bar": "hello",
+						},
+					},
+				}
+				return test{
+					target: &doc{},
+					expected: &doc{
+						Foo: &x,
+					},
+				}
+			},
+		},
+		{
+			desc: "array table into maps with pointer on last key with invalid leaf type",
+			input: `[[foo]]
+			bar = "hello"`,
+			gen: func() test {
+				type doc struct {
+					Foo **[]map[string]int
+				}
+				return test{
+					target: &doc{},
+					err:    true,
+				}
+			},
+		},
 		{
 			desc:  "unexported struct fields are ignored",
 			input: `foo = "bar"`,
@@ -1471,7 +1679,7 @@ B = "data"`,
 					target: &map[string]interface{}{},
 					expected: &map[string]interface{}{
 						"products": []interface{}{
-							map[string]interface{}(nil),
+							map[string]interface{}{},
 						},
 					},
 				}
@@ -1487,6 +1695,16 @@ B = "data"`,
 				}
 			},
 		},
+		{
+			desc:  "empty map into map with invalid key type",
+			input: ``,
+			gen: func() test {
+				return test{
+					target:   &map[int]string{},
+					expected: &map[int]string{},
+				}
+			},
+		},
 		{
 			desc:  "into map with convertible key type",
 			input: `A = "hello"`,
@@ -1701,6 +1919,28 @@ B = "data"`,
 				}
 			},
 		},
+		{
+			desc:  "kv that points to a slice",
+			input: "a.b.c = 'foo'",
+			gen: func() test {
+				doc := map[string][]string{}
+				return test{
+					target: &doc,
+					err:    true,
+				}
+			},
+		},
+		{
+			desc:  "kv that points to a pointer to a slice",
+			input: "a.b.c = 'foo'",
+			gen: func() test {
+				doc := map[string]*[]string{}
+				return test{
+					target: &doc,
+					err:    true,
+				}
+			},
+		},
 	}
 
 	for _, e := range examples {
@@ -1721,7 +1961,11 @@ B = "data"`,
 				require.Error(t, err)
 			} else {
 				require.NoError(t, err)
-				assert.Equal(t, test.expected, test.target)
+				if test.assert != nil {
+					test.assert(t, test)
+				} else {
+					assert.Equal(t, test.expected, test.target)
+				}
 			}
 		})
 	}
@@ -1804,6 +2048,68 @@ func TestUnmarshalErrors(t *testing.T) {
 	require.Equal(t, "toml: cannot decode TOML integer into struct field toml_test.mystruct.Bar of type string", err.Error())
 }
 
+func TestUnmarshalStringInvalidStructField(t *testing.T) {
+	type Server struct {
+		Path string
+		Port int
+	}
+
+	type Cfg struct {
+		Server Server
+	}
+
+	var cfg Cfg
+
+	data := `[server]
+path = "/my/path"
+port = "bad"
+`
+
+	file := strings.NewReader(data)
+	err := toml.NewDecoder(file).Decode(&cfg)
+	require.Error(t, err)
+
+	x := err.(*toml.DecodeError)
+	require.Equal(t, "toml: cannot decode TOML string into struct field toml_test.Server.Port of type int", x.Error())
+	expected := `1| [server]
+2| path = "/my/path"
+3| port = "bad"
+ |        ~~~~~ cannot decode TOML string into struct field toml_test.Server.Port of type int`
+
+	require.Equal(t, expected, x.String())
+}
+
+func TestUnmarshalIntegerInvalidStructField(t *testing.T) {
+	type Server struct {
+		Path string
+		Port int
+	}
+
+	type Cfg struct {
+		Server Server
+	}
+
+	var cfg Cfg
+
+	data := `[server]
+path = 100
+port = 50
+`
+
+	file := strings.NewReader(data)
+	err := toml.NewDecoder(file).Decode(&cfg)
+	require.Error(t, err)
+
+	x := err.(*toml.DecodeError)
+	require.Equal(t, "toml: cannot decode TOML integer into struct field toml_test.Server.Path of type string", x.Error())
+	expected := `1| [server]
+2| path = 100
+ |        ~~~ cannot decode TOML integer into struct field toml_test.Server.Path of type string
+3| port = 50`
+
+	require.Equal(t, expected, x.String())
+}
+
 func TestUnmarshalInvalidTarget(t *testing.T) {
 	x := "foo"
 	err := toml.Unmarshal([]byte{}, x)
@@ -1842,8 +2148,7 @@ key2 = "missing2"
 key3 = "missing3"
 key4 = "value4"
 `,
-			expected: `
+			expected: `2| key1 = "value1"
-2| key1 = "value1"
 3| key2 = "missing2"
  | ~~~~ missing field
 4| key3 = "missing3"
@@ -1853,8 +2158,7 @@ key4 = "value4"
 3| key2 = "missing2"
 4| key3 = "missing3"
  | ~~~~ missing field
-5| key4 = "value4"
+5| key4 = "value4"`,
-`,
 			target: &struct {
 				Key1 string
 				Key4 string
@@ -1863,10 +2167,8 @@ key4 = "value4"
 		{
 			desc:  "multi-part key",
 			input: `a.short.key="foo"`,
-			expected: `
+			expected: `1| a.short.key="foo"
-1| a.short.key="foo"
+ | ~~~~~~~~~~~ missing field`,
- | ~~~~~~~~~~~ missing field
-`,
 		},
 		{
 			desc: "missing table",
@@ -1874,24 +2176,19 @@ key4 = "value4"
 [foo]
 bar = 42
 `,
-			expected: `
+			expected: `2| [foo]
-2| [foo]
  |  ~~~ missing table
-3| bar = 42
+3| bar = 42`,
-`,
 		},
 
 		{
 			desc: "missing array table",
 			input: `
 [[foo]]
-bar = 42
+bar = 42`,
-`,
+			expected: `2| [[foo]]
-			expected: `
-2| [[foo]]
  |   ~~~ missing table
-3| bar = 42
+3| bar = 42`,
-`,
 		},
 	}
 
@@ -1901,7 +2198,7 @@ bar = 42
 			t.Run("strict", func(t *testing.T) {
 				r := strings.NewReader(e.input)
 				d := toml.NewDecoder(r)
-				d.SetStrict(true)
+				d.DisallowUnknownFields()
 				x := e.target
 				if x == nil {
 					x = &struct{}{}
@@ -1910,7 +2207,7 @@ bar = 42
 
 				var tsm *toml.StrictMissingError
 				if errors.As(err, &tsm) {
-					equalStringsIgnoreNewlines(t, e.expected, tsm.String())
+					assert.Equal(t, e.expected, tsm.String())
 				} else {
 					t.Fatalf("err should have been a *toml.StrictMissingError, but got %s (%T)", err, err)
 				}
@@ -1919,7 +2216,6 @@ bar = 42
 			t.Run("default", func(t *testing.T) {
 				r := strings.NewReader(e.input)
 				d := toml.NewDecoder(r)
-				d.SetStrict(false)
 				x := e.target
 				if x == nil {
 					x = &struct{}{}
@@ -2347,6 +2643,164 @@ func TestIssue714(t *testing.T) {
 	require.Error(t, err)
 }
 
+func TestIssue772(t *testing.T) {
+	type FileHandling struct {
+		FilePattern string `toml:"pattern"`
+	}
+
+	type Config struct {
+		FileHandling `toml:"filehandling"`
+	}
+
+	var defaultConfigFile = []byte(`
+		[filehandling]
+		pattern = "reach-masterdev-"`)
+
+	config := Config{}
+	err := toml.Unmarshal(defaultConfigFile, &config)
+	require.NoError(t, err)
+	require.Equal(t, "reach-masterdev-", config.FileHandling.FilePattern)
+}
+
+func TestIssue774(t *testing.T) {
+	type ScpData struct {
+		Host string `json:"host"`
+	}
+
+	type GenConfig struct {
+		SCP []ScpData `toml:"scp" comment:"Array of Secure Copy Configurations"`
+	}
+
+	c := &GenConfig{}
+	c.SCP = []ScpData{{Host: "main.domain.com"}}
+
+	b, err := toml.Marshal(c)
+	require.NoError(t, err)
+
+	expected := `# Array of Secure Copy Configurations
+[[scp]]
+Host = 'main.domain.com'
+`
+
+	require.Equal(t, expected, string(b))
+}
+
+func TestIssue799(t *testing.T) {
+	const testTOML = `
+# notice the double brackets
+[[test]]
+answer = 42
+`
+
+	var s struct {
+		// should be []map[string]int
+		Test map[string]int `toml:"test"`
+	}
+
+	err := toml.Unmarshal([]byte(testTOML), &s)
+	require.Error(t, err)
+}
+
+func TestIssue807(t *testing.T) {
+	type A struct {
+		Name string `toml:"name"`
+	}
+
+	type M struct {
+		*A
+	}
+
+	var m M
+	err := toml.Unmarshal([]byte(`name = 'foo'`), &m)
+	require.NoError(t, err)
+	require.Equal(t, "foo", m.Name)
+}
+
+func TestIssue850(t *testing.T) {
+	data := make(map[string]string)
+	err := toml.Unmarshal([]byte("foo = {}"), &data)
+	require.Error(t, err)
+}
+
+func TestIssue851(t *testing.T) {
+	type Target struct {
+		Params map[string]string `toml:"params"`
+	}
+
+	content := "params = {a=\"1\",b=\"2\"}"
+	var target Target
+	err := toml.Unmarshal([]byte(content), &target)
+	require.NoError(t, err)
+	require.Equal(t, map[string]string{"a": "1", "b": "2"}, target.Params)
+	err = toml.Unmarshal([]byte(content), &target)
+	require.NoError(t, err)
+	require.Equal(t, map[string]string{"a": "1", "b": "2"}, target.Params)
+}
+
+func TestIssue866(t *testing.T) {
+	type Pipeline struct {
+		Mapping map[string]struct {
+			Req [][]string `toml:"req"`
+			Res [][]string `toml:"res"`
+		} `toml:"mapping"`
+	}
+
+	type Pipelines struct {
+		PipelineMapping map[string]*Pipeline `toml:"pipelines"`
+	}
+
+	var badToml = `
+[pipelines.register]
+mapping.inst.req = [
+    ["param1", "value1"],
+]
+mapping.inst.res = [
+    ["param2", "value2"],
+]
+`
+
+	pipelines := new(Pipelines)
+	if err := toml.NewDecoder(bytes.NewBufferString(badToml)).DisallowUnknownFields().Decode(pipelines); err != nil {
+		t.Fatal(err)
+	}
+	if pipelines.PipelineMapping["register"].Mapping["inst"].Req[0][0] != "param1" {
+		t.Fatal("unmarshal failed with mismatch value")
+	}
+
+	var goodTooToml = `
+[pipelines.register]
+mapping.inst.req = [
+    ["param1", "value1"],
+]
+`
+
+	pipelines = new(Pipelines)
+	if err := toml.NewDecoder(bytes.NewBufferString(goodTooToml)).DisallowUnknownFields().Decode(pipelines); err != nil {
+		t.Fatal(err)
+	}
+	if pipelines.PipelineMapping["register"].Mapping["inst"].Req[0][0] != "param1" {
+		t.Fatal("unmarshal failed with mismatch value")
+	}
+
+	var goodToml = `
+[pipelines.register.mapping.inst]
+req = [
+    ["param1", "value1"],
+]
+res = [
+    ["param2", "value2"],
+]
+`
+
+	pipelines = new(Pipelines)
+	if err := toml.NewDecoder(bytes.NewBufferString(goodToml)).DisallowUnknownFields().Decode(pipelines); err != nil {
+		t.Fatal(err)
+	}
+	if pipelines.PipelineMapping["register"].Mapping["inst"].Req[0][0] != "param1" {
+		t.Fatal("unmarshal failed with mismatch value")
+	}
+}
+
 func TestUnmarshalDecodeErrors(t *testing.T) {
 	examples := []struct {
 		desc string
@@ -2623,7 +3077,7 @@ world'`,
 			data: "a = \"aaaa\xE2\x80\x00\"",
 		},
 		{
-			desc: "invalid 4rd byte of 4-byte utf8 character in string with no escape sequence",
+			desc: "invalid 4th byte of 4-byte utf8 character in string with no escape sequence",
 			data: "a = \"aaaa\xF2\x81\x81\x00\"",
 		},
 		{
@@ -2639,7 +3093,7 @@ world'`,
 			data: "a = 'aaaa\xE2\x80\x00'",
 		},
 		{
-			desc: "invalid 4rd byte of 4-byte utf8 character in literal string",
+			desc: "invalid 4th byte of 4-byte utf8 character in literal string",
 			data: "a = 'aaaa\xF2\x81\x81\x00'",
 		},
 		{
@@ -2798,6 +3252,36 @@ world'`,
 	}
 }
 
+func TestOmitEmpty(t *testing.T) {
+	type inner struct {
+		private string
+		Skip    string `toml:"-"`
+		V       string
+	}
+
+	type elem struct {
+		Foo   string `toml:",omitempty"`
+		Bar   string `toml:",omitempty"`
+		Inner inner  `toml:",omitempty"`
+	}
+
+	type doc struct {
+		X []elem `toml:",inline"`
+	}
+
+	d := doc{X: []elem{elem{
+		Foo: "test",
+		Inner: inner{
+			V: "alue",
+		},
+	}}}
+
+	b, err := toml.Marshal(d)
+	require.NoError(t, err)
+
+	require.Equal(t, "X = [{Foo = 'test', Inner = {V = 'alue'}}]\n", string(b))
+}
+
 func TestUnmarshalTags(t *testing.T) {
 	type doc struct {
 		Dash   string `toml:"-,"`
@@ -3139,3 +3623,60 @@ func TestUnmarshal_RecursiveTableArray(t *testing.T) {
 		})
 	}
 }
+
+func TestUnmarshalEmbedNonString(t *testing.T) {
+	type Foo []byte
+	type doc struct {
+		Foo
+	}
+
+	d := doc{}
+
+	err := toml.Unmarshal([]byte(`foo = 'bar'`), &d)
+	require.NoError(t, err)
+	require.Nil(t, d.Foo)
+}
+
+func TestUnmarshal_Nil(t *testing.T) {
+	type Foo struct {
+		Foo *Foo `toml:"foo,omitempty"`
+		Bar *Foo `toml:"bar,omitempty"`
+	}
+
+	examples := []struct {
+		desc     string
+		input    string
+		expected string
+		err      bool
+	}{
+		{
+			desc:     "empty",
+			input:    ``,
+			expected: ``,
+		},
+		{
+			desc: "simplest",
+			input: `
+            [foo]
+            [foo.foo]
+            `,
+			expected: "[foo]\n[foo.foo]\n",
+		},
+	}
+
+	for _, ex := range examples {
+		e := ex
+		t.Run(e.desc, func(t *testing.T) {
+			foo := Foo{}
+			err := toml.Unmarshal([]byte(e.input), &foo)
+			if e.err {
+				require.Error(t, err)
+			} else {
+				require.NoError(t, err)
+				j, err := toml.Marshal(foo)
+				require.NoError(t, err)
+				assert.Equal(t, e.expected, string(j))
+			}
+		})
+	}
+}

unstable/ast.go

@@ -1,4 +1,4 @@
-package ast
+package unstable
 
 import (
 	"fmt"
@@ -7,14 +7,17 @@ import (
 	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
-// Iterator starts uninitialized, you need to call Next() first.
+// Iterator over a sequence of nodes.
+//
+// Starts uninitialized, you need to call Next() first.
 //
 // For example:
 //
-// it := n.Children()
+//	it := n.Children()
-// for it.Next() {
+//	for it.Next() {
-// 		it.Node()
+//		n := it.Node()
-// }
+//		// do something with n
+//	}
 type Iterator struct {
 	started bool
 	node    *Node
@@ -32,42 +35,31 @@ func (c *Iterator) Next() bool {
 }
 
 // IsLast returns true if the current node of the iterator is the last
-// one.  Subsequent call to Next() will return false.
+// one.  Subsequent calls 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.
+// Node returns a pointer to the node pointed at by the iterator.
 func (c *Iterator) Node() *Node {
 	return c.node
 }
 
-// Root contains a full AST.
+// Node in a TOML expression AST.
 //
-// It is immutable once constructed with Builder.
+// Depending on Kind, its sequence of children should be interpreted
-type Root struct {
+// differently.
-	nodes []Node
+//
-}
+//   - Array have one child per element in the array.
-
+//   - InlineTable have one child per key-value in the table (each of kind
-// Iterator over the top level nodes.
+//     InlineTable).
-func (r *Root) Iterator() Iterator {
+//   - KeyValue have at least two children. The first one is the value. The rest
-	it := Iterator{}
+//     make a potentially dotted key.
-	if len(r.nodes) > 0 {
+//   - Table and ArrayTable's children represent a dotted key (same as
-		it.node = &r.nodes[0]
+//     KeyValue, but without the first node being the value).
-	}
+//
-	return it
+// When relevant, Raw describes the range of bytes this node is referring to in
-}
+// the input document. Use Parser.Raw() to retrieve the actual bytes.
-
-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).
 type Node struct {
 	Kind Kind
 	Raw  Range  // Raw bytes from the input.
@@ -80,13 +72,13 @@ type Node struct {
 	child int // 0 if no child
 }
 
+// Range of bytes in the document.
 type Range struct {
 	Offset uint32
 	Length uint32
 }
 
-// Next returns a copy of the next node, or an invalid Node if there
+// Next returns a pointer to the next node, or nil if there is no next node.
-// is no next node.
 func (n *Node) Next() *Node {
 	if n.next == 0 {
 		return nil
@@ -96,9 +88,9 @@ func (n *Node) Next() *Node {
 	return (*Node)(danger.Stride(ptr, size, n.next))
 }
 
-// Child returns a copy of the first child node of this node. Other
+// Child returns a pointer to the first child node of this node. Other children
-// children can be accessed calling Next on the first child.  Returns
+// can be accessed calling Next on the first child.  Returns an nil if this Node
-// an invalid Node if there is none.
+// has no child.
 func (n *Node) Child() *Node {
 	if n.child == 0 {
 		return nil
@@ -113,9 +105,9 @@ func (n *Node) Valid() bool {
 	return n != nil
 }
 
-// Key returns the child nodes making the Key on a supported
+// Key returns the children nodes making the Key on a supported node. Panics
-// node. Panics otherwise.  They are guaranteed to be all be of the
+// otherwise.  They are guaranteed to be all be of the Kind Key. A simple key
-// Kind Key. A simple key would return just one element.
+// would return just one element.
 func (n *Node) Key() Iterator {
 	switch n.Kind {
 	case KeyValue:

unstable/benchmark_test.go

@@ -1,4 +1,4 @@
-package toml
+package unstable
 
 import (
 	"bytes"
@@ -55,7 +55,7 @@ func BenchmarkParseLiteralStringValid(b *testing.B) {
 
 	for name, input := range inputs {
 		b.Run(name, func(b *testing.B) {
-			p := parser{}
+			p := Parser{}
 			b.SetBytes(int64(len(input)))
 			b.ReportAllocs()
 			b.ResetTimer()

unstable/builder.go

@@ -0,0 +1,71 @@
+package unstable
+
+// root contains a full AST.
+//
+// It is immutable once constructed with Builder.
+type root struct {
+	nodes []Node
+}
+
+// Iterator over the top level nodes.
+func (r *root) Iterator() Iterator {
+	it := Iterator{}
+	if len(r.nodes) > 0 {
+		it.node = &r.nodes[0]
+	}
+	return it
+}
+
+func (r *root) at(idx reference) *Node {
+	return &r.nodes[idx]
+}
+
+type reference int
+
+const invalidReference reference = -1
+
+func (r reference) Valid() bool {
+	return r != invalidReference
+}
+
+type builder struct {
+	tree    root
+	lastIdx int
+}
+
+func (b *builder) Tree() *root {
+	return &b.tree
+}
+
+func (b *builder) NodeAt(ref reference) *Node {
+	return b.tree.at(ref)
+}
+
+func (b *builder) Reset() {
+	b.tree.nodes = b.tree.nodes[:0]
+	b.lastIdx = 0
+}
+
+func (b *builder) Push(n Node) reference {
+	b.lastIdx = len(b.tree.nodes)
+	b.tree.nodes = append(b.tree.nodes, n)
+	return reference(b.lastIdx)
+}
+
+func (b *builder) PushAndChain(n Node) reference {
+	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.lastIdx
+	}
+	b.lastIdx = newIdx
+	return reference(b.lastIdx)
+}
+
+func (b *builder) AttachChild(parent reference, child reference) {
+	b.tree.nodes[parent].child = int(child) - int(parent)
+}
+
+func (b *builder) Chain(from reference, to reference) {
+	b.tree.nodes[from].next = int(to) - int(from)
+}

unstable/doc.go

@@ -0,0 +1,3 @@
+// Package unstable provides APIs that do not meet the backward compatibility
+// guarantees yet.
+package unstable

unstable/kind.go

@@ -1,25 +1,26 @@
-package ast
+package unstable
 
 import "fmt"
 
+// Kind represents the type of TOML structure contained in a given Node.
 type Kind int
 
 const (
-	// meta
+	// Meta
 	Invalid Kind = iota
 	Comment
 	Key
 
-	// top level structures
+	// Top level structures
 	Table
 	ArrayTable
 	KeyValue
 
-	// containers values
+	// Containers values
 	Array
 	InlineTable
 
-	// values
+	// Values
 	String
 	Bool
 	Float
@@ -30,6 +31,7 @@ const (
 	DateTime
 )
 
+// String implementation of fmt.Stringer.
 func (k Kind) String() string {
 	switch k {
 	case Invalid:

unstable/parser.go

@@ -1,50 +1,108 @@
-package toml
+package unstable
 
 import (
 	"bytes"
+	"fmt"
 	"unicode"
 
-	"github.com/pelletier/go-toml/v2/internal/ast"
+	"github.com/pelletier/go-toml/v2/internal/characters"
 	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
-type parser struct {
+// ParserError describes an error relative to the content of the document.
-	builder ast.Builder
+//
-	ref     ast.Reference
+// It cannot outlive the instance of Parser it refers to, and may cause panics
+// if the parser is reset.
+type ParserError struct {
+	Highlight []byte
+	Message   string
+	Key       []string // optional
+}
+
+// Error is the implementation of the error interface.
+func (e *ParserError) Error() string {
+	return e.Message
+}
+
+// NewParserError is a convenience function to create a ParserError
+//
+// Warning: Highlight needs to be a subslice of Parser.data, so only slices
+// returned by Parser.Raw are valid candidates.
+func NewParserError(highlight []byte, format string, args ...interface{}) error {
+	return &ParserError{
+		Highlight: highlight,
+		Message:   fmt.Errorf(format, args...).Error(),
+	}
+}
+
+// Parser scans over a TOML-encoded document and generates an iterative AST.
+//
+// To prime the Parser, first reset it with the contents of a TOML document.
+// Then, process all top-level expressions sequentially. See Example.
+//
+// Don't forget to check Error() after you're done parsing.
+//
+// Each top-level expression needs to be fully processed before calling
+// NextExpression() again. Otherwise, calls to various Node methods may panic if
+// the parser has moved on the next expression.
+//
+// For performance reasons, go-toml doesn't make a copy of the input bytes to
+// the parser. Make sure to copy all the bytes you need to outlive the slice
+// given to the parser.
+type Parser struct {
 	data    []byte
+	builder builder
+	ref     reference
 	left    []byte
 	err     error
 	first   bool
+
+	KeepComments bool
 }
 
-func (p *parser) Range(b []byte) ast.Range {
+// Data returns the slice provided to the last call to Reset.
-	return ast.Range{
+func (p *Parser) Data() []byte {
+	return p.data
+}
+
+// Range returns a range description that corresponds to a given slice of the
+// input. If the argument is not a subslice of the parser input, this function
+// panics.
+func (p *Parser) Range(b []byte) Range {
+	return Range{
 		Offset: uint32(danger.SubsliceOffset(p.data, b)),
 		Length: uint32(len(b)),
 	}
 }
 
-func (p *parser) Raw(raw ast.Range) []byte {
+// Raw returns the slice corresponding to the bytes in the given range.
+func (p *Parser) Raw(raw Range) []byte {
 	return p.data[raw.Offset : raw.Offset+raw.Length]
 }
 
-func (p *parser) Reset(b []byte) {
+// Reset brings the parser to its initial state for a given input. It wipes an
+// reuses internal storage to reduce allocation.
+func (p *Parser) Reset(b []byte) {
 	p.builder.Reset()
-	p.ref = ast.InvalidReference
+	p.ref = invalidReference
 	p.data = b
 	p.left = b
 	p.err = nil
 	p.first = true
 }
 
-//nolint:cyclop
+// NextExpression parses the next top-level expression. If an expression was
-func (p *parser) NextExpression() bool {
+// successfully parsed, it returns true. If the parser is at the end of the
+// document or an error occurred, it returns false.
+//
+// Retrieve the parsed expression with Expression().
+func (p *Parser) NextExpression() bool {
 	if len(p.left) == 0 || p.err != nil {
 		return false
 	}
 
 	p.builder.Reset()
-	p.ref = ast.InvalidReference
+	p.ref = invalidReference
 
 	for {
 		if len(p.left) == 0 || p.err != nil {
@@ -73,15 +131,56 @@ func (p *parser) NextExpression() bool {
 	}
 }
 
-func (p *parser) Expression() *ast.Node {
+// Expression returns a pointer to the node representing the last successfully
+// parsed expression.
+func (p *Parser) Expression() *Node {
 	return p.builder.NodeAt(p.ref)
 }
 
-func (p *parser) Error() error {
+// Error returns any error that has occurred during parsing.
+func (p *Parser) Error() error {
 	return p.err
 }
 
-func (p *parser) parseNewline(b []byte) ([]byte, error) {
+// Position describes a position in the input.
+type Position struct {
+	// Number of bytes from the beginning of the input.
+	Offset int
+	// Line number, starting at 1.
+	Line int
+	// Column number, starting at 1.
+	Column int
+}
+
+// Shape describes the position of a range in the input.
+type Shape struct {
+	Start Position
+	End   Position
+}
+
+func (p *Parser) position(b []byte) Position {
+	offset := danger.SubsliceOffset(p.data, b)
+
+	lead := p.data[:offset]
+
+	return Position{
+		Offset: offset,
+		Line:   bytes.Count(lead, []byte{'\n'}) + 1,
+		Column: len(lead) - bytes.LastIndex(lead, []byte{'\n'}),
+	}
+}
+
+// Shape returns the shape of the given range in the input.  Will
+// panic if the range is not a subslice of the input.
+func (p *Parser) Shape(r Range) Shape {
+	raw := p.Raw(r)
+	return Shape{
+		Start: p.position(raw),
+		End:   p.position(raw[r.Length:]),
+	}
+}
+
+func (p *Parser) parseNewline(b []byte) ([]byte, error) {
 	if b[0] == '\n' {
 		return b[1:], nil
 	}
@@ -91,14 +190,27 @@ func (p *parser) parseNewline(b []byte) ([]byte, error) {
 		return rest, err
 	}
 
-	return nil, newDecodeError(b[0:1], "expected newline but got %#U", b[0])
+	return nil, NewParserError(b[0:1], "expected newline but got %#U", b[0])
 }
 
-func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseComment(b []byte) (reference, []byte, error) {
+	ref := invalidReference
+	data, rest, err := scanComment(b)
+	if p.KeepComments && err == nil {
+		ref = p.builder.Push(Node{
+			Kind: Comment,
+			Raw:  p.Range(data),
+			Data: data,
+		})
+	}
+	return ref, rest, err
+}
+
+func (p *Parser) parseExpression(b []byte) (reference, []byte, error) {
 	// expression =  ws [ comment ]
 	// expression =/ ws keyval ws [ comment ]
 	// expression =/ ws table ws [ comment ]
-	ref := ast.InvalidReference
+	ref := invalidReference
 
 	b = p.parseWhitespace(b)
 
@@ -107,7 +219,7 @@ func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
 	}
 
 	if b[0] == '#' {
-		_, rest, err := scanComment(b)
+		ref, rest, err := p.parseComment(b)
 		return ref, rest, err
 	}
 
@@ -129,14 +241,17 @@ func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
 	b = p.parseWhitespace(b)
 
 	if len(b) > 0 && b[0] == '#' {
-		_, rest, err := scanComment(b)
+		cref, rest, err := p.parseComment(b)
+		if cref != invalidReference {
+			p.builder.Chain(ref, cref)
+		}
 		return ref, rest, err
 	}
 
 	return ref, b, nil
 }
 
-func (p *parser) parseTable(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseTable(b []byte) (reference, []byte, error) {
 	// table = std-table / array-table
 	if len(b) > 1 && b[1] == '[' {
 		return p.parseArrayTable(b)
@@ -145,12 +260,12 @@ func (p *parser) parseTable(b []byte) (ast.Reference, []byte, error) {
 	return p.parseStdTable(b)
 }
 
-func (p *parser) parseArrayTable(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseArrayTable(b []byte) (reference, []byte, error) {
 	// array-table = array-table-open key array-table-close
 	// array-table-open  = %x5B.5B ws  ; [[ Double left square bracket
 	// array-table-close = ws %x5D.5D  ; ]] Double right square bracket
-	ref := p.builder.Push(ast.Node{
+	ref := p.builder.Push(Node{
-		Kind: ast.ArrayTable,
+		Kind: ArrayTable,
 	})
 
 	b = b[2:]
@@ -174,12 +289,12 @@ func (p *parser) parseArrayTable(b []byte) (ast.Reference, []byte, error) {
 	return ref, b, err
 }
 
-func (p *parser) parseStdTable(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseStdTable(b []byte) (reference, []byte, error) {
 	// std-table = std-table-open key std-table-close
 	// std-table-open  = %x5B ws     ; [ Left square bracket
 	// std-table-close = ws %x5D     ; ] Right square bracket
-	ref := p.builder.Push(ast.Node{
+	ref := p.builder.Push(Node{
-		Kind: ast.Table,
+		Kind: Table,
 	})
 
 	b = b[1:]
@@ -199,15 +314,15 @@ func (p *parser) parseStdTable(b []byte) (ast.Reference, []byte, error) {
 	return ref, b, err
 }
 
-func (p *parser) parseKeyval(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseKeyval(b []byte) (reference, []byte, error) {
 	// keyval = key keyval-sep val
-	ref := p.builder.Push(ast.Node{
+	ref := p.builder.Push(Node{
-		Kind: ast.KeyValue,
+		Kind: KeyValue,
 	})
 
 	key, b, err := p.parseKey(b)
 	if err != nil {
-		return ast.InvalidReference, nil, err
+		return invalidReference, nil, err
 	}
 
 	// keyval-sep = ws %x3D ws ; =
@@ -215,12 +330,12 @@ func (p *parser) parseKeyval(b []byte) (ast.Reference, []byte, error) {
 	b = p.parseWhitespace(b)
 
 	if len(b) == 0 {
-		return ast.InvalidReference, nil, newDecodeError(b, "expected = after a key, but the document ends there")
+		return invalidReference, nil, NewParserError(b, "expected = after a key, but the document ends there")
 	}
 
 	b, err = expect('=', b)
 	if err != nil {
-		return ast.InvalidReference, nil, err
+		return invalidReference, nil, err
 	}
 
 	b = p.parseWhitespace(b)
@@ -237,12 +352,12 @@ func (p *parser) parseKeyval(b []byte) (ast.Reference, []byte, error) {
 }
 
 //nolint:cyclop,funlen
-func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseVal(b []byte) (reference, []byte, error) {
 	// val = string / boolean / array / inline-table / date-time / float / integer
-	ref := ast.InvalidReference
+	ref := invalidReference
 
 	if len(b) == 0 {
-		return ref, nil, newDecodeError(b, "expected value, not eof")
+		return ref, nil, NewParserError(b, "expected value, not eof")
 	}
 
 	var err error
@@ -259,8 +374,8 @@ func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 		}
 
 		if err == nil {
-			ref = p.builder.Push(ast.Node{
+			ref = p.builder.Push(Node{
-				Kind: ast.String,
+				Kind: String,
 				Raw:  p.Range(raw),
 				Data: v,
 			})
@@ -277,8 +392,8 @@ func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 		}
 
 		if err == nil {
-			ref = p.builder.Push(ast.Node{
+			ref = p.builder.Push(Node{
-				Kind: ast.String,
+				Kind: String,
 				Raw:  p.Range(raw),
 				Data: v,
 			})
@@ -287,22 +402,22 @@ func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 		return ref, b, err
 	case 't':
 		if !scanFollowsTrue(b) {
-			return ref, nil, newDecodeError(atmost(b, 4), "expected 'true'")
+			return ref, nil, NewParserError(atmost(b, 4), "expected 'true'")
 		}
 
-		ref = p.builder.Push(ast.Node{
+		ref = p.builder.Push(Node{
-			Kind: ast.Bool,
+			Kind: Bool,
 			Data: b[:4],
 		})
 
 		return ref, b[4:], nil
 	case 'f':
 		if !scanFollowsFalse(b) {
-			return ref, nil, newDecodeError(atmost(b, 5), "expected 'false'")
+			return ref, nil, NewParserError(atmost(b, 5), "expected 'false'")
 		}
 
-		ref = p.builder.Push(ast.Node{
+		ref = p.builder.Push(Node{
-			Kind: ast.Bool,
+			Kind: Bool,
 			Data: b[:5],
 		})
 
@@ -324,7 +439,7 @@ func atmost(b []byte, n int) []byte {
 	return b[:n]
 }
 
-func (p *parser) parseLiteralString(b []byte) ([]byte, []byte, []byte, error) {
+func (p *Parser) parseLiteralString(b []byte) ([]byte, []byte, []byte, error) {
 	v, rest, err := scanLiteralString(b)
 	if err != nil {
 		return nil, nil, nil, err
@@ -333,19 +448,20 @@ func (p *parser) parseLiteralString(b []byte) ([]byte, []byte, []byte, error) {
 	return v, v[1 : len(v)-1], rest, nil
 }
 
-func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseInlineTable(b []byte) (reference, []byte, error) {
 	// inline-table = inline-table-open [ inline-table-keyvals ] inline-table-close
 	// inline-table-open  = %x7B ws     ; {
 	// inline-table-close = ws %x7D     ; }
 	// inline-table-sep   = ws %x2C ws  ; , Comma
 	// inline-table-keyvals = keyval [ inline-table-sep inline-table-keyvals ]
-	parent := p.builder.Push(ast.Node{
+	parent := p.builder.Push(Node{
-		Kind: ast.InlineTable,
+		Kind: InlineTable,
+		Raw:  p.Range(b[:1]),
 	})
 
 	first := true
 
-	var child ast.Reference
+	var child reference
 
 	b = b[1:]
 
@@ -356,7 +472,7 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 		b = p.parseWhitespace(b)
 
 		if len(b) == 0 {
-			return parent, nil, newDecodeError(previousB[:1], "inline table is incomplete")
+			return parent, nil, NewParserError(previousB[:1], "inline table is incomplete")
 		}
 
 		if b[0] == '}' {
@@ -371,7 +487,7 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 			b = p.parseWhitespace(b)
 		}
 
-		var kv ast.Reference
+		var kv reference
 
 		kv, b, err = p.parseKeyval(b)
 		if err != nil {
@@ -394,7 +510,7 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 }
 
 //nolint:funlen,cyclop
-func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseValArray(b []byte) (reference, []byte, error) {
 	// array = array-open [ array-values ] ws-comment-newline array-close
 	// array-open =  %x5B ; [
 	// array-close = %x5D ; ]
@@ -405,23 +521,39 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 	arrayStart := b
 	b = b[1:]
 
-	parent := p.builder.Push(ast.Node{
+	parent := p.builder.Push(Node{
-		Kind: ast.Array,
+		Kind: Array,
 	})
 
+	// First indicates whether the parser is looking for the first element
+	// (non-comment) of the array.
 	first := true
 
-	var lastChild ast.Reference
+	lastChild := invalidReference
+
+	addChild := func(valueRef reference) {
+		if lastChild == invalidReference {
+			p.builder.AttachChild(parent, valueRef)
+		} else {
+			p.builder.Chain(lastChild, valueRef)
+		}
+		lastChild = valueRef
+	}
 
 	var err error
 	for len(b) > 0 {
-		b, err = p.parseOptionalWhitespaceCommentNewline(b)
+		cref := invalidReference
+		cref, b, err = p.parseOptionalWhitespaceCommentNewline(b)
 		if err != nil {
 			return parent, nil, err
 		}
 
+		if cref != invalidReference {
+			addChild(cref)
+		}
+
 		if len(b) == 0 {
-			return parent, nil, newDecodeError(arrayStart[:1], "array is incomplete")
+			return parent, nil, NewParserError(arrayStart[:1], "array is incomplete")
 		}
 
 		if b[0] == ']' {
@@ -430,16 +562,19 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 
 		if b[0] == ',' {
 			if first {
-				return parent, nil, newDecodeError(b[0:1], "array cannot start with comma")
+				return parent, nil, NewParserError(b[0:1], "array cannot start with comma")
 			}
 			b = b[1:]
 
-			b, err = p.parseOptionalWhitespaceCommentNewline(b)
+			cref, b, err = p.parseOptionalWhitespaceCommentNewline(b)
 			if err != nil {
 				return parent, nil, err
 			}
+			if cref != invalidReference {
+				addChild(cref)
+			}
 		} else if !first {
-			return parent, nil, newDecodeError(b[0:1], "array elements must be separated by commas")
+			return parent, nil, NewParserError(b[0:1], "array elements must be separated by commas")
 		}
 
 		// TOML allows trailing commas in arrays.
@@ -447,23 +582,22 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 			break
 		}
 
-		var valueRef ast.Reference
+		var valueRef reference
 		valueRef, b, err = p.parseVal(b)
 		if err != nil {
 			return parent, nil, err
 		}
 
-		if first {
+		addChild(valueRef)
-			p.builder.AttachChild(parent, valueRef)
-		} else {
-			p.builder.Chain(lastChild, valueRef)
-		}
-		lastChild = valueRef
 
-		b, err = p.parseOptionalWhitespaceCommentNewline(b)
+		cref, b, err = p.parseOptionalWhitespaceCommentNewline(b)
 		if err != nil {
 			return parent, nil, err
 		}
+		if cref != invalidReference {
+			addChild(cref)
+		}
+
 		first = false
 	}
 
@@ -472,15 +606,34 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 	return parent, rest, err
 }
 
-func (p *parser) parseOptionalWhitespaceCommentNewline(b []byte) ([]byte, error) {
+func (p *Parser) parseOptionalWhitespaceCommentNewline(b []byte) (reference, []byte, error) {
+	rootCommentRef := invalidReference
+	latestCommentRef := invalidReference
+
+	addComment := func(ref reference) {
+		if rootCommentRef == invalidReference {
+			rootCommentRef = ref
+		} else if latestCommentRef == invalidReference {
+			p.builder.AttachChild(rootCommentRef, ref)
+			latestCommentRef = ref
+		} else {
+			p.builder.Chain(latestCommentRef, ref)
+			latestCommentRef = ref
+		}
+	}
+
 	for len(b) > 0 {
 		var err error
 		b = p.parseWhitespace(b)
 
 		if len(b) > 0 && b[0] == '#' {
-			_, b, err = scanComment(b)
+			var ref reference
+			ref, b, err = p.parseComment(b)
 			if err != nil {
-				return nil, err
+				return invalidReference, nil, err
+			}
+			if ref != invalidReference {
+				addComment(ref)
 			}
 		}
 
@@ -491,17 +644,17 @@ func (p *parser) parseOptionalWhitespaceCommentNewline(b []byte) ([]byte, error)
 		if b[0] == '\n' || b[0] == '\r' {
 			b, err = p.parseNewline(b)
 			if err != nil {
-				return nil, err
+				return invalidReference, nil, err
 			}
 		} else {
 			break
 		}
 	}
 
-	return b, nil
+	return rootCommentRef, b, nil
 }
 
-func (p *parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, []byte, error) {
+func (p *Parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, []byte, error) {
 	token, rest, err := scanMultilineLiteralString(b)
 	if err != nil {
 		return nil, nil, nil, err
@@ -520,7 +673,7 @@ func (p *parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, []byte,
 }
 
 //nolint:funlen,gocognit,cyclop
-func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, error) {
+func (p *Parser) parseMultilineBasicString(b []byte) ([]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
@@ -551,11 +704,11 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, er
 
 	if !escaped {
 		str := token[startIdx:endIdx]
-		verr := utf8TomlValidAlreadyEscaped(str)
+		verr := characters.Utf8TomlValidAlreadyEscaped(str)
 		if verr.Zero() {
 			return token, str, rest, nil
 		}
-		return nil, nil, nil, newDecodeError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
+		return nil, nil, nil, NewParserError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
 	}
 
 	var builder bytes.Buffer
@@ -617,6 +770,8 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, er
 				builder.WriteByte('\r')
 			case 't':
 				builder.WriteByte('\t')
+			case 'e':
+				builder.WriteByte(0x1B)
 			case 'u':
 				x, err := hexToRune(atmost(token[i+1:], 4), 4)
 				if err != nil {
@@ -633,13 +788,13 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, er
 				builder.WriteRune(x)
 				i += 8
 			default:
-				return nil, nil, nil, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
+				return nil, nil, nil, NewParserError(token[i:i+1], "invalid escaped character %#U", c)
 			}
 			i++
 		} else {
-			size := utf8ValidNext(token[i:])
+			size := characters.Utf8ValidNext(token[i:])
 			if size == 0 {
-				return nil, nil, nil, newDecodeError(token[i:i+1], "invalid character %#U", c)
+				return nil, nil, nil, NewParserError(token[i:i+1], "invalid character %#U", c)
 			}
 			builder.Write(token[i : i+size])
 			i += size
@@ -649,7 +804,7 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, er
 	return token, builder.Bytes(), rest, nil
 }
 
-func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseKey(b []byte) (reference, []byte, error) {
 	// key = simple-key / dotted-key
 	// simple-key = quoted-key / unquoted-key
 	//
@@ -660,11 +815,11 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 	// dot-sep   = ws %x2E ws  ; . Period
 	raw, key, b, err := p.parseSimpleKey(b)
 	if err != nil {
-		return ast.InvalidReference, nil, err
+		return invalidReference, nil, err
 	}
 
-	ref := p.builder.Push(ast.Node{
+	ref := p.builder.Push(Node{
-		Kind: ast.Key,
+		Kind: Key,
 		Raw:  p.Range(raw),
 		Data: key,
 	})
@@ -679,8 +834,8 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 				return ref, nil, err
 			}
 
-			p.builder.PushAndChain(ast.Node{
+			p.builder.PushAndChain(Node{
-				Kind: ast.Key,
+				Kind: Key,
 				Raw:  p.Range(raw),
 				Data: key,
 			})
@@ -692,9 +847,9 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 	return ref, b, nil
 }
 
-func (p *parser) parseSimpleKey(b []byte) (raw, key, rest []byte, err error) {
+func (p *Parser) parseSimpleKey(b []byte) (raw, key, rest []byte, err error) {
 	if len(b) == 0 {
-		return nil, nil, nil, newDecodeError(b, "expected key but found none")
+		return nil, nil, nil, NewParserError(b, "expected key but found none")
 	}
 
 	// simple-key = quoted-key / unquoted-key
@@ -709,12 +864,12 @@ func (p *parser) parseSimpleKey(b []byte) (raw, key, rest []byte, err error) {
 		key, rest = scanUnquotedKey(b)
 		return key, key, rest, nil
 	default:
-		return nil, nil, nil, newDecodeError(b[0:1], "invalid character at start of key: %c", b[0])
+		return nil, nil, nil, NewParserError(b[0:1], "invalid character at start of key: %c", b[0])
 	}
 }
 
 //nolint:funlen,cyclop
-func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
+func (p *Parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 	// basic-string = quotation-mark *basic-char quotation-mark
 	// quotation-mark = %x22            ; "
 	// basic-char = basic-unescaped / escaped
@@ -742,11 +897,11 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 	// validate the string and return a direct reference to the buffer.
 	if !escaped {
 		str := token[startIdx:endIdx]
-		verr := utf8TomlValidAlreadyEscaped(str)
+		verr := characters.Utf8TomlValidAlreadyEscaped(str)
 		if verr.Zero() {
 			return token, str, rest, nil
 		}
-		return nil, nil, nil, newDecodeError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
+		return nil, nil, nil, NewParserError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
 	}
 
 	i := startIdx
@@ -774,6 +929,8 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 				builder.WriteByte('\r')
 			case 't':
 				builder.WriteByte('\t')
+			case 'e':
+				builder.WriteByte(0x1B)
 			case 'u':
 				x, err := hexToRune(token[i+1:len(token)-1], 4)
 				if err != nil {
@@ -791,13 +948,13 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 				builder.WriteRune(x)
 				i += 8
 			default:
-				return nil, nil, nil, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
+				return nil, nil, nil, NewParserError(token[i:i+1], "invalid escaped character %#U", c)
 			}
 			i++
 		} else {
-			size := utf8ValidNext(token[i:])
+			size := characters.Utf8ValidNext(token[i:])
 			if size == 0 {
-				return nil, nil, nil, newDecodeError(token[i:i+1], "invalid character %#U", c)
+				return nil, nil, nil, NewParserError(token[i:i+1], "invalid character %#U", c)
 			}
 			builder.Write(token[i : i+size])
 			i += size
@@ -809,7 +966,7 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 
 func hexToRune(b []byte, length int) (rune, error) {
 	if len(b) < length {
-		return -1, newDecodeError(b, "unicode point needs %d character, not %d", length, len(b))
+		return -1, NewParserError(b, "unicode point needs %d character, not %d", length, len(b))
 	}
 	b = b[:length]
 
@@ -824,19 +981,19 @@ func hexToRune(b []byte, length int) (rune, error) {
 		case 'A' <= c && c <= 'F':
 			d = uint32(c - 'A' + 10)
 		default:
-			return -1, newDecodeError(b[i:i+1], "non-hex character")
+			return -1, NewParserError(b[i:i+1], "non-hex character")
 		}
 		r = r*16 + d
 	}
 
 	if r > unicode.MaxRune || 0xD800 <= r && r < 0xE000 {
-		return -1, newDecodeError(b, "escape sequence is invalid Unicode code point")
+		return -1, NewParserError(b, "escape sequence is invalid Unicode code point")
 	}
 
 	return rune(r), nil
 }
 
-func (p *parser) parseWhitespace(b []byte) []byte {
+func (p *Parser) parseWhitespace(b []byte) []byte {
 	// ws = *wschar
 	// wschar =  %x20  ; Space
 	// wschar =/ %x09  ; Horizontal tab
@@ -846,25 +1003,27 @@ func (p *parser) parseWhitespace(b []byte) []byte {
 }
 
 //nolint:cyclop
-func (p *parser) parseIntOrFloatOrDateTime(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) parseIntOrFloatOrDateTime(b []byte) (reference, []byte, error) {
 	switch b[0] {
 	case 'i':
 		if !scanFollowsInf(b) {
-			return ast.InvalidReference, nil, newDecodeError(atmost(b, 3), "expected 'inf'")
+			return invalidReference, nil, NewParserError(atmost(b, 3), "expected 'inf'")
 		}
 
-		return p.builder.Push(ast.Node{
+		return p.builder.Push(Node{
-			Kind: ast.Float,
+			Kind: Float,
 			Data: b[:3],
+			Raw:  p.Range(b[:3]),
 		}), b[3:], nil
 	case 'n':
 		if !scanFollowsNan(b) {
-			return ast.InvalidReference, nil, newDecodeError(atmost(b, 3), "expected 'nan'")
+			return invalidReference, nil, NewParserError(atmost(b, 3), "expected 'nan'")
 		}
 
-		return p.builder.Push(ast.Node{
+		return p.builder.Push(Node{
-			Kind: ast.Float,
+			Kind: Float,
 			Data: b[:3],
+			Raw:  p.Range(b[:3]),
 		}), b[3:], nil
 	case '+', '-':
 		return p.scanIntOrFloat(b)
@@ -894,7 +1053,7 @@ func (p *parser) parseIntOrFloatOrDateTime(b []byte) (ast.Reference, []byte, err
 	return p.scanIntOrFloat(b)
 }
 
-func (p *parser) scanDateTime(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) scanDateTime(b []byte) (reference, []byte, error) {
 	// scans for contiguous characters in [0-9T:Z.+-], and up to one space if
 	// followed by a digit.
 	hasDate := false
@@ -937,30 +1096,30 @@ byteLoop:
 		}
 	}
 
-	var kind ast.Kind
+	var kind Kind
 
 	if hasTime {
 		if hasDate {
 			if hasTz {
-				kind = ast.DateTime
+				kind = DateTime
 			} else {
-				kind = ast.LocalDateTime
+				kind = LocalDateTime
 			}
 		} else {
-			kind = ast.LocalTime
+			kind = LocalTime
 		}
 	} else {
-		kind = ast.LocalDate
+		kind = LocalDate
 	}
 
-	return p.builder.Push(ast.Node{
+	return p.builder.Push(Node{
 		Kind: kind,
 		Data: b[:i],
 	}), b[i:], nil
 }
 
 //nolint:funlen,gocognit,cyclop
-func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
+func (p *Parser) scanIntOrFloat(b []byte) (reference, []byte, error) {
 	i := 0
 
 	if len(b) > 2 && b[0] == '0' && b[1] != '.' && b[1] != 'e' && b[1] != 'E' {
@@ -986,9 +1145,10 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 			}
 		}
 
-		return p.builder.Push(ast.Node{
+		return p.builder.Push(Node{
-			Kind: ast.Integer,
+			Kind: Integer,
 			Data: b[:i],
+			Raw:  p.Range(b[:i]),
 		}), b[i:], nil
 	}
 
@@ -1009,42 +1169,45 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 
 		if c == 'i' {
 			if scanFollowsInf(b[i:]) {
-				return p.builder.Push(ast.Node{
+				return p.builder.Push(Node{
-					Kind: ast.Float,
+					Kind: Float,
 					Data: b[:i+3],
+					Raw:  p.Range(b[:i+3]),
 				}), b[i+3:], nil
 			}
 
-			return ast.InvalidReference, nil, newDecodeError(b[i:i+1], "unexpected character 'i' while scanning for a number")
+			return invalidReference, nil, NewParserError(b[i:i+1], "unexpected character 'i' while scanning for a number")
 		}
 
 		if c == 'n' {
 			if scanFollowsNan(b[i:]) {
-				return p.builder.Push(ast.Node{
+				return p.builder.Push(Node{
-					Kind: ast.Float,
+					Kind: Float,
 					Data: b[:i+3],
+					Raw:  p.Range(b[:i+3]),
 				}), b[i+3:], nil
 			}
 
-			return ast.InvalidReference, nil, newDecodeError(b[i:i+1], "unexpected character 'n' while scanning for a number")
+			return invalidReference, nil, NewParserError(b[i:i+1], "unexpected character 'n' while scanning for a number")
 		}
 
 		break
 	}
 
 	if i == 0 {
-		return ast.InvalidReference, b, newDecodeError(b, "incomplete number")
+		return invalidReference, b, NewParserError(b, "incomplete number")
 	}
 
-	kind := ast.Integer
+	kind := Integer
 
 	if isFloat {
-		kind = ast.Float
+		kind = Float
 	}
 
-	return p.builder.Push(ast.Node{
+	return p.builder.Push(Node{
 		Kind: kind,
 		Data: b[:i],
+		Raw:  p.Range(b[:i]),
 	}), b[i:], nil
 }
 
@@ -1071,11 +1234,11 @@ func isValidBinaryRune(r byte) bool {
 
 func expect(x byte, b []byte) ([]byte, error) {
 	if len(b) == 0 {
-		return nil, newDecodeError(b, "expected character %c but the document ended here", x)
+		return nil, NewParserError(b, "expected character %c but the document ended here", x)
 	}
 
 	if b[0] != x {
-		return nil, newDecodeError(b[0:1], "expected character %c", x)
+		return nil, NewParserError(b[0:1], "expected character %c", x)
 	}
 
 	return b[1:], nil

unstable/parser_test.go

@@ -0,0 +1,629 @@
+package unstable
+
+import (
+	"fmt"
+	"strconv"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestParser_AST_Numbers(t *testing.T) {
+	examples := []struct {
+		desc  string
+		input string
+		kind  Kind
+		err   bool
+	}{
+		{
+			desc:  "integer just digits",
+			input: `1234`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer zero",
+			input: `0`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer sign",
+			input: `+99`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer hex uppercase",
+			input: `0xDEADBEEF`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer hex lowercase",
+			input: `0xdead_beef`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer octal",
+			input: `0o01234567`,
+			kind:  Integer,
+		},
+		{
+			desc:  "integer binary",
+			input: `0b11010110`,
+			kind:  Integer,
+		},
+		{
+			desc:  "float zero",
+			input: `0.0`,
+			kind:  Float,
+		},
+		{
+			desc:  "float positive zero",
+			input: `+0.0`,
+			kind:  Float,
+		},
+		{
+			desc:  "float negative zero",
+			input: `-0.0`,
+			kind:  Float,
+		},
+		{
+			desc:  "float pi",
+			input: `3.1415`,
+			kind:  Float,
+		},
+		{
+			desc:  "float negative",
+			input: `-0.01`,
+			kind:  Float,
+		},
+		{
+			desc:  "float signed exponent",
+			input: `5e+22`,
+			kind:  Float,
+		},
+		{
+			desc:  "float exponent lowercase",
+			input: `1e06`,
+			kind:  Float,
+		},
+		{
+			desc:  "float exponent uppercase",
+			input: `-2E-2`,
+			kind:  Float,
+		},
+		{
+			desc:  "float fractional with exponent",
+			input: `6.626e-34`,
+			kind:  Float,
+		},
+		{
+			desc:  "float underscores",
+			input: `224_617.445_991_228`,
+			kind:  Float,
+		},
+		{
+			desc:  "inf",
+			input: `inf`,
+			kind:  Float,
+		},
+		{
+			desc:  "inf negative",
+			input: `-inf`,
+			kind:  Float,
+		},
+		{
+			desc:  "inf positive",
+			input: `+inf`,
+			kind:  Float,
+		},
+		{
+			desc:  "nan",
+			input: `nan`,
+			kind:  Float,
+		},
+		{
+			desc:  "nan negative",
+			input: `-nan`,
+			kind:  Float,
+		},
+		{
+			desc:  "nan positive",
+			input: `+nan`,
+			kind:  Float,
+		},
+	}
+
+	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: KeyValue,
+					Children: []astNode{
+						{Kind: e.kind, Data: []byte(e.input)},
+						{Kind: Key, Data: []byte(`A`)},
+					},
+				}
+				compareNode(t, expected, p.Expression())
+			}
+		})
+	}
+}
+
+type (
+	astNode struct {
+		Kind     Kind
+		Data     []byte
+		Children []astNode
+	}
+)
+
+func compareNode(t *testing.T, e astNode, n *Node) {
+	t.Helper()
+	require.Equal(t, e.Kind, n.Kind)
+	require.Equal(t, e.Data, n.Data)
+
+	compareIterator(t, e.Children, n.Children())
+}
+
+func compareIterator(t *testing.T, expected []astNode, actual Iterator) {
+	t.Helper()
+	idx := 0
+
+	for actual.Next() {
+		n := actual.Node()
+
+		if idx >= len(expected) {
+			t.Fatal("extra child in actual tree")
+		}
+		e := expected[idx]
+
+		compareNode(t, e, n)
+
+		idx++
+	}
+
+	if idx < len(expected) {
+		t.Fatal("missing children in actual", "idx =", idx, "expected =", len(expected))
+	}
+}
+
+//nolint:funlen
+func TestParser_AST(t *testing.T) {
+	examples := []struct {
+		desc  string
+		input string
+		ast   astNode
+		err   bool
+	}{
+		{
+			desc:  "simple string assignment",
+			input: `A = "hello"`,
+			ast: astNode{
+				Kind: KeyValue,
+				Children: []astNode{
+					{
+						Kind: String,
+						Data: []byte(`hello`),
+					},
+					{
+						Kind: Key,
+						Data: []byte(`A`),
+					},
+				},
+			},
+		},
+		{
+			desc:  "simple bool assignment",
+			input: `A = true`,
+			ast: astNode{
+				Kind: KeyValue,
+				Children: []astNode{
+					{
+						Kind: Bool,
+						Data: []byte(`true`),
+					},
+					{
+						Kind: Key,
+						Data: []byte(`A`),
+					},
+				},
+			},
+		},
+		{
+			desc:  "array of strings",
+			input: `A = ["hello", ["world", "again"]]`,
+			ast: astNode{
+				Kind: KeyValue,
+				Children: []astNode{
+					{
+						Kind: Array,
+						Children: []astNode{
+							{
+								Kind: String,
+								Data: []byte(`hello`),
+							},
+							{
+								Kind: Array,
+								Children: []astNode{
+									{
+										Kind: String,
+										Data: []byte(`world`),
+									},
+									{
+										Kind: String,
+										Data: []byte(`again`),
+									},
+								},
+							},
+						},
+					},
+					{
+						Kind: Key,
+						Data: []byte(`A`),
+					},
+				},
+			},
+		},
+		{
+			desc:  "array of arrays of strings",
+			input: `A = ["hello", "world"]`,
+			ast: astNode{
+				Kind: KeyValue,
+				Children: []astNode{
+					{
+						Kind: Array,
+						Children: []astNode{
+							{
+								Kind: String,
+								Data: []byte(`hello`),
+							},
+							{
+								Kind: String,
+								Data: []byte(`world`),
+							},
+						},
+					},
+					{
+						Kind: Key,
+						Data: []byte(`A`),
+					},
+				},
+			},
+		},
+		{
+			desc:  "inline table",
+			input: `name = { first = "Tom", last = "Preston-Werner" }`,
+			ast: astNode{
+				Kind: KeyValue,
+				Children: []astNode{
+					{
+						Kind: InlineTable,
+						Children: []astNode{
+							{
+								Kind: KeyValue,
+								Children: []astNode{
+									{Kind: String, Data: []byte(`Tom`)},
+									{Kind: Key, Data: []byte(`first`)},
+								},
+							},
+							{
+								Kind: KeyValue,
+								Children: []astNode{
+									{Kind: String, Data: []byte(`Preston-Werner`)},
+									{Kind: Key, Data: []byte(`last`)},
+								},
+							},
+						},
+					},
+					{
+						Kind: Key,
+						Data: []byte(`name`),
+					},
+				},
+			},
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			p := Parser{}
+			p.Reset([]byte(e.input))
+			p.NextExpression()
+			err := p.Error()
+			if e.err {
+				require.Error(t, err)
+			} else {
+				require.NoError(t, err)
+				compareNode(t, e.ast, p.Expression())
+			}
+		})
+	}
+}
+
+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  Kind
+		err   bool
+	}{
+		{
+			desc:  "offset-date-time with delim 'T' and UTC offset",
+			input: `2021-07-21T12:08:05Z`,
+			kind:  DateTime,
+		},
+		{
+			desc:  "offset-date-time with space delim and +8hours offset",
+			input: `2021-07-21 12:08:05+08:00`,
+			kind:  DateTime,
+		},
+		{
+			desc:  "local-date-time with nano second",
+			input: `2021-07-21T12:08:05.666666666`,
+			kind:  LocalDateTime,
+		},
+		{
+			desc:  "local-date-time",
+			input: `2021-07-21T12:08:05`,
+			kind:  LocalDateTime,
+		},
+		{
+			desc:  "local-date",
+			input: `2021-07-21`,
+			kind:  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: KeyValue,
+					Children: []astNode{
+						{Kind: e.kind, Data: []byte(e.input)},
+						{Kind: Key, Data: []byte(`A`)},
+					},
+				}
+				compareNode(t, expected, p.Expression())
+			}
+		})
+	}
+}
+
+// This example demonstrates how to parse a TOML document and preserving
+// comments.  Comments are stored in the AST as Comment nodes. This example
+// displays the structure of the full AST generated by the parser using the
+// following structure:
+//
+//  1. Each root-level expression is separated by three dashes.
+//  2. Bytes associated to a node are displayed in square brackets.
+//  3. Siblings have the same indentation.
+//  4. Children of a node are indented one level.
+func ExampleParser_comments() {
+	doc := `# Top of the document comment.
+# Optional, any amount of lines.
+
+# Above table.
+[table] # Next to table.
+# Above simple value.
+key = "value" # Next to simple value.
+# Below simple value.
+
+# Some comment alone.
+
+# Multiple comments, on multiple lines.
+
+# Above inline table.
+name = { first = "Tom", last = "Preston-Werner" } # Next to inline table.
+# Below inline table.
+
+# Above array.
+array = [ 1, 2, 3 ] # Next to one-line array.
+# Below array.
+
+# Above multi-line array.
+key5 = [ # Next to start of inline array.
+  # Second line before array content.
+  1, # Next to first element.
+  # After first element.
+  # Before second element.
+  2,
+  3, # Next to last element
+  # After last element.
+] # Next to end of array.
+# Below multi-line array.
+
+# Before array table.
+[[products]] # Next to array table.
+# After array table.
+`
+
+	var printGeneric func(*Parser, int, *Node)
+	printGeneric = func(p *Parser, indent int, e *Node) {
+		if e == nil {
+			return
+		}
+		s := p.Shape(e.Raw)
+		x := fmt.Sprintf("%d:%d->%d:%d (%d->%d)", s.Start.Line, s.Start.Column, s.End.Line, s.End.Column, s.Start.Offset, s.End.Offset)
+		fmt.Printf("%-25s | %s%s [%s]\n", x, strings.Repeat("  ", indent), e.Kind, e.Data)
+		printGeneric(p, indent+1, e.Child())
+		printGeneric(p, indent, e.Next())
+	}
+
+	printTree := func(p *Parser) {
+		for p.NextExpression() {
+			e := p.Expression()
+			fmt.Println("---")
+			printGeneric(p, 0, e)
+		}
+		if err := p.Error(); err != nil {
+			panic(err)
+		}
+	}
+
+	p := &Parser{
+		KeepComments: true,
+	}
+	p.Reset([]byte(doc))
+	printTree(p)
+
+	// Output:
+	// ---
+	// 1:1->1:31 (0->30)         | Comment [# Top of the document comment.]
+	// ---
+	// 2:1->2:33 (31->63)        | Comment [# Optional, any amount of lines.]
+	// ---
+	// 4:1->4:15 (65->79)        | Comment [# Above table.]
+	// ---
+	// 1:1->1:1 (0->0)           | Table []
+	// 5:2->5:7 (81->86)         |   Key [table]
+	// 5:9->5:25 (88->104)       | Comment [# Next to table.]
+	// ---
+	// 6:1->6:22 (105->126)      | Comment [# Above simple value.]
+	// ---
+	// 1:1->1:1 (0->0)           | KeyValue []
+	// 7:7->7:14 (133->140)      |   String [value]
+	// 7:1->7:4 (127->130)       |   Key [key]
+	// 7:15->7:38 (141->164)     | Comment [# Next to simple value.]
+	// ---
+	// 8:1->8:22 (165->186)      | Comment [# Below simple value.]
+	// ---
+	// 10:1->10:22 (188->209)    | Comment [# Some comment alone.]
+	// ---
+	// 12:1->12:40 (211->250)    | Comment [# Multiple comments, on multiple lines.]
+	// ---
+	// 14:1->14:22 (252->273)    | Comment [# Above inline table.]
+	// ---
+	// 1:1->1:1 (0->0)           | KeyValue []
+	// 15:8->15:9 (281->282)     |   InlineTable []
+	// 1:1->1:1 (0->0)           |     KeyValue []
+	// 15:18->15:23 (291->296)   |       String [Tom]
+	// 15:10->15:15 (283->288)   |       Key [first]
+	// 1:1->1:1 (0->0)           |     KeyValue []
+	// 15:32->15:48 (305->321)   |       String [Preston-Werner]
+	// 15:25->15:29 (298->302)   |       Key [last]
+	// 15:1->15:5 (274->278)     |   Key [name]
+	// 15:51->15:74 (324->347)   | Comment [# Next to inline table.]
+	// ---
+	// 16:1->16:22 (348->369)    | Comment [# Below inline table.]
+	// ---
+	// 18:1->18:15 (371->385)    | Comment [# Above array.]
+	// ---
+	// 1:1->1:1 (0->0)           | KeyValue []
+	// 1:1->1:1 (0->0)           |   Array []
+	// 19:11->19:12 (396->397)   |     Integer [1]
+	// 19:14->19:15 (399->400)   |     Integer [2]
+	// 19:17->19:18 (402->403)   |     Integer [3]
+	// 19:1->19:6 (386->391)     |   Key [array]
+	// 19:21->19:46 (406->431)   | Comment [# Next to one-line array.]
+	// ---
+	// 20:1->20:15 (432->446)    | Comment [# Below array.]
+	// ---
+	// 22:1->22:26 (448->473)    | Comment [# Above multi-line array.]
+	// ---
+	// 1:1->1:1 (0->0)           | KeyValue []
+	// 1:1->1:1 (0->0)           |   Array []
+	// 23:10->23:42 (483->515)   |     Comment [# Next to start of inline array.]
+	// 24:3->24:38 (518->553)    |       Comment [# Second line before array content.]
+	// 25:3->25:4 (556->557)     |     Integer [1]
+	// 25:6->25:30 (559->583)    |     Comment [# Next to first element.]
+	// 26:3->26:25 (586->608)    |       Comment [# After first element.]
+	// 27:3->27:27 (611->635)    |       Comment [# Before second element.]
+	// 28:3->28:4 (638->639)     |     Integer [2]
+	// 29:3->29:4 (643->644)     |     Integer [3]
+	// 29:6->29:28 (646->668)    |     Comment [# Next to last element]
+	// 30:3->30:24 (671->692)    |       Comment [# After last element.]
+	// 23:1->23:5 (474->478)     |   Key [key5]
+	// 31:3->31:26 (695->718)    | Comment [# Next to end of array.]
+	// ---
+	// 32:1->32:26 (719->744)    | Comment [# Below multi-line array.]
+	// ---
+	// 34:1->34:22 (746->767)    | Comment [# Before array table.]
+	// ---
+	// 1:1->1:1 (0->0)           | ArrayTable []
+	// 35:3->35:11 (770->778)    |   Key [products]
+	// 35:14->35:36 (781->803)   | Comment [# Next to array table.]
+	// ---
+	// 36:1->36:21 (804->824)    | Comment [# After array table.]
+}
+
+func ExampleParser() {
+	doc := `
+	hello = "world"
+	value = 42
+	`
+	p := Parser{}
+	p.Reset([]byte(doc))
+	for p.NextExpression() {
+		e := p.Expression()
+		fmt.Printf("Expression: %s\n", e.Kind)
+		value := e.Value()
+		it := e.Key()
+		k := it.Node() // shortcut: we know there is no dotted key in the example
+		fmt.Printf("%s -> (%s) %s\n", k.Data, value.Kind, value.Data)
+	}
+
+	// Output:
+	// Expression: KeyValue
+	// hello -> (String) world
+	// Expression: KeyValue
+	// value -> (Integer) 42
+}

unstable/scanner.go

@@ -1,4 +1,6 @@
-package toml
+package unstable
+
+import "github.com/pelletier/go-toml/v2/internal/characters"
 
 func scanFollows(b []byte, pattern string) bool {
 	n := len(pattern)
@@ -54,16 +56,16 @@ func scanLiteralString(b []byte) ([]byte, []byte, error) {
 		case '\'':
 			return b[:i+1], b[i+1:], nil
 		case '\n', '\r':
-			return nil, nil, newDecodeError(b[i:i+1], "literal strings cannot have new lines")
+			return nil, nil, NewParserError(b[i:i+1], "literal strings cannot have new lines")
 		}
-		size := utf8ValidNext(b[i:])
+		size := characters.Utf8ValidNext(b[i:])
 		if size == 0 {
-			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+			return nil, nil, NewParserError(b[i:i+1], "invalid character")
 		}
 		i += size
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], "unterminated literal string")
+	return nil, nil, NewParserError(b[len(b):], "unterminated literal string")
 }
 
 func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
@@ -98,39 +100,39 @@ func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
 				i++
 
 				if i < len(b) && b[i] == '\'' {
-					return nil, nil, newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+					return nil, nil, NewParserError(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`)
+				return nil, nil, NewParserError(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`)
+				return nil, nil, NewParserError(b[i:i+2], `need a \n after \r`)
 			}
 			i += 2 // skip the \n
 			continue
 		}
-		size := utf8ValidNext(b[i:])
+		size := characters.Utf8ValidNext(b[i:])
 		if size == 0 {
-			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+			return nil, nil, NewParserError(b[i:i+1], "invalid character")
 		}
 		i += size
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], `multiline literal string not terminated by '''`)
+	return nil, nil, NewParserError(b[len(b):], `multiline literal string not terminated by '''`)
 }
 
 func scanWindowsNewline(b []byte) ([]byte, []byte, error) {
 	const lenCRLF = 2
 	if len(b) < lenCRLF {
-		return nil, nil, newDecodeError(b, "windows new line expected")
+		return nil, nil, NewParserError(b, "windows new line expected")
 	}
 
 	if b[1] != '\n' {
-		return nil, nil, newDecodeError(b, `windows new line should be \r\n`)
+		return nil, nil, NewParserError(b, `windows new line should be \r\n`)
 	}
 
 	return b[:lenCRLF], b[lenCRLF:], nil
@@ -149,7 +151,6 @@ func scanWhitespace(b []byte) ([]byte, []byte) {
 	return b, b[len(b):]
 }
 
-//nolint:unparam
 func scanComment(b []byte) ([]byte, []byte, error) {
 	// comment-start-symbol = %x23 ; #
 	// non-ascii = %x80-D7FF / %xE000-10FFFF
@@ -165,11 +166,11 @@ func scanComment(b []byte) ([]byte, []byte, error) {
 			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")
+			return nil, nil, NewParserError(b[i:i+1], "invalid character in comment")
 		}
-		size := utf8ValidNext(b[i:])
+		size := characters.Utf8ValidNext(b[i:])
 		if size == 0 {
-			return nil, nil, newDecodeError(b[i:i+1], "invalid character in comment")
+			return nil, nil, NewParserError(b[i:i+1], "invalid character in comment")
 		}
 
 		i += size
@@ -192,17 +193,17 @@ func scanBasicString(b []byte) ([]byte, bool, []byte, error) {
 		case '"':
 			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")
+			return nil, escaped, nil, NewParserError(b[i:i+1], "basic strings cannot have new lines")
 		case '\\':
 			if len(b) < i+2 {
-				return nil, escaped, nil, newDecodeError(b[i:i+1], "need a character after \\")
+				return nil, escaped, nil, NewParserError(b[i:i+1], "need a character after \\")
 			}
 			escaped = true
 			i++ // skip the next character
 		}
 	}
 
-	return nil, escaped, nil, newDecodeError(b[len(b):], `basic string not terminated by "`)
+	return nil, escaped, nil, NewParserError(b[len(b):], `basic string not terminated by "`)
 }
 
 func scanMultilineBasicString(b []byte) ([]byte, bool, []byte, error) {
@@ -243,27 +244,27 @@ func scanMultilineBasicString(b []byte) ([]byte, bool, []byte, error) {
 				i++
 
 				if i < len(b) && b[i] == '"' {
-					return nil, escaped, nil, newDecodeError(b[i-3:i+1], `""" not allowed in multiline basic string`)
+					return nil, escaped, nil, NewParserError(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, escaped, nil, newDecodeError(b[len(b):], "need a character after \\")
+				return nil, escaped, nil, NewParserError(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`)
+				return nil, escaped, nil, NewParserError(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`)
+				return nil, escaped, nil, NewParserError(b[i:i+2], `need a \n after \r`)
 			}
 			i++ // skip the \n
 		}
 	}
 
-	return nil, escaped, nil, newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
+	return nil, escaped, nil, NewParserError(b[len(b):], `multiline basic string not terminated by """`)
 }