Keep separate fn for []interface{} unmarshal

```
name                               old time/op    new time/op    delta
UnmarshalDataset/config-2            24.9ms ± 0%    24.6ms ± 0%  -1.09%  (p=0.029 n=4+4)
UnmarshalDataset/canada-2            61.7ms ± 1%    62.1ms ± 3%    ~     (p=1.000 n=5+5)
UnmarshalDataset/citm_catalog-2      24.7ms ± 1%    24.2ms ± 0%  -2.30%  (p=0.008 n=5+5)
UnmarshalDataset/twitter-2           10.9ms ± 2%    10.7ms ± 1%  -1.46%  (p=0.008 n=5+5)
UnmarshalDataset/code-2               108ms ± 0%     106ms ± 0%  -1.91%  (p=0.008 n=5+5)
UnmarshalDataset/example-2            176µs ± 0%     173µs ± 0%  -1.83%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/struct-2     586ns ± 1%     587ns ± 0%    ~     (p=0.690 n=5+5)
Unmarshal/SimpleDocument/map-2        876ns ± 0%     872ns ± 0%    ~     (p=0.095 n=5+5)
Unmarshal/ReferenceFile/struct-2     49.5µs ± 0%    49.5µs ± 0%    ~     (p=0.222 n=5+5)
Unmarshal/ReferenceFile/map-2        79.6µs ± 0%    79.1µs ± 0%  -0.62%  (p=0.008 n=5+5)
Unmarshal/HugoFrontMatter-2          13.7µs ± 0%    13.5µs ± 0%  -0.91%  (p=0.008 n=5+5)

name                               old speed      new speed      delta
UnmarshalDataset/config-2          42.2MB/s ± 0%  42.7MB/s ± 0%  +1.10%  (p=0.029 n=4+4)
UnmarshalDataset/canada-2          35.7MB/s ± 1%  35.5MB/s ± 3%    ~     (p=1.000 n=5+5)
UnmarshalDataset/citm_catalog-2    22.6MB/s ± 1%  23.1MB/s ± 0%  +2.36%  (p=0.008 n=5+5)
UnmarshalDataset/twitter-2         40.6MB/s ± 2%  41.2MB/s ± 1%  +1.47%  (p=0.008 n=5+5)
UnmarshalDataset/code-2            24.9MB/s ± 0%  25.4MB/s ± 0%  +1.95%  (p=0.008 n=5+5)
UnmarshalDataset/example-2         46.0MB/s ± 0%  46.9MB/s ± 0%  +1.86%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/struct-2  18.8MB/s ± 1%  18.7MB/s ± 0%    ~     (p=0.651 n=5+5)
Unmarshal/SimpleDocument/map-2     12.6MB/s ± 0%  12.6MB/s ± 0%    ~     (p=0.087 n=5+5)
Unmarshal/ReferenceFile/struct-2    106MB/s ± 0%   106MB/s ± 0%    ~     (p=0.222 n=5+5)
Unmarshal/ReferenceFile/map-2      65.8MB/s ± 0%  66.2MB/s ± 0%  +0.63%  (p=0.008 n=5+5)
Unmarshal/HugoFrontMatter-2        40.0MB/s ± 0%  40.3MB/s ± 0%  +0.92%  (p=0.008 n=5+5)

name                               old alloc/op   new alloc/op   delta
UnmarshalDataset/config-2            5.85MB ± 0%    5.85MB ± 0%    ~     (p=1.000 n=5+5)
UnmarshalDataset/canada-2            75.2MB ± 0%    75.2MB ± 0%    ~     (p=1.000 n=5+5)
UnmarshalDataset/citm_catalog-2      35.0MB ± 0%    35.0MB ± 0%    ~     (p=0.841 n=5+5)
UnmarshalDataset/twitter-2           13.5MB ± 0%    13.5MB ± 0%    ~     (p=0.548 n=5+5)
UnmarshalDataset/code-2              22.0MB ± 0%    22.0MB ± 0%    ~     (p=0.738 n=5+5)
UnmarshalDataset/example-2            203kB ± 0%     203kB ± 0%    ~     (p=0.714 n=5+5)
Unmarshal/SimpleDocument/struct-2      709B ± 0%      709B ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/map-2       1.08kB ± 0%    1.08kB ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/struct-2     19.7kB ± 0%    19.7kB ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/map-2        37.0kB ± 0%    37.0kB ± 0%    ~     (p=0.333 n=4+5)
Unmarshal/HugoFrontMatter-2          7.22kB ± 0%    7.22kB ± 0%    ~     (all equal)

name                               old allocs/op  new allocs/op  delta
UnmarshalDataset/config-2              230k ± 0%      230k ± 0%    ~     (p=0.556 n=4+5)
UnmarshalDataset/canada-2              391k ± 0%      391k ± 0%    ~     (all equal)
UnmarshalDataset/citm_catalog-2        158k ± 0%      158k ± 0%    ~     (p=1.000 n=4+5)
UnmarshalDataset/twitter-2            54.7k ± 0%     54.7k ± 0%    ~     (p=1.000 n=4+5)
UnmarshalDataset/code-2               1.05M ± 0%     1.05M ± 0%    ~     (all equal)
UnmarshalDataset/example-2            1.28k ± 0%     1.28k ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/struct-2      8.00 ± 0%      8.00 ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/map-2         13.0 ± 0%      13.0 ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/struct-2        123 ± 0%       123 ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/map-2           590 ± 0%       590 ± 0%    ~     (all equal)
Unmarshal/HugoFrontMatter-2             130 ± 0%       130 ± 0%    ~     (all equal)
```

.dockerignore

@@ -0,0 +1,2 @@
+cmd/tomll/tomll
+cmd/tomljson/tomljson

.gitattributes

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

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,22 @@
+---
+name: Bug report
+about: Create a report to help us improve
+
+---
+
+**Describe the bug**
+A clear and concise description of what the bug is.
+
+**To Reproduce**
+Steps to reproduce the behavior. Including TOML files.
+
+**Expected behavior**
+A clear and concise description of what you expected to happen, if other than "should work".
+
+**Versions**
+ - go-toml: version (git sha)
+ - go: version
+ - operating system: e.g. macOS, Windows, Linux
+
+**Additional context**
+Add any other context about the problem here that you think may help to diagnose.

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,19 @@
+<!--
+
+Thank you for your pull request!
+
+Please read the Code changes section of the CONTRIBUTING.md file,
+and make sure you have followed the instructions.
+
+https://github.com/pelletier/go-toml/blob/v2/CONTRIBUTING.md#code-changes
+
+-->
+
+Explanation of what this pull request does.
+
+More detailed description of the decisions being made and the reasons why (if
+the patch is non-trivial).
+
+---
+
+Paste `benchstat` results here

.github/dependabot.yml

@@ -0,0 +1,17 @@
+version: 2
+updates:
+- package-ecosystem: gomod
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10
+- package-ecosystem: github-actions
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10
+- package-ecosystem: docker
+  directory: /
+  schedule:
+    interval: daily
+  open-pull-requests-limit: 10

.github/release.yml

@@ -0,0 +1,20 @@
+changelog:
+  exclude:
+    labels:
+      - build
+  categories:
+    - title: What's new
+      labels:
+        - feature
+    - title: Performance
+      labels:
+        - performance
+    - title: Fixed bugs
+      labels:
+        - bug
+    - title: Documentation
+      labels:
+        - doc
+    - title: Other changes
+      labels:
+        - "*"

.github/workflows/codeql-analysis.yml

@@ -0,0 +1,67 @@
+# For most projects, this workflow file will not need changing; you simply need
+# to commit it to your repository.
+#
+# You may wish to alter this file to override the set of languages analyzed,
+# or to provide custom queries or build logic.
+#
+# ******** NOTE ********
+# We have attempted to detect the languages in your repository. Please check
+# the `language` matrix defined below to confirm you have the correct set of
+# supported CodeQL languages.
+#
+name: "CodeQL"
+
+on:
+  push:
+    branches: [ master, v2 ]
+  pull_request:
+    # The branches below must be a subset of the branches above
+    branches: [ master ]
+  schedule:
+    - cron: '26 19 * * 0'
+
+jobs:
+  analyze:
+    name: Analyze
+    runs-on: ubuntu-latest
+
+    strategy:
+      fail-fast: false
+      matrix:
+        language: [ 'go' ]
+        # CodeQL supports [ 'cpp', 'csharp', 'go', 'java', 'javascript', 'python' ]
+        # Learn more:
+        # https://docs.github.com/en/free-pro-team@latest/github/finding-security-vulnerabilities-and-errors-in-your-code/configuring-code-scanning#changing-the-languages-that-are-analyzed
+
+    steps:
+    - name: Checkout repository
+      uses: actions/checkout@v2
+
+    # Initializes the CodeQL tools for scanning.
+    - name: Initialize CodeQL
+      uses: github/codeql-action/init@v1
+      with:
+        languages: ${{ matrix.language }}
+        # If you wish to specify custom queries, you can do so here or in a config file.
+        # By default, queries listed here will override any specified in a config file.
+        # Prefix the list here with "+" to use these queries and those in the config file.
+        # queries: ./path/to/local/query, your-org/your-repo/queries@main
+
+    # 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
+
+    # ℹ️ Command-line programs to run using the OS shell.
+    # 📚 https://git.io/JvXDl
+
+    # ✏️ If the Autobuild fails above, remove it and uncomment the following three lines
+    #    and modify them (or add more) to build your code if your project
+    #    uses a compiled language
+
+    #- run: |
+    #   make bootstrap
+    #   make release
+
+    - name: Perform CodeQL Analysis
+      uses: github/codeql-action/analyze@v1

.github/workflows/coverage.yml

@@ -0,0 +1,20 @@
+name: coverage
+on:
+  pull_request:
+    branches:
+      - v2
+
+jobs:
+  report:
+    runs-on: "ubuntu-latest"
+    name: report
+    steps:
+      - uses: actions/checkout@master
+        with:
+          fetch-depth: 0
+      - name: Setup go
+        uses: actions/setup-go@master
+        with:
+          go-version: 1.16
+      - name: Run tests with coverage
+        run: ./ci.sh coverage -d "${GITHUB_BASE_REF-HEAD}"

.github/workflows/workflow.yml

@@ -0,0 +1,25 @@
+name: test
+on:
+  push:
+    branches:
+      - v2
+  pull_request:
+    branches:
+      - v2
+
+jobs:
+  build:
+    strategy:
+      matrix:
+        os: [ 'ubuntu-latest', 'windows-latest', 'macos-latest']
+        go: [ '1.16', '1.17' ]
+    runs-on: ${{ matrix.os }}
+    name: ${{ matrix.go }}/${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@master
+      - name: Setup go ${{ matrix.go }}
+        uses: actions/setup-go@master
+        with:
+          go-version: ${{ matrix.go }}
+      - name: Run unit tests
+        run: go test -race ./...

.gitignore

@@ -0,0 +1,5 @@
+test_program/test_program_bin
+fuzz/
+cmd/tomll/tomll
+cmd/tomljson/tomljson
+cmd/tomltestgen/tomltestgen

.golangci.toml

@@ -0,0 +1,84 @@
+[service]
+golangci-lint-version =  "1.39.0"
+
+[linters-settings.wsl]
+allow-assign-and-anything = true
+
+[linters-settings.exhaustive]
+default-signifies-exhaustive = true
+
+[linters]
+disable-all = true
+enable = [
+    "asciicheck",
+    "bodyclose",
+    "cyclop",
+    "deadcode",
+    "depguard",
+    "dogsled",
+    "dupl",
+    "durationcheck",
+    "errcheck",
+    "errorlint",
+    "exhaustive",
+    # "exhaustivestruct",
+    "exportloopref",
+    "forbidigo",
+    # "forcetypeassert",
+    "funlen",
+    "gci",
+    # "gochecknoglobals",
+    "gochecknoinits",
+    "gocognit",
+    "goconst",
+    "gocritic",
+    "gocyclo",
+    "godot",
+    "godox",
+    # "goerr113",
+    "gofmt",
+    "gofumpt",
+    "goheader",
+    "goimports",
+    "golint",
+    "gomnd",
+    # "gomoddirectives",
+    "gomodguard",
+    "goprintffuncname",
+    "gosec",
+    "gosimple",
+    "govet",
+    # "ifshort",
+    "importas",
+    "ineffassign",
+    "lll",
+    "makezero",
+    "misspell",
+    "nakedret",
+    "nestif",
+    "nilerr",
+    # "nlreturn",
+    "noctx",
+    "nolintlint",
+    #"paralleltest",
+    "prealloc",
+    "predeclared",
+    "revive",
+    "rowserrcheck",
+    "sqlclosecheck",
+    "staticcheck",
+    "structcheck",
+    "stylecheck",
+    # "testpackage",
+    "thelper",
+    "tparallel",
+    "typecheck",
+    "unconvert",
+    "unparam",
+    "unused",
+    "varcheck",
+    "wastedassign",
+    "whitespace",
+    # "wrapcheck",
+    # "wsl"
+]

.travis.yml

@@ -1 +0,0 @@
-language: go

CONTRIBUTING.md

@@ -0,0 +1,182 @@
+# Contributing
+
+Thank you for your interest in go-toml! We appreciate you considering
+contributing to go-toml!
+
+The main goal is the project is to provide an easy-to-use and efficient TOML
+implementation for Go that gets the job done and gets out of your way – dealing
+with TOML is probably not the central piece of your project.
+
+As the single maintainer of go-toml, time is scarce. All help, big or small, is
+more than welcomed!
+
+## Ask questions
+
+Any question you may have, somebody else might have it too. Always feel free to
+ask them on the [discussion board][discussions]. We will try to answer them as
+clearly and quickly as possible, time permitting.
+
+Asking questions also helps us identify areas where the documentation needs
+improvement, or new features that weren't envisioned before. Sometimes, a
+seemingly innocent question leads to the fix of a bug. Don't hesitate and ask
+away!
+
+[discussions]: https://github.com/pelletier/go-toml/discussions
+
+## Improve the documentation
+
+The best way to share your knowledge and experience with go-toml is to improve
+the documentation. Fix a typo, clarify an interface, add an example, anything
+goes!
+
+The documentation is present in the [README][readme] and thorough the source
+code. On release, it gets updated on [pkg.go.dev][pkg.go.dev]. To make a change
+to the documentation, create a pull request with your proposed changes. For
+simple changes like that, the easiest way to go is probably the "Fork this
+project and edit the file" button on Github, displayed at the top right of the
+file. Unless it's a trivial change (for example a typo), provide a little bit of
+context in your pull request description or commit message.
+
+## Report a bug
+
+Found a bug! Sorry to hear that :(. Help us and other track them down and fix by
+reporting it. [File a new bug report][bug-report] on the [issues
+tracker][issues-tracker]. The template should provide enough guidance on what to
+include. When in doubt: add more details! By reducing ambiguity and providing
+more information, it decreases back and forth and saves everyone time.
+
+## Code changes
+
+Want to contribute a patch? Very happy to hear that!
+
+First, some high-level rules:
+
+- A short proposal with some POC code is better than a lengthy piece of text
+  with no code. Code speaks louder than words. That being said, bigger changes
+  should probably start with a [discussion][discussions].
+- No backward-incompatible patch will be accepted unless discussed. Sometimes
+  it's hard, but we try not to break people's programs unless we absolutely have
+  to.
+- If you are writing a new feature or extending an existing one, make sure to
+  write some documentation.
+- Bug fixes need to be accompanied with regression tests.
+- New code needs to be tested.
+- Your commit messages need to explain why the change is needed, even if already
+  included in the PR description.
+
+It does sound like a lot, but those best practices are here to save time overall
+and continuously improve the quality of the project, which is something everyone
+benefits from.
+
+### Get started
+
+The fairly standard code contribution process looks like that:
+
+1. [Fork the project][fork].
+2. Make your changes, commit on any branch you like.
+3. [Open up a pull request][pull-request]
+4. Review, potential ask for changes.
+5. Merge.
+
+Feel free to ask for help! You can create draft pull requests to gather
+some early feedback!
+
+### Run the tests
+
+You can run tests for go-toml using Go's test tool: `go test -race ./...`.
+
+During the pull request process, all tests will be ran on Linux, Windows, and
+MacOS on the last two versions of Go.
+
+However, given GitHub's new policy to _not_ run Actions on pull requests until a
+maintainer clicks on button, it is highly recommended that you run them locally
+as you make changes.
+
+### Check coverage
+
+We use `go tool cover` to compute test coverage. Most code editors have a way to
+run and display code coverage, but at the end of the day, we do this:
+
+```
+go test -covermode=atomic -coverprofile=coverage.out
+go tool cover -func=coverage.out
+```
+
+and verify that the overall percentage of tested code does not go down. This is
+a requirement. As a rule of thumb, all lines of code touched by your changes
+should be covered. On Unix you can use `./ci.sh coverage -d v2` to check if your
+code lowers the coverage.
+
+### Verify performance
+
+Go-toml aims to stay efficient. We rely on a set of scenarios executed with Go's
+builtin benchmark systems. Because of their noisy nature, containers provided by
+Github Actions cannot be reliably used for benchmarking. As a result, you are
+responsible for checking that your changes do not incur a performance penalty.
+You can run their following to execute benchmarks:
+
+```
+go test ./... -bench=. -count=10
+```
+
+Benchmark results should be compared against each other with
+[benchstat][benchstat]. Typical flow looks like this:
+
+1. On the `v2` branch, run `go test ./... -bench=. -count 10` and save output to
+   a file (for example `old.txt`).
+2. Make some code changes.
+3. Run `go test ....` again, and save the output to an other file (for example
+   `new.txt`).
+4. Run `benchstat old.txt new.txt` to check that time/op does not go up in any
+   test.
+
+On Unix you can use `./ci.sh benchmark -d v2` to verify how your code impacts
+performance.
+
+It is highly encouraged to add the benchstat results to your pull request
+description. Pull requests that lower performance will receive more scrutiny.
+
+[benchstat]: https://pkg.go.dev/golang.org/x/perf/cmd/benchstat
+
+### Style
+
+Try to look around and follow the same format and structure as the rest of the
+code. We enforce using `go fmt` on the whole code base.
+
+---
+
+## Maintainers-only
+
+### Merge pull request
+
+Checklist:
+
+- Passing CI.
+- Does not introduce backward-incompatible changes (unless discussed).
+- Has relevant doc changes.
+- Benchstat does not show performance regression.
+
+1. Merge using "squash and merge".
+2. Make sure to edit the commit message to keep all the useful information
+   nice and clean.
+3. Make sure the commit title is clear and contains the PR number (#123).
+
+### New release
+
+1. Go to [releases][releases]. Click on "X commits to master since this
+   release".
+2. Make note of all the changes. Look for backward incompatible changes,
+   new features, and bug fixes.
+3. Pick the new version using the above and semver.
+4. Create a [new release][new-release].
+5. Follow the same format as [1.1.0][release-110].
+
+[issues-tracker]: https://github.com/pelletier/go-toml/issues
+[bug-report]: https://github.com/pelletier/go-toml/issues/new?template=bug_report.md
+[pkg.go.dev]: https://pkg.go.dev/github.com/pelletier/go-toml
+[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

LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2013 - 2021 Thomas Pelletier, Eric Anderton
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,80 +1,427 @@
-# go-toml
+# go-toml v2
 
-Go library for the [TOML](https://github.com/mojombo/toml) format.
+Go library for the [TOML](https://toml.io/en/) format.
 
-This library supports TOML version
-[v0.1.0](https://github.com/mojombo/toml/blob/master/versions/toml-v0.1.0.md)
+This library supports [TOML v1.0.0](https://toml.io/en/v1.0.0).
 
-[![Build Status](https://travis-ci.org/pelletier/go-toml.png?branch=master)](https://travis-ci.org/pelletier/go-toml)
+## Development status
 
-## Import
+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.
 
-    import "github.com/pelletier/go-toml"
+If you do not need the advanced document editing features of v1, you are
+encouraged to try out this version.
 
-## Usage
+[👉 Roadmap for v2](https://github.com/pelletier/go-toml/discussions/506)
 
-Say you have a TOML file that looks like this:
+[🐞 Bug Reports](https://github.com/pelletier/go-toml/issues)
 
-```toml
-[postgres]
-user = "pelletier"
-password = "mypassword"
-```
-
-Read the username and password like this:
-
-```go
-import (
-    "fmt"
-    "github.com/pelletier/go-toml"
-)
-
-config, err := toml.LoadFile("config.toml")
-if err != nil {
-    fmt.Println("Error ", err.Error())
-} else {
-    // retrieve data directly
-    user := config.Get("postgres.user").(string)
-    password := config.Get("postgres.password").(string)
-
-    // or using an intermediate object
-    configTree := config.Get("postgres").(*toml.TomlTree)
-    user = configTree.Get("user").(string)
-    password = configTree.Get("password").(string)
-    fmt.Println("User is ", user, ". Password is ", password)
-}
-```
+[💬 Anything else](https://github.com/pelletier/go-toml/discussions)
 
 ## Documentation
 
-The documentation is available at
-[godoc.org](http://godoc.org/github.com/pelletier/go-toml).
+Full API, examples, and implementation notes are available in the Go documentation.
 
-## Contribute
+[![Go Reference](https://pkg.go.dev/badge/github.com/pelletier/go-toml/v2.svg)](https://pkg.go.dev/github.com/pelletier/go-toml/v2)
 
-Feel free to report bugs and patches using GitHub's pull requests system on
-[pelletier/go-toml](https://github.com/pelletier/go-toml). Any feedback would be
-much appreciated!
+## Import
 
+```go
+import "github.com/pelletier/go-toml/v2"
+```
+
+See [Modules](#Modules).
+
+## Features
+
+### Stdlib behavior
+
+As much as possible, this library is designed to behave similarly as the
+standard library's `encoding/json`.
+
+### Performance
+
+While go-toml favors usability, it is written with performance in mind. Most
+operations should not be shockingly slow. See [benchmarks](#benchmarks).
+
+### Strict mode
+
+`Decoder` can be set to "strict mode", which makes it error when some parts of
+the TOML document was not prevent in the target structure. This is a great way
+to check for typos. [See example in the documentation][strict].
+
+[strict]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#example-Decoder.SetStrict
+
+### Contextualized errors
+
+When decoding errors occur, go-toml returns [`DecodeError`][decode-err]), which
+contains a human readable contextualized version of the error. For example:
+
+```
+2| key1 = "value1"
+3| key2 = "missing2"
+ | ~~~~ missing field
+4| key3 = "missing3"
+5| key4 = "value4"
+```
+
+[decode-err]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#DecodeError
+
+### Local date and time support
+
+TOML supports native [local date/times][ldt]. It allows to represent a given
+date, time, or date-time without relation to a timezone or offset. To support
+this use-case, go-toml provides [`LocalDate`][tld], [`LocalTime`][tlt], and
+[`LocalDateTime`][tldt]. Those types can be transformed to and from `time.Time`,
+making them convenient yet unambiguous structures for their respective TOML
+representation.
+
+[ldt]: https://toml.io/en/v1.0.0#local-date-time
+[tld]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalDate
+[tlt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalTime
+[tldt]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#LocalDateTime
+
+## Getting started
+
+Given the following struct, let's see how to read it and write it as TOML:
+
+```go
+type MyConfig struct {
+      Version int
+      Name    string
+      Tags    []string
+}
+```
+
+### Unmarshaling
+
+[`Unmarshal`][unmarshal] reads a TOML document and fills a Go structure with its
+content. For example:
+
+```go
+doc := `
+version = 2
+name = "go-toml"
+tags = ["go", "toml"]
+`
+
+var cfg MyConfig
+err := toml.Unmarshal([]byte(doc), &cfg)
+if err != nil {
+      panic(err)
+}
+fmt.Println("version:", cfg.Version)
+fmt.Println("name:", cfg.Name)
+fmt.Println("tags:", cfg.Tags)
+
+// Output:
+// version: 2
+// name: go-toml
+// tags: [go toml]
+```
+
+[unmarshal]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Unmarshal
+
+### Marshaling
+
+[`Marshal`][marshal] is the opposite of Unmarshal: it represents a Go structure
+as a TOML document:
+
+```go
+cfg := MyConfig{
+      Version: 2,
+      Name:    "go-toml",
+      Tags:    []string{"go", "toml"},
+}
+
+b, err := toml.Marshal(cfg)
+if err != nil {
+      panic(err)
+}
+fmt.Println(string(b))
+
+// Output:
+// Version = 2
+// Name = 'go-toml'
+// Tags = ['go', 'toml']
+```
+
+[marshal]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Marshal
+
+## Benchmarks
+
+Execution time speedup compared to other Go TOML libraries:
+
+<table>
+    <thead>
+        <tr><th>Benchmark</th><th>go-toml v1</th><th>BurntSushi/toml</th></tr>
+    </thead>
+    <tbody>
+        <tr><td>Marshal/HugoFrontMatter-2</td><td>1.9x</td><td>1.9x</td></tr>
+        <tr><td>Marshal/ReferenceFile/map-2</td><td>1.7x</td><td>1.9x</td></tr>
+        <tr><td>Marshal/ReferenceFile/struct-2</td><td>2.4x</td><td>2.6x</td></tr>
+        <tr><td>Unmarshal/HugoFrontMatter-2</td><td>2.9x</td><td>2.5x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/map-2</td><td>2.7x</td><td>2.6x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/struct-2</td><td>4.8x</td><td>5.1x</td></tr>
+     </tbody>
+</table>
+<details><summary>See more</summary>
+<p>The table above has the results of the most common use-cases. The table below
+contains the results of all benchmarks, including unrealistic ones. It is
+provided for completeness.</p>
+
+<table>
+    <thead>
+        <tr><th>Benchmark</th><th>go-toml v1</th><th>BurntSushi/toml</th></tr>
+    </thead>
+    <tbody>
+        <tr><td>Marshal/SimpleDocument/map-2</td><td>1.7x</td><td>2.1x</td></tr>
+        <tr><td>Marshal/SimpleDocument/struct-2</td><td>2.5x</td><td>2.8x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/map-2</td><td>4.1x</td><td>3.1x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/struct-2</td><td>6.4x</td><td>4.3x</td></tr>
+        <tr><td>UnmarshalDataset/example-2</td><td>3.4x</td><td>3.2x</td></tr>
+        <tr><td>UnmarshalDataset/code-2</td><td>2.2x</td><td>2.5x</td></tr>
+        <tr><td>UnmarshalDataset/twitter-2</td><td>2.8x</td><td>2.7x</td></tr>
+        <tr><td>UnmarshalDataset/citm_catalog-2</td><td>2.2x</td><td>2.0x</td></tr>
+        <tr><td>UnmarshalDataset/canada-2</td><td>1.8x</td><td>1.4x</td></tr>
+        <tr><td>UnmarshalDataset/config-2</td><td>4.4x</td><td>2.9x</td></tr>
+        <tr><td>[Geo mean]</td><td>2.8x</td><td>2.6x</td></tr>
+     </tbody>
+</table>
+<p>This table can be generated with <code>./ci.sh benchmark -a -html</code>.</p>
+</details>
+
+## Modules
+
+go-toml uses Go's standard modules system.
+
+Installation instructions:
+
+- Go ≥ 1.16: Nothing to do. Use the import in your code. The `go` command deals
+  with it automatically.
+- Go ≥ 1.13: `GO111MODULE=on go get github.com/pelletier/go-toml/v2`.
+
+In case of trouble: [Go Modules FAQ][mod-faq].
+
+[mod-faq]: https://github.com/golang/go/wiki/Modules#why-does-installing-a-tool-via-go-get-fail-with-error-cannot-find-main-module
+
+## Migrating from v1
+
+This section describes the differences between v1 and v2, with some pointers on
+how to get the original behavior when possible.
+
+### Decoding / Unmarshal
+
+#### Automatic field name guessing
+
+When unmarshaling to a struct, if a key in the TOML document does not exactly
+match the name of a struct field or any of the `toml`-tagged field, v1 tries
+multiple variations of the key ([code][v1-keys]).
+
+V2 instead does a case-insensitive matching, like `encoding/json`.
+
+This could impact you if you are relying on casing to differentiate two fields,
+and one of them is a not using the `toml` struct tag. The recommended solution
+is to be specific about tag names for those fields using the `toml` struct tag.
+
+[v1-keys]: https://github.com/pelletier/go-toml/blob/a2e52561804c6cd9392ebf0048ca64fe4af67a43/marshal.go#L775-L781
+
+#### Ignore preexisting value in interface
+
+When decoding into a non-nil `interface{}`, go-toml v1 uses the type of the
+element in the interface to decode the object. For example:
+
+```go
+type inner struct {
+  B interface{}
+}
+type doc struct {
+  A interface{}
+}
+
+d := doc{
+  A: inner{
+    B: "Before",
+  },
+}
+
+data := `
+[A]
+B = "After"
+`
+
+toml.Unmarshal([]byte(data), &d)
+fmt.Printf("toml v1: %#v\n", d)
+
+// toml v1: main.doc{A:main.inner{B:"After"}}
+```
+
+In this case, field `A` is of type `interface{}`, containing a `inner` struct.
+V1 sees that type and uses it when decoding the object.
+
+When decoding an object into an `interface{}`, V2 instead disregards whatever
+value the `interface{}` may contain and replaces it with a
+`map[string]interface{}`. With the same data structure as above, here is what
+the result looks like:
+
+```go
+toml.Unmarshal([]byte(data), &d)
+fmt.Printf("toml v2: %#v\n", d)
+
+// toml v2: main.doc{A:map[string]interface {}{"B":"After"}}
+```
+
+This is to match `encoding/json`'s behavior. There is no way to make the v2
+decoder behave like v1.
+
+#### Values out of array bounds ignored
+
+When decoding into an array, v1 returns an error when the number of elements
+contained in the doc is superior to the capacity of the array. For example:
+
+```go
+type doc struct {
+  A [2]string
+}
+d := doc{}
+err := toml.Unmarshal([]byte(`A = ["one", "two", "many"]`), &d)
+fmt.Println(err)
+
+// (1, 1): unmarshal: TOML array length (3) exceeds destination array length (2)
+```
+
+In the same situation, v2 ignores the last value:
+
+```go
+err := toml.Unmarshal([]byte(`A = ["one", "two", "many"]`), &d)
+fmt.Println("err:", err, "d:", d)
+// err: <nil> d: {[one two]}
+```
+
+This is to match `encoding/json`'s behavior. There is no way to make the v2
+decoder behave like v1.
+
+#### Support for `toml.Unmarshaler` has been dropped
+
+This method was not widely used, poorly defined, and added a lot of complexity.
+A similar effect can be achieved by implementing the `encoding.TextUnmarshaler`
+interface and use strings.
+
+#### Support for `default` struct tag has been dropped
+
+This feature adds complexity and a poorly defined API for an effect that can be
+accomplished outside of the library.
+
+It does not seem like other format parsers in Go support that feature (the
+project referenced in the original ticket #202 has not been updated since 2017).
+Given that go-toml v2 should not touch values not in the document, the same
+effect can be achieved by pre-filling the struct with defaults (libraries like
+[go-defaults][go-defaults] can help). Also, string representation is not well
+defined for all types: it creates issues like #278.
+
+The recommended replacement is pre-filling the struct before unmarshaling.
+
+[go-defaults]: https://github.com/mcuadros/go-defaults
+
+### Encoding / Marshal
+
+#### Default struct fields order
+
+V1 emits struct fields order alphabetically by default. V2 struct fields are
+emitted in order they are defined. For example:
+
+```go
+type S struct {
+	B string
+	A string
+}
+
+data := S{
+	B: "B",
+	A: "A",
+}
+
+b, _ := tomlv1.Marshal(data)
+fmt.Println("v1:\n" + string(b))
+
+b, _ = tomlv2.Marshal(data)
+fmt.Println("v2:\n" + string(b))
+
+// Output:
+// v1:
+// A = "A"
+// B = "B"
+
+// v2:
+// B = 'B'
+// A = 'A'
+```
+
+There is no way to make v2 encoder behave like v1. A workaround could be to
+manually sort the fields alphabetically in the struct definition.
+
+#### No indentation by default
+
+V1 automatically indents content of tables by default. V2 does not. However the
+same behavior can be obtained using [`Encoder.SetIndentTables`][sit]. For example:
+
+```go
+data := map[string]interface{}{
+	"table": map[string]string{
+		"key": "value",
+	},
+}
+
+b, _ := tomlv1.Marshal(data)
+fmt.Println("v1:\n" + string(b))
+
+b, _ = tomlv2.Marshal(data)
+fmt.Println("v2:\n" + string(b))
+
+buf := bytes.Buffer{}
+enc := tomlv2.NewEncoder(&buf)
+enc.SetIndentTables(true)
+enc.Encode(data)
+fmt.Println("v2 Encoder:\n" + string(buf.Bytes()))
+
+// Output:
+// v1:
+//
+// [table]
+//   key = "value"
+//
+// v2:
+// [table]
+// key = 'value'
+//
+//
+// v2 Encoder:
+// [table]
+//   key = 'value'
+```
+
+[sit]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#Encoder.SetIndentTables
+
+#### Keys and strings are single quoted
+
+V1 always uses double quotes (`"`) around strings and keys that cannot be
+represented bare (unquoted). V2 uses single quotes instead by default (`'`),
+unless a character cannot be represented, then falls back to double quotes.
+
+There is no way to make v2 encoder behave like v1.
+
+#### `TextMarshaler` emits as a string, not TOML
+
+Types that implement [`encoding.TextMarshaler`][tm] can emit arbitrary TOML in
+v1. The encoder would append the result to the output directly. In v2 the result
+is wrapped in a string. As a result, this interface cannot be implemented by the
+root object.
+
+There is no way to make v2 encoder behave like v1.
+
+[tm]: https://golang.org/pkg/encoding/#TextMarshaler
 
 ## License
 
-Copyright (c) 2013 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 in
-the Software without restriction, including without limitation the rights to
-use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
-of the Software, and to permit persons to whom the Software is furnished to do
-so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+The MIT License (MIT). Read [LICENSE](LICENSE).

SECURITY.md

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

benchmark/bench_datasets_test.go

@@ -0,0 +1,80 @@
+package benchmark_test
+
+import (
+	"compress/gzip"
+	"encoding/json"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+var bench_inputs = []struct {
+	name    string
+	jsonLen int
+}{
+	// from https://gist.githubusercontent.com/feeeper/2197d6d734729625a037af1df14cf2aa/raw/2f22b120e476d897179be3c1e2483d18067aa7df/config.toml
+	{"config", 806507},
+
+	// converted from https://github.com/miloyip/nativejson-benchmark
+	{"canada", 2090234},
+	{"citm_catalog", 479897},
+	{"twitter", 428778},
+	{"code", 1940472},
+
+	// converted from https://raw.githubusercontent.com/mailru/easyjson/master/benchmark/example.json
+	{"example", 7779},
+}
+
+func TestUnmarshalDatasetCode(t *testing.T) {
+	for _, tc := range bench_inputs {
+		t.Run(tc.name, func(t *testing.T) {
+			buf := fixture(t, tc.name)
+
+			var v interface{}
+			require.NoError(t, toml.Unmarshal(buf, &v))
+
+			b, err := json.Marshal(v)
+			require.NoError(t, err)
+			require.Equal(t, len(b), tc.jsonLen)
+		})
+	}
+}
+
+func BenchmarkUnmarshalDataset(b *testing.B) {
+	for _, tc := range bench_inputs {
+		b.Run(tc.name, func(b *testing.B) {
+			buf := fixture(b, tc.name)
+			b.SetBytes(int64(len(buf)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				var v interface{}
+				require.NoError(b, toml.Unmarshal(buf, &v))
+			}
+		})
+	}
+}
+
+// fixture returns the uncompressed contents of path.
+func fixture(tb testing.TB, path string) []byte {
+	tb.Helper()
+
+	file := path + ".toml.gz"
+	f, err := os.Open(filepath.Join("testdata", file))
+	if os.IsNotExist(err) {
+		tb.Skip("benchmark fixture not found:", file)
+	}
+	require.NoError(tb, err)
+	defer f.Close()
+
+	gz, err := gzip.NewReader(f)
+	require.NoError(tb, err)
+
+	buf, err := ioutil.ReadAll(gz)
+	require.NoError(tb, err)
+	return buf
+}

benchmark/benchmark.toml

@@ -0,0 +1,244 @@
+################################################################################
+## Comment
+
+# Speak your mind with the hash symbol. They go from the symbol to the end of
+# the line.
+
+
+################################################################################
+## Table
+
+# Tables (also known as hash tables or dictionaries) are collections of
+# key/value pairs. They appear in square brackets on a line by themselves.
+
+[table]
+
+key = "value" # Yeah, you can do this.
+
+# Nested tables are denoted by table names with dots in them. Name your tables
+# whatever crap you please, just don't use #, ., [ or ].
+
+[table.subtable]
+
+key = "another value"
+
+# You don't need to specify all the super-tables if you don't want to. TOML
+# knows how to do it for you.
+
+# [x] you
+# [x.y] don't
+# [x.y.z] need these
+[x.y.z.w] # for this to work
+
+
+################################################################################
+## Inline Table
+
+# Inline tables provide a more compact syntax for expressing tables. They are
+# especially useful for grouped data that can otherwise quickly become verbose.
+# Inline tables are enclosed in curly braces `{` and `}`. No newlines are
+# allowed between the curly braces unless they are valid within a value.
+
+[table.inline]
+
+name = { first = "Tom", last = "Preston-Werner" }
+point = { x = 1, y = 2 }
+
+
+################################################################################
+## String
+
+# There are four ways to express strings: basic, multi-line basic, literal, and
+# multi-line literal. All strings must contain only valid UTF-8 characters.
+
+[string.basic]
+
+basic = "I'm a string. \"You can quote me\". Name\tJos\u00E9\nLocation\tSF."
+
+[string.multiline]
+
+# The following strings are byte-for-byte equivalent:
+key1 = "One\nTwo"
+key2 = """One\nTwo"""
+key3 = """
+One
+Two"""
+
+[string.multiline.continued]
+
+# The following strings are byte-for-byte equivalent:
+key1 = "The quick brown fox jumps over the lazy dog."
+
+key2 = """
+The quick brown \
+
+
+  fox jumps over \
+    the lazy dog."""
+
+key3 = """\
+       The quick brown \
+       fox jumps over \
+       the lazy dog.\
+       """
+
+[string.literal]
+
+# What you see is what you get.
+winpath  = 'C:\Users\nodejs\templates'
+winpath2 = '\\ServerX\admin$\system32\'
+quoted   = 'Tom "Dubs" Preston-Werner'
+regex    = '<\i\c*\s*>'
+
+
+[string.literal.multiline]
+
+regex2 = '''I [dw]on't need \d{2} apples'''
+lines  = '''
+The first newline is
+trimmed in raw strings.
+   All other whitespace
+   is preserved.
+'''
+
+
+################################################################################
+## Integer
+
+# Integers are whole numbers. Positive numbers may be prefixed with a plus sign.
+# Negative numbers are prefixed with a minus sign.
+
+[integer]
+
+key1 = +99
+key2 = 42
+key3 = 0
+key4 = -17
+
+[integer.underscores]
+
+# For large numbers, you may use underscores to enhance readability. Each
+# underscore must be surrounded by at least one digit.
+key1 = 1_000
+key2 = 5_349_221
+key3 = 1_2_3_4_5     # valid but inadvisable
+
+
+################################################################################
+## Float
+
+# A float consists of an integer part (which may be prefixed with a plus or
+# minus sign) followed by a fractional part and/or an exponent part.
+
+[float.fractional]
+
+key1 = +1.0
+key2 = 3.1415
+key3 = -0.01
+
+[float.exponent]
+
+key1 = 5e+22
+key2 = 1e6
+key3 = -2E-2
+
+[float.both]
+
+key = 6.626e-34
+
+[float.underscores]
+
+key1 = 9_224_617.445_991_228_313
+key2 = 1e1_00
+
+
+################################################################################
+## Boolean
+
+# Booleans are just the tokens you're used to. Always lowercase.
+
+[boolean]
+
+True = true
+False = false
+
+
+################################################################################
+## Datetime
+
+# Datetimes are RFC 3339 dates.
+
+[datetime]
+
+key1 = 1979-05-27T07:32:00Z
+key2 = 1979-05-27T00:32:00-07:00
+key3 = 1979-05-27T00:32:00.999999-07:00
+
+
+################################################################################
+## Array
+
+# Arrays are square brackets with other primitives inside. Whitespace is
+# ignored. Elements are separated by commas. Data types may not be mixed.
+
+[array]
+
+key1 = [ 1, 2, 3 ]
+key2 = [ "red", "yellow", "green" ]
+key3 = [ [ 1, 2 ], [3, 4, 5] ]
+key4 = [ [ 1, 2 ], ["a", "b", "c"] ] # this is ok
+
+# Arrays can also be multiline. So in addition to ignoring whitespace, arrays
+# also ignore newlines between the brackets.  Terminating commas are ok before
+# the closing bracket.
+
+key5 = [
+  1, 2, 3
+]
+key6 = [
+  1,
+  2, # this is ok
+]
+
+
+################################################################################
+## Array of Tables
+
+# These can be expressed by using a table name in double brackets. Each table
+# with the same double bracketed name will be an element in the array. The
+# tables are inserted in the order encountered.
+
+[[products]]
+
+name = "Hammer"
+sku = 738594937
+
+[[products]]
+
+[[products]]
+
+name = "Nail"
+sku = 284758393
+color = "gray"
+
+
+# You can create nested arrays of tables as well.
+
+[[fruit]]
+  name = "apple"
+
+  [fruit.physical]
+    color = "red"
+    shape = "round"
+
+  [[fruit.variety]]
+    name = "red delicious"
+
+  [[fruit.variety]]
+    name = "granny smith"
+
+[[fruit]]
+  name = "banana"
+
+  [[fruit.variety]]
+    name = "plantain"

benchmark/benchmark_test.go

@@ -0,0 +1,653 @@
+package benchmark_test
+
+import (
+	"bytes"
+	"io/ioutil"
+	"testing"
+	"time"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+func TestUnmarshalSimple(t *testing.T) {
+	doc := []byte(`A = "hello"`)
+	d := struct {
+		A string
+	}{}
+
+	err := toml.Unmarshal(doc, &d)
+	if err != nil {
+		panic(err)
+	}
+}
+
+func BenchmarkUnmarshal(b *testing.B) {
+	b.Run("SimpleDocument", func(b *testing.B) {
+		doc := []byte(`A = "hello"`)
+
+		b.Run("struct", func(b *testing.B) {
+			b.SetBytes(int64(len(doc)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				d := struct {
+					A string
+				}{}
+
+				err := toml.Unmarshal(doc, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+
+		b.Run("map", func(b *testing.B) {
+			b.SetBytes(int64(len(doc)))
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			for i := 0; i < b.N; i++ {
+				d := map[string]interface{}{}
+				err := toml.Unmarshal(doc, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+	})
+
+	b.Run("ReferenceFile", func(b *testing.B) {
+		bytes, err := ioutil.ReadFile("benchmark.toml")
+		if err != nil {
+			b.Fatal(err)
+		}
+
+		b.Run("struct", func(b *testing.B) {
+			b.SetBytes(int64(len(bytes)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				d := benchmarkDoc{}
+				err := toml.Unmarshal(bytes, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+
+		b.Run("map", func(b *testing.B) {
+			b.SetBytes(int64(len(bytes)))
+			b.ReportAllocs()
+			b.ResetTimer()
+			for i := 0; i < b.N; i++ {
+				d := map[string]interface{}{}
+				err := toml.Unmarshal(bytes, &d)
+				if err != nil {
+					panic(err)
+				}
+			}
+		})
+	})
+
+	b.Run("HugoFrontMatter", func(b *testing.B) {
+		b.SetBytes(int64(len(hugoFrontMatterbytes)))
+		b.ReportAllocs()
+		b.ResetTimer()
+		for i := 0; i < b.N; i++ {
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(hugoFrontMatterbytes, &d)
+			if err != nil {
+				panic(err)
+			}
+		}
+	})
+}
+
+func marshal(v interface{}) ([]byte, error) {
+	var b bytes.Buffer
+	enc := toml.NewEncoder(&b)
+	err := enc.Encode(v)
+	return b.Bytes(), err
+}
+
+func BenchmarkMarshal(b *testing.B) {
+	b.Run("SimpleDocument", func(b *testing.B) {
+		doc := []byte(`A = "hello"`)
+
+		b.Run("struct", func(b *testing.B) {
+			d := struct {
+				A string
+			}{}
+
+			err := toml.Unmarshal(doc, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+
+		b.Run("map", func(b *testing.B) {
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(doc, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+	})
+
+	b.Run("ReferenceFile", func(b *testing.B) {
+		bytes, err := ioutil.ReadFile("benchmark.toml")
+		if err != nil {
+			b.Fatal(err)
+		}
+
+		b.Run("struct", func(b *testing.B) {
+			d := benchmarkDoc{}
+			err := toml.Unmarshal(bytes, &d)
+			if err != nil {
+				panic(err)
+			}
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+
+		b.Run("map", func(b *testing.B) {
+			d := map[string]interface{}{}
+			err := toml.Unmarshal(bytes, &d)
+			if err != nil {
+				panic(err)
+			}
+
+			b.ReportAllocs()
+			b.ResetTimer()
+
+			var out []byte
+			for i := 0; i < b.N; i++ {
+				out, err = marshal(d)
+				if err != nil {
+					panic(err)
+				}
+			}
+
+			b.SetBytes(int64(len(out)))
+		})
+	})
+
+	b.Run("HugoFrontMatter", func(b *testing.B) {
+		d := map[string]interface{}{}
+		err := toml.Unmarshal(hugoFrontMatterbytes, &d)
+		if err != nil {
+			panic(err)
+		}
+
+		b.ReportAllocs()
+		b.ResetTimer()
+
+		var out []byte
+
+		for i := 0; i < b.N; i++ {
+			out, err = marshal(d)
+			if err != nil {
+				panic(err)
+			}
+		}
+
+		b.SetBytes(int64(len(out)))
+	})
+}
+
+type benchmarkDoc struct {
+	Table struct {
+		Key      string
+		Subtable struct {
+			Key string
+		}
+		Inline struct {
+			Name struct {
+				First string
+				Last  string
+			}
+			Point struct {
+				X int64
+				Y int64
+			}
+		}
+	}
+	String struct {
+		Basic struct {
+			Basic string
+		}
+		Multiline struct {
+			Key1      string
+			Key2      string
+			Key3      string
+			Continued struct {
+				Key1 string
+				Key2 string
+				Key3 string
+			}
+		}
+		Literal struct {
+			Winpath   string
+			Winpath2  string
+			Quoted    string
+			Regex     string
+			Multiline struct {
+				Regex2 string
+				Lines  string
+			}
+		}
+	}
+	Integer struct {
+		Key1        int64
+		Key2        int64
+		Key3        int64
+		Key4        int64
+		Underscores struct {
+			Key1 int64
+			Key2 int64
+			Key3 int64
+		}
+	}
+	Float struct {
+		Fractional struct {
+			Key1 float64
+			Key2 float64
+			Key3 float64
+		}
+		Exponent struct {
+			Key1 float64
+			Key2 float64
+			Key3 float64
+		}
+		Both struct {
+			Key float64
+		}
+		Underscores struct {
+			Key1 float64
+			Key2 float64
+		}
+	}
+	Boolean struct {
+		True  bool
+		False bool
+	}
+	Datetime struct {
+		Key1 time.Time
+		Key2 time.Time
+		Key3 time.Time
+	}
+	Array struct {
+		Key1 []int64
+		Key2 []string
+		Key3 [][]int64
+		Key4 []interface{}
+		Key5 []int64
+		Key6 []int64
+	}
+	Products []struct {
+		Name  string
+		Sku   int64
+		Color string
+	}
+	Fruit []struct {
+		Name     string
+		Physical struct {
+			Color string
+			Shape string
+		}
+		Variety []struct {
+			Name string
+		}
+	}
+}
+
+func TestUnmarshalReferenceFile(t *testing.T) {
+	bytes, err := ioutil.ReadFile("benchmark.toml")
+	require.NoError(t, err)
+	d := benchmarkDoc{}
+	err = toml.Unmarshal(bytes, &d)
+	require.NoError(t, err)
+
+	expected := benchmarkDoc{
+		Table: struct {
+			Key      string
+			Subtable struct{ Key string }
+			Inline   struct {
+				Name struct {
+					First string
+					Last  string
+				}
+				Point struct {
+					X int64
+					Y int64
+				}
+			}
+		}{
+			Key: "value",
+			Subtable: struct{ Key string }{
+				Key: "another value",
+			},
+			// note: x.y.z.w is purposefully missing
+			Inline: struct {
+				Name struct {
+					First string
+					Last  string
+				}
+				Point struct {
+					X int64
+					Y int64
+				}
+			}{
+				Name: struct {
+					First string
+					Last  string
+				}{
+					First: "Tom",
+					Last:  "Preston-Werner",
+				},
+				Point: struct {
+					X int64
+					Y int64
+				}{
+					X: 1,
+					Y: 2,
+				},
+			},
+		},
+		String: struct {
+			Basic     struct{ Basic string }
+			Multiline struct {
+				Key1      string
+				Key2      string
+				Key3      string
+				Continued struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}
+			}
+			Literal struct {
+				Winpath   string
+				Winpath2  string
+				Quoted    string
+				Regex     string
+				Multiline struct {
+					Regex2 string
+					Lines  string
+				}
+			}
+		}{
+			Basic: struct{ Basic string }{
+				Basic: "I'm a string. \"You can quote me\". Name\tJos\u00E9\nLocation\tSF.",
+			},
+			Multiline: struct {
+				Key1      string
+				Key2      string
+				Key3      string
+				Continued struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}
+			}{
+				Key1: "One\nTwo",
+				Key2: "One\nTwo",
+				Key3: "One\nTwo",
+
+				Continued: struct {
+					Key1 string
+					Key2 string
+					Key3 string
+				}{
+					Key1: `The quick brown fox jumps over the lazy dog.`,
+					Key2: `The quick brown fox jumps over the lazy dog.`,
+					Key3: `The quick brown fox jumps over the lazy dog.`,
+				},
+			},
+			Literal: struct {
+				Winpath   string
+				Winpath2  string
+				Quoted    string
+				Regex     string
+				Multiline struct {
+					Regex2 string
+					Lines  string
+				}
+			}{
+				Winpath:  `C:\Users\nodejs\templates`,
+				Winpath2: `\\ServerX\admin$\system32\`,
+				Quoted:   `Tom "Dubs" Preston-Werner`,
+				Regex:    `<\i\c*\s*>`,
+
+				Multiline: struct {
+					Regex2 string
+					Lines  string
+				}{
+					Regex2: `I [dw]on't need \d{2} apples`,
+					Lines: `The first newline is
+trimmed in raw strings.
+   All other whitespace
+   is preserved.
+`,
+				},
+			},
+		},
+		Integer: struct {
+			Key1        int64
+			Key2        int64
+			Key3        int64
+			Key4        int64
+			Underscores struct {
+				Key1 int64
+				Key2 int64
+				Key3 int64
+			}
+		}{
+			Key1: 99,
+			Key2: 42,
+			Key3: 0,
+			Key4: -17,
+
+			Underscores: struct {
+				Key1 int64
+				Key2 int64
+				Key3 int64
+			}{
+				Key1: 1000,
+				Key2: 5349221,
+				Key3: 12345,
+			},
+		},
+		Float: struct {
+			Fractional struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}
+			Exponent struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}
+			Both        struct{ Key float64 }
+			Underscores struct {
+				Key1 float64
+				Key2 float64
+			}
+		}{
+			Fractional: struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}{
+				Key1: 1.0,
+				Key2: 3.1415,
+				Key3: -0.01,
+			},
+			Exponent: struct {
+				Key1 float64
+				Key2 float64
+				Key3 float64
+			}{
+				Key1: 5e+22,
+				Key2: 1e6,
+				Key3: -2e-2,
+			},
+			Both: struct{ Key float64 }{
+				Key: 6.626e-34,
+			},
+			Underscores: struct {
+				Key1 float64
+				Key2 float64
+			}{
+				Key1: 9224617.445991228313,
+				Key2: 1e100,
+			},
+		},
+		Boolean: struct {
+			True  bool
+			False bool
+		}{
+			True:  true,
+			False: false,
+		},
+		Datetime: struct {
+			Key1 time.Time
+			Key2 time.Time
+			Key3 time.Time
+		}{
+			Key1: time.Date(1979, 5, 27, 7, 32, 0, 0, time.UTC),
+			Key2: time.Date(1979, 5, 27, 0, 32, 0, 0, time.FixedZone("", -7*3600)),
+			Key3: time.Date(1979, 5, 27, 0, 32, 0, 999999000, time.FixedZone("", -7*3600)),
+		},
+		Array: struct {
+			Key1 []int64
+			Key2 []string
+			Key3 [][]int64
+			Key4 []interface{}
+			Key5 []int64
+			Key6 []int64
+		}{
+			Key1: []int64{1, 2, 3},
+			Key2: []string{"red", "yellow", "green"},
+			Key3: [][]int64{{1, 2}, {3, 4, 5}},
+			Key4: []interface{}{
+				[]interface{}{int64(1), int64(2)},
+				[]interface{}{"a", "b", "c"},
+			},
+			Key5: []int64{1, 2, 3},
+			Key6: []int64{1, 2},
+		},
+		Products: []struct {
+			Name  string
+			Sku   int64
+			Color string
+		}{
+			{
+				Name: "Hammer",
+				Sku:  738594937,
+			},
+			{},
+			{
+				Name:  "Nail",
+				Sku:   284758393,
+				Color: "gray",
+			},
+		},
+		Fruit: []struct {
+			Name     string
+			Physical struct {
+				Color string
+				Shape string
+			}
+			Variety []struct{ Name string }
+		}{
+			{
+				Name: "apple",
+				Physical: struct {
+					Color string
+					Shape string
+				}{
+					Color: "red",
+					Shape: "round",
+				},
+				Variety: []struct{ Name string }{
+					{Name: "red delicious"},
+					{Name: "granny smith"},
+				},
+			},
+			{
+				Name: "banana",
+				Variety: []struct{ Name string }{
+					{Name: "plantain"},
+				},
+			},
+		},
+	}
+
+	require.Equal(t, expected, d)
+}
+
+var hugoFrontMatterbytes = []byte(`
+categories = ["Development", "VIM"]
+date = "2012-04-06"
+description = "spf13-vim is a cross platform distribution of vim plugins and resources for Vim."
+slug = "spf13-vim-3-0-release-and-new-website"
+tags = [".vimrc", "plugins", "spf13-vim", "vim"]
+title = "spf13-vim 3.0 release and new website"
+include_toc = true
+show_comments = false
+
+[[cascade]]
+  background = "yosemite.jpg"
+  [cascade._target]
+    kind = "page"
+    lang = "en"
+    path = "/blog/**"
+
+[[cascade]]
+  background = "goldenbridge.jpg"
+  [cascade._target]
+    kind = "section"
+`)

benchmark_test.go

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

ci.sh

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

cmd/gotoml-test-decoder/main.go

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

cmd/tomltestgen/main.go

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

decode.go

@@ -0,0 +1,496 @@
+package toml
+
+import (
+	"fmt"
+	"math"
+	"strconv"
+	"time"
+)
+
+func parseInteger(b []byte) (int64, error) {
+	if len(b) > 2 && b[0] == '0' {
+		switch b[1] {
+		case 'x':
+			return parseIntHex(b)
+		case 'b':
+			return parseIntBin(b)
+		case 'o':
+			return parseIntOct(b)
+		default:
+			panic(fmt.Errorf("invalid base '%c', should have been checked by scanIntOrFloat", b[1]))
+		}
+	}
+
+	return parseIntDec(b)
+}
+
+func parseLocalDate(b []byte) (LocalDate, error) {
+	// full-date      = date-fullyear "-" date-month "-" date-mday
+	// date-fullyear  = 4DIGIT
+	// date-month     = 2DIGIT  ; 01-12
+	// date-mday      = 2DIGIT  ; 01-28, 01-29, 01-30, 01-31 based on month/year
+	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")
+	}
+
+	var err error
+
+	date.Year, err = parseDecimalDigits(b[0:4])
+	if err != nil {
+		return LocalDate{}, err
+	}
+
+	date.Month, err = parseDecimalDigits(b[5:7])
+	if err != nil {
+		return LocalDate{}, err
+	}
+
+	date.Day, err = parseDecimalDigits(b[8:10])
+	if err != nil {
+		return LocalDate{}, err
+	}
+
+	if !isValidDate(date.Year, date.Month, date.Day) {
+		return LocalDate{}, newDecodeError(b, "impossible date")
+	}
+
+	return date, nil
+}
+
+func parseDecimalDigits(b []byte) (int, error) {
+	v := 0
+
+	for i, c := range b {
+		if c < '0' || c > '9' {
+			return 0, newDecodeError(b[i:i+1], "expected digit (0-9)")
+		}
+		v *= 10
+		v += int(c - '0')
+	}
+
+	return v, nil
+}
+
+func parseDateTime(b []byte) (time.Time, error) {
+	// offset-date-time = full-date time-delim full-time
+	// full-time      = partial-time time-offset
+	// time-offset    = "Z" / time-numoffset
+	// time-numoffset = ( "+" / "-" ) time-hour ":" time-minute
+
+	dt, b, err := parseLocalDateTime(b)
+	if err != nil {
+		return time.Time{}, err
+	}
+
+	var zone *time.Location
+
+	if len(b) == 0 {
+		// parser should have checked that when assigning the date time node
+		panic("date time should have a timezone")
+	}
+
+	if b[0] == 'Z' || b[0] == 'z' {
+		b = b[1:]
+		zone = time.UTC
+	} else {
+		const dateTimeByteLen = 6
+		if len(b) != dateTimeByteLen {
+			return time.Time{}, newDecodeError(b, "invalid date-time timezone")
+		}
+		direction := 1
+		if b[0] == '-' {
+			direction = -1
+		}
+
+		hours := digitsToInt(b[1:3])
+		minutes := digitsToInt(b[4:6])
+		seconds := direction * (hours*3600 + minutes*60)
+		zone = time.FixedZone("", seconds)
+		b = b[dateTimeByteLen:]
+	}
+
+	if len(b) > 0 {
+		return time.Time{}, newDecodeError(b, "extra bytes at the end of the timezone")
+	}
+
+	t := time.Date(
+		dt.Year,
+		time.Month(dt.Month),
+		dt.Day,
+		dt.Hour,
+		dt.Minute,
+		dt.Second,
+		dt.Nanosecond,
+		zone)
+
+	return t, nil
+}
+
+func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
+	var dt LocalDateTime
+
+	const localDateTimeByteMinLen = 11
+	if len(b) < localDateTimeByteMinLen {
+		return dt, nil, newDecodeError(b, "local datetimes are expected to have the format YYYY-MM-DDTHH:MM:SS[.NNNNNNNNN]")
+	}
+
+	date, err := parseLocalDate(b[:10])
+	if err != nil {
+		return dt, nil, err
+	}
+	dt.LocalDate = date
+
+	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")
+	}
+
+	t, rest, err := parseLocalTime(b[11:])
+	if err != nil {
+		return dt, nil, err
+	}
+	dt.LocalTime = t
+
+	return dt, rest, nil
+}
+
+// parseLocalTime is a bit different because it also returns the remaining
+// []byte that is didn't need. This is to allow parseDateTime to parse those
+// remaining bytes as a timezone.
+func parseLocalTime(b []byte) (LocalTime, []byte, error) {
+	var (
+		nspow = [10]int{0, 1e8, 1e7, 1e6, 1e5, 1e4, 1e3, 1e2, 1e1, 1e0}
+		t     LocalTime
+	)
+
+	// check if b matches to have expected format HH:MM:SS[.NNNNNN]
+	const localTimeByteLen = 8
+	if len(b) < localTimeByteLen {
+		return t, nil, newDecodeError(b, "times are expected to have the format HH:MM:SS[.NNNNNN]")
+	}
+
+	var err error
+
+	t.Hour, err = parseDecimalDigits(b[0:2])
+	if err != nil {
+		return t, nil, err
+	}
+
+	if t.Hour > 23 {
+		return t, nil, newDecodeError(b[0:2], "hour cannot be greater 23")
+	}
+	if b[2] != ':' {
+		return t, nil, newDecodeError(b[2:3], "expecting colon between hours and minutes")
+	}
+
+	t.Minute, err = parseDecimalDigits(b[3:5])
+	if err != nil {
+		return t, nil, err
+	}
+	if t.Minute > 59 {
+		return t, nil, newDecodeError(b[3:5], "minutes cannot be greater 59")
+	}
+	if b[5] != ':' {
+		return t, nil, newDecodeError(b[5:6], "expecting colon between minutes and seconds")
+	}
+
+	t.Second, err = parseDecimalDigits(b[6:8])
+	if err != nil {
+		return t, nil, err
+	}
+
+	if t.Second > 59 {
+		return t, nil, newDecodeError(b[3:5], "seconds cannot be greater 59")
+	}
+
+	const minLengthWithFrac = 9
+	if len(b) >= minLengthWithFrac && b[minLengthWithFrac-1] == '.' {
+		frac := 0
+		digits := 0
+
+		for i, c := range b[minLengthWithFrac:] {
+			if !isDigit(c) {
+				if i == 0 {
+					return t, nil, newDecodeError(b[i:i+1], "need at least one digit after fraction point")
+				}
+
+				break
+			}
+
+			const maxFracPrecision = 9
+			if i >= maxFracPrecision {
+				return t, nil, newDecodeError(b[i:i+1], "maximum precision for date time is nanosecond")
+			}
+
+			frac *= 10
+			frac += int(c - '0')
+			digits++
+		}
+
+		if digits == 0 {
+			return t, nil, newDecodeError(b[minLengthWithFrac-1:minLengthWithFrac], "nanoseconds need at least one digit")
+		}
+
+		t.Nanosecond = frac * nspow[digits]
+		t.Precision = digits
+
+		return t, b[9+digits:], nil
+	}
+
+	return t, b[8:], nil
+}
+
+//nolint:cyclop
+func parseFloat(b []byte) (float64, error) {
+	if len(b) == 4 && (b[0] == '+' || b[0] == '-') && b[1] == 'n' && b[2] == 'a' && b[3] == 'n' {
+		return math.NaN(), nil
+	}
+
+	cleaned, err := checkAndRemoveUnderscoresFloats(b)
+	if err != nil {
+		return 0, err
+	}
+
+	if cleaned[0] == '.' {
+		return 0, newDecodeError(b, "float cannot start with a dot")
+	}
+
+	if cleaned[len(cleaned)-1] == '.' {
+		return 0, newDecodeError(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")
+			}
+			if !isDigit(cleaned[i-1]) {
+				return 0, newDecodeError(b[i-1:i+1], "float decimal point must be preceded by a digit")
+			}
+			if !isDigit(cleaned[i+1]) {
+				return 0, newDecodeError(b[i:i+2], "float decimal point must be followed by a digit")
+			}
+			dotAlreadySeen = true
+		}
+	}
+
+	start := 0
+	if b[0] == '+' || b[0] == '-' {
+		start = 1
+	}
+	if b[start] == '0' && isDigit(b[start+1]) {
+		return 0, newDecodeError(b, "float integer part cannot have leading zeroes")
+	}
+
+	f, err := strconv.ParseFloat(string(cleaned), 64)
+	if err != nil {
+		return 0, newDecodeError(b, "unable to parse float: %w", err)
+	}
+
+	return f, nil
+}
+
+func parseIntHex(b []byte) (int64, error) {
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
+	if err != nil {
+		return 0, err
+	}
+
+	i, err := strconv.ParseInt(string(cleaned), 16, 64)
+	if err != nil {
+		return 0, newDecodeError(b, "couldn't parse hexadecimal number: %w", err)
+	}
+
+	return i, nil
+}
+
+func parseIntOct(b []byte) (int64, error) {
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
+	if err != nil {
+		return 0, err
+	}
+
+	i, err := strconv.ParseInt(string(cleaned), 8, 64)
+	if err != nil {
+		return 0, newDecodeError(b, "couldn't parse octal number: %w", err)
+	}
+
+	return i, nil
+}
+
+func parseIntBin(b []byte) (int64, error) {
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
+	if err != nil {
+		return 0, err
+	}
+
+	i, err := strconv.ParseInt(string(cleaned), 2, 64)
+	if err != nil {
+		return 0, newDecodeError(b, "couldn't parse binary number: %w", err)
+	}
+
+	return i, nil
+}
+
+func isSign(b byte) bool {
+	return b == '+' || b == '-'
+}
+
+func parseIntDec(b []byte) (int64, error) {
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b)
+	if err != nil {
+		return 0, err
+	}
+
+	startIdx := 0
+
+	if isSign(cleaned[0]) {
+		startIdx++
+	}
+
+	if len(cleaned) > startIdx+1 && cleaned[startIdx] == '0' {
+		return 0, newDecodeError(b, "leading zero not allowed on decimal number")
+	}
+
+	i, err := strconv.ParseInt(string(cleaned), 10, 64)
+	if err != nil {
+		return 0, newDecodeError(b, "couldn't parse decimal number: %w", err)
+	}
+
+	return i, nil
+}
+
+func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
+	if b[0] == '_' {
+		return nil, newDecodeError(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")
+	}
+
+	// fast path
+	i := 0
+	for ; i < len(b); i++ {
+		if b[i] == '_' {
+			break
+		}
+	}
+	if i == len(b) {
+		return b, nil
+	}
+
+	before := false
+	cleaned := make([]byte, i, len(b))
+	copy(cleaned, b)
+
+	for i++; i < len(b); i++ {
+		c := b[i]
+		if c == '_' {
+			if !before {
+				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+			}
+			before = false
+		} else {
+			before = true
+			cleaned = append(cleaned, c)
+		}
+	}
+
+	return cleaned, nil
+}
+
+func checkAndRemoveUnderscoresFloats(b []byte) ([]byte, error) {
+	if b[0] == '_' {
+		return nil, newDecodeError(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")
+	}
+
+	// fast path
+	i := 0
+	for ; i < len(b); i++ {
+		if b[i] == '_' {
+			break
+		}
+	}
+	if i == len(b) {
+		return b, nil
+	}
+
+	before := false
+	cleaned := make([]byte, 0, len(b))
+
+	for i := 0; i < len(b); i++ {
+		c := b[i]
+
+		switch c {
+		case '_':
+			if !before {
+				return nil, newDecodeError(b[i-1:i+1], "number must have at least one digit between underscores")
+			}
+			if i < len(b)-1 && (b[i+1] == 'e' || b[i+1] == 'E') {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore before exponent")
+			}
+			before = false
+		case 'e', 'E':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after exponent")
+			}
+			cleaned = append(cleaned, c)
+		case '.':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after decimal point")
+			}
+			if i > 0 && b[i-1] == '_' {
+				return nil, newDecodeError(b[i-1:i], "cannot have underscore before decimal point")
+			}
+			cleaned = append(cleaned, c)
+		default:
+			before = true
+			cleaned = append(cleaned, c)
+		}
+	}
+
+	return cleaned, nil
+}
+
+// isValidDate checks if a provided date is a date that exists.
+func isValidDate(year int, month int, day int) bool {
+	return day <= daysIn(month, year)
+}
+
+// daysBefore[m] counts the number of days in a non-leap year
+// before month m begins. There is an entry for m=12, counting
+// the number of days before January of next year (365).
+var daysBefore = [...]int32{
+	0,
+	31,
+	31 + 28,
+	31 + 28 + 31,
+	31 + 28 + 31 + 30,
+	31 + 28 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31 + 30,
+	31 + 28 + 31 + 30 + 31 + 30 + 31 + 31 + 30 + 31 + 30 + 31,
+}
+
+func daysIn(m int, year int) int {
+	if m == 2 && isLeap(year) {
+		return 29
+	}
+	return int(daysBefore[m] - daysBefore[m-1])
+}
+
+func isLeap(year int) bool {
+	return year%4 == 0 && (year%100 != 0 || year%400 == 0)
+}

doc.go

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

errors.go

@@ -0,0 +1,269 @@
+package toml
+
+import (
+	"fmt"
+	"strconv"
+	"strings"
+
+	"github.com/pelletier/go-toml/v2/internal/danger"
+)
+
+// DecodeError represents an error encountered during the parsing or decoding
+// of a TOML document.
+//
+// In addition to the error message, it contains the position in the document
+// where it happened, as well as a human-readable representation that shows
+// where the error occurred in the document.
+type DecodeError struct {
+	message string
+	line    int
+	column  int
+	key     Key
+
+	human string
+}
+
+// StrictMissingError occurs in a TOML document that does not have a
+// corresponding field in the target value. It contains all the missing fields
+// in Errors.
+//
+// Emitted by Decoder when SetStrict(true) was called.
+type StrictMissingError struct {
+	// One error per field that could not be found.
+	Errors []DecodeError
+}
+
+// Error returns the canonical string for this error.
+func (s *StrictMissingError) Error() string {
+	return "strict mode: fields in the document are missing in the target struct"
+}
+
+// String returns a human readable description of all errors.
+func (s *StrictMissingError) String() string {
+	var buf strings.Builder
+
+	for i, e := range s.Errors {
+		if i > 0 {
+			buf.WriteString("\n---\n")
+		}
+
+		buf.WriteString(e.String())
+	}
+
+	return buf.String()
+}
+
+type Key []string
+
+// internal version of DecodeError that is used as the base to create a
+// DecodeError with full context.
+type decodeError struct {
+	highlight []byte
+	message   string
+	key       Key // optional
+}
+
+func (de *decodeError) Error() string {
+	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
+}
+
+// String returns the human-readable contextualized error. This string is multi-line.
+func (e *DecodeError) String() string {
+	return e.human
+}
+
+// Position returns the (line, column) pair indicating where the error
+// occurred in the document. Positions are 1-indexed.
+func (e *DecodeError) Position() (row int, column int) {
+	return e.line, e.column
+}
+
+// Key that was being processed when the error occurred. The key is present only
+// if this DecodeError is part of a StrictMissingError.
+func (e *DecodeError) Key() Key {
+	return e.key
+}
+
+// decodeErrorFromHighlight creates a DecodeError referencing a highlighted
+// range of bytes from document.
+//
+// highlight needs to be a sub-slice of document, or this function panics.
+//
+// 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 {
+	offset := danger.SubsliceOffset(document, de.highlight)
+
+	errMessage := de.Error()
+	errLine, errColumn := positionAtEnd(document[:offset])
+	before, after := linesOfContext(document, de.highlight, offset, 3)
+
+	var buf strings.Builder
+
+	maxLine := errLine + len(after) - 1
+	lineColumnWidth := len(strconv.Itoa(maxLine))
+
+	// Write the lines of context strictly before the error.
+	for i := len(before) - 1; i > 0; i-- {
+		line := errLine - i
+		buf.WriteString(formatLineNumber(line, lineColumnWidth))
+		buf.WriteString("|")
+
+		if len(before[i]) > 0 {
+			buf.WriteString(" ")
+			buf.Write(before[i])
+		}
+
+		buf.WriteRune('\n')
+	}
+
+	// Write the document line that contains the error.
+
+	buf.WriteString(formatLineNumber(errLine, lineColumnWidth))
+	buf.WriteString("| ")
+
+	if len(before) > 0 {
+		buf.Write(before[0])
+	}
+
+	buf.Write(de.highlight)
+
+	if len(after) > 0 {
+		buf.Write(after[0])
+	}
+
+	buf.WriteRune('\n')
+
+	// Write the line with the error message itself (so it does not have a line
+	// number).
+
+	buf.WriteString(strings.Repeat(" ", lineColumnWidth))
+	buf.WriteString("| ")
+
+	if len(before) > 0 {
+		buf.WriteString(strings.Repeat(" ", len(before[0])))
+	}
+
+	buf.WriteString(strings.Repeat("~", len(de.highlight)))
+
+	if len(errMessage) > 0 {
+		buf.WriteString(" ")
+		buf.WriteString(errMessage)
+	}
+
+	// Write the lines of context strictly after the error.
+
+	for i := 1; i < len(after); i++ {
+		buf.WriteRune('\n')
+		line := errLine + i
+		buf.WriteString(formatLineNumber(line, lineColumnWidth))
+		buf.WriteString("|")
+
+		if len(after[i]) > 0 {
+			buf.WriteString(" ")
+			buf.Write(after[i])
+		}
+	}
+
+	return &DecodeError{
+		message: errMessage,
+		line:    errLine,
+		column:  errColumn,
+		key:     de.key,
+		human:   buf.String(),
+	}
+}
+
+func formatLineNumber(line int, width int) string {
+	format := "%" + strconv.Itoa(width) + "d"
+
+	return fmt.Sprintf(format, line)
+}
+
+func linesOfContext(document []byte, highlight []byte, offset int, linesAround int) ([][]byte, [][]byte) {
+	return beforeLines(document, offset, linesAround), afterLines(document, highlight, offset, linesAround)
+}
+
+func beforeLines(document []byte, offset int, linesAround int) [][]byte {
+	var beforeLines [][]byte
+
+	// Walk the document backward from the highlight to find previous lines
+	// of context.
+	rest := document[:offset]
+backward:
+	for o := len(rest) - 1; o >= 0 && len(beforeLines) <= linesAround && len(rest) > 0; {
+		switch {
+		case rest[o] == '\n':
+			// handle individual lines
+			beforeLines = append(beforeLines, rest[o+1:])
+			rest = rest[:o]
+			o = len(rest) - 1
+		case o == 0:
+			// add the first line only if it's non-empty
+			beforeLines = append(beforeLines, rest)
+
+			break backward
+		default:
+			o--
+		}
+	}
+
+	return beforeLines
+}
+
+func afterLines(document []byte, highlight []byte, offset int, linesAround int) [][]byte {
+	var afterLines [][]byte
+
+	// Walk the document forward from the highlight to find the following
+	// lines of context.
+	rest := document[offset+len(highlight):]
+forward:
+	for o := 0; o < len(rest) && len(afterLines) <= linesAround; {
+		switch {
+		case rest[o] == '\n':
+			// handle individual lines
+			afterLines = append(afterLines, rest[:o])
+			rest = rest[o+1:]
+			o = 0
+
+		case o == len(rest)-1:
+			// add last line only if it's non-empty
+			afterLines = append(afterLines, rest)
+
+			break forward
+		default:
+			o++
+		}
+	}
+
+	return afterLines
+}
+
+func positionAtEnd(b []byte) (row int, column int) {
+	row = 1
+	column = 1
+
+	for _, c := range b {
+		if c == '\n' {
+			row++
+			column = 1
+		} else {
+			column++
+		}
+	}
+
+	return
+}

errors_test.go

@@ -0,0 +1,226 @@
+package toml
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+//nolint:funlen
+func TestDecodeError(t *testing.T) {
+
+	examples := []struct {
+		desc     string
+		doc      [3]string
+		msg      string
+		expected string
+	}{
+		{
+			desc: "no context",
+			doc:  [3]string{"", "morning", ""},
+			msg:  "this is wrong",
+			expected: `
+1| morning
+ | ~~~~~~~ this is wrong`,
+		},
+		{
+			desc: "one line",
+			doc:  [3]string{"good ", "morning", " everyone"},
+			msg:  "this is wrong",
+			expected: `
+1| good morning everyone
+ |      ~~~~~~~ this is wrong`,
+		},
+		{
+			desc: "exactly 3 lines",
+			doc: [3]string{`line1
+line2
+line3
+before `, "highlighted", ` after
+post line 1
+post line 2
+post line 3`},
+			msg: "this is wrong",
+			expected: `
+1| line1
+2| line2
+3| line3
+4| before highlighted after
+ |        ~~~~~~~~~~~ this is wrong
+5| post line 1
+6| post line 2
+7| post line 3`,
+		},
+		{
+			desc: "more than 3 lines",
+			doc: [3]string{`should not be seen1
+should not be seen2
+line1
+line2
+line3
+before `, "highlighted", ` after
+post line 1
+post line 2
+post line 3
+should not be seen3
+should not be seen4`},
+			msg: "this is wrong",
+			expected: `
+3| line1
+4| line2
+5| line3
+6| before highlighted after
+ |        ~~~~~~~~~~~ this is wrong
+7| post line 1
+8| post line 2
+9| post line 3`,
+		},
+		{
+			desc: "more than 10 total lines",
+			doc: [3]string{`should not be seen 0
+should not be seen1
+should not be seen2
+should not be seen3
+line1
+line2
+line3
+before `, "highlighted", ` after
+post line 1
+post line 2
+post line 3
+should not be seen3
+should not be seen4`},
+			msg: "this is wrong",
+			expected: `
+ 5| line1
+ 6| line2
+ 7| line3
+ 8| before highlighted after
+  |        ~~~~~~~~~~~ this is wrong
+ 9| post line 1
+10| post line 2
+11| post line 3`,
+		},
+		{
+			desc: "last line of more than 10",
+			doc: [3]string{`should not be seen
+should not be seen
+should not be seen
+should not be seen
+should not be seen
+should not be seen
+should not be seen
+line1
+line2
+line3
+before `, "highlighted", ``},
+			msg: "this is wrong",
+			expected: `
+ 8| line1
+ 9| line2
+10| line3
+11| before highlighted
+  |        ~~~~~~~~~~~ this is wrong
+`,
+		},
+		{
+			desc: "handle empty lines in the before/after blocks",
+			doc: [3]string{
+				`line1
+
+line 2
+before `, "highlighted", ` after
+line 3
+
+line 4
+line 5`,
+			},
+			expected: `1| line1
+2|
+3| line 2
+4| before highlighted after
+ |        ~~~~~~~~~~~
+5| line 3
+6|
+7| line 4`,
+		},
+		{
+			desc: "handle remainder of the error line when there is only one line",
+			doc:  [3]string{`P=`, `[`, `#`},
+			msg:  "array is incomplete",
+			expected: `1| P=[#
+ |   ~ array is incomplete`,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+
+			b := bytes.Buffer{}
+			b.Write([]byte(e.doc[0]))
+			start := b.Len()
+			b.Write([]byte(e.doc[1]))
+			end := b.Len()
+			b.Write([]byte(e.doc[2]))
+			doc := b.Bytes()
+			hl := doc[start:end]
+
+			err := wrapDecodeError(doc, &decodeError{
+				highlight: hl,
+				message:   e.msg,
+			})
+
+			var derr *DecodeError
+			if !errors.As(err, &derr) {
+				t.Errorf("error not in expected format")
+
+				return
+			}
+
+			assert.Equal(t, strings.Trim(e.expected, "\n"), derr.String())
+		})
+	}
+}
+
+func TestDecodeError_Accessors(t *testing.T) {
+
+	e := DecodeError{
+		message: "foo",
+		line:    1,
+		column:  2,
+		key:     []string{"one", "two"},
+		human:   "bar",
+	}
+	assert.Equal(t, "toml: foo", e.Error())
+	r, c := e.Position()
+	assert.Equal(t, 1, r)
+	assert.Equal(t, 2, c)
+	assert.Equal(t, Key{"one", "two"}, e.Key())
+	assert.Equal(t, "bar", e.String())
+}
+
+func ExampleDecodeError() {
+	doc := `name = 123__456`
+
+	s := map[string]interface{}{}
+	err := Unmarshal([]byte(doc), &s)
+
+	fmt.Println(err)
+
+	//nolint:errorlint
+	de := err.(*DecodeError)
+	fmt.Println(de.String())
+
+	row, col := de.Position()
+	fmt.Println("error occurred at row", row, "column", col)
+	// Output:
+	// toml: number must have at least one digit between underscores
+	// 1| name = 123__456
+	//  |           ~~ number must have at least one digit between underscores
+	// error occurred at row 1 column 11
+}

example.toml

@@ -1,29 +0,0 @@
-# This is a TOML document. Boom.
-
-title = "TOML Example"
-
-[owner]
-name = "Tom Preston-Werner"
-organization = "GitHub"
-bio = "GitHub Cofounder & CEO\nLikes tater tots and beer."
-dob = 1979-05-27T07:32:00Z # First class dates? Why not?
-
-[database]
-server = "192.168.1.1"
-ports = [ 8001, 8001, 8002 ]
-connection_max = 5000
-enabled = true
-
-[servers]
-
-  # You can indent as you please. Tabs or spaces. TOML don't care.
-  [servers.alpha]
-  ip = "10.0.0.1"
-  dc = "eqdc10"
-
-  [servers.beta]
-  ip = "10.0.0.2"
-  dc = "eqdc10"
-
-[clients]
-data = [ ["gamma", "delta"], [1, 2] ] # just an update to make sure parsers support it

fast_test.go

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

go.mod

@@ -0,0 +1,6 @@
+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 v1.7.1-0.20210427113832-6241f9ab9942

go.sum

@@ -0,0 +1,11 @@
+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/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/testify v1.7.1-0.20210427113832-6241f9ab9942/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
+gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
+gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c h1:dUUwHk2QECo/6vqA44rthZ8ie2QXMNeKRTHCNY2nXvo=
+gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

internal/ast/ast.go

@@ -0,0 +1,145 @@
+package ast
+
+import (
+	"fmt"
+	"unsafe"
+
+	"github.com/pelletier/go-toml/v2/internal/danger"
+)
+
+// Iterator starts uninitialized, you need to call Next() first.
+//
+// For example:
+//
+// it := n.Children()
+// for it.Next() {
+// 		it.Node()
+// }
+type Iterator struct {
+	started bool
+	node    *Node
+}
+
+// Next moves the iterator forward and returns true if points to a node, false
+// otherwise.
+func (c *Iterator) Next() bool {
+	if !c.started {
+		c.started = true
+	} else if c.node.Valid() {
+		c.node = c.node.Next()
+	}
+	return c.node.Valid()
+}
+
+// IsLast returns true if the current node of the iterator is the last one.
+// Subsequent call to Next() will return false.
+func (c *Iterator) IsLast() bool {
+	return c.node.next == 0
+}
+
+// Node returns a copy of the node pointed at by the iterator.
+func (c *Iterator) Node() *Node {
+	return c.node
+}
+
+// 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]
+}
+
+// Arrays have one child per element in the array.
+// InlineTables have one child per key-value pair in the table.
+// KeyValues have at least two children. The first one is the value. The
+// rest make a potentially dotted key.
+// Table and Array table have one child per element of the key they
+// represent (same as KeyValue, but without the last node being the value).
+// children []Node
+type Node struct {
+	Kind Kind
+	Raw  Range  // Raw bytes from the input.
+	Data []byte // Node value (could be either allocated or referencing the input).
+
+	// References to other nodes, as offsets in the backing array from this
+	// node. References can go backward, so those can be negative.
+	next  int // 0 if last element
+	child int // 0 if no child
+}
+
+type Range struct {
+	Offset uint32
+	Length uint32
+}
+
+// Next returns a copy of the next node, or an invalid Node if there is no
+// next node.
+func (n *Node) Next() *Node {
+	if n.next == 0 {
+		return nil
+	}
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.next))
+}
+
+// Child returns a copy of the first child node of this node. Other children
+// can be accessed calling Next on the first child.
+// Returns an invalid Node if there is none.
+func (n *Node) Child() *Node {
+	if n.child == 0 {
+		return nil
+	}
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.child))
+}
+
+// Valid returns true if the node's kind is set (not to Invalid).
+func (n *Node) Valid() bool {
+	return n != nil
+}
+
+// Key returns the child nodes making the Key on a supported node. Panics
+// otherwise.
+// They are guaranteed to be all be of the Kind Key. A simple key would return
+// just one element.
+func (n *Node) Key() Iterator {
+	switch n.Kind {
+	case KeyValue:
+		value := n.Child()
+		if !value.Valid() {
+			panic(fmt.Errorf("KeyValue should have at least two children"))
+		}
+		return Iterator{node: value.Next()}
+	case Table, ArrayTable:
+		return Iterator{node: n.Child()}
+	default:
+		panic(fmt.Errorf("Key() is not supported on a %s", n.Kind))
+	}
+}
+
+// Value returns a pointer to the value node of a KeyValue.
+// Guaranteed to be non-nil.
+// Panics if not called on a KeyValue node, or if the Children are malformed.
+func (n *Node) Value() *Node {
+	return n.Child()
+}
+
+// Children returns an iterator over a node's children.
+func (n *Node) Children() Iterator {
+	return Iterator{node: n.Child()}
+}

internal/ast/builder.go

@@ -0,0 +1,51 @@
+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/ast/kind.go

@@ -0,0 +1,69 @@
+package ast
+
+import "fmt"
+
+type Kind int
+
+const (
+	// meta
+	Invalid Kind = iota
+	Comment
+	Key
+
+	// top level structures
+	Table
+	ArrayTable
+	KeyValue
+
+	// containers values
+	Array
+	InlineTable
+
+	// values
+	String
+	Bool
+	Float
+	Integer
+	LocalDate
+	LocalTime
+	LocalDateTime
+	DateTime
+)
+
+func (k Kind) String() string {
+	switch k {
+	case Invalid:
+		return "Invalid"
+	case Comment:
+		return "Comment"
+	case Key:
+		return "Key"
+	case Table:
+		return "Table"
+	case ArrayTable:
+		return "ArrayTable"
+	case KeyValue:
+		return "KeyValue"
+	case Array:
+		return "Array"
+	case InlineTable:
+		return "InlineTable"
+	case String:
+		return "String"
+	case Bool:
+		return "Bool"
+	case Float:
+		return "Float"
+	case Integer:
+		return "Integer"
+	case LocalDate:
+		return "LocalDate"
+	case LocalTime:
+		return "LocalTime"
+	case LocalDateTime:
+		return "LocalDateTime"
+	case DateTime:
+		return "DateTime"
+	}
+	panic(fmt.Errorf("Kind.String() not implemented for '%d'", k))
+}

internal/danger/danger.go

@@ -0,0 +1,76 @@
+package danger
+
+import (
+	"fmt"
+	"reflect"
+	"unsafe"
+)
+
+const maxInt = uintptr(int(^uint(0) >> 1))
+
+func SubsliceOffset(data []byte, subslice []byte) int {
+	datap := (*reflect.SliceHeader)(unsafe.Pointer(&data))
+	hlp := (*reflect.SliceHeader)(unsafe.Pointer(&subslice))
+
+	if hlp.Data < datap.Data {
+		panic(fmt.Errorf("subslice address (%d) is before data address (%d)", hlp.Data, datap.Data))
+	}
+	offset := hlp.Data - datap.Data
+
+	if offset > maxInt {
+		panic(fmt.Errorf("slice offset larger than int (%d)", offset))
+	}
+
+	intoffset := int(offset)
+
+	if intoffset > datap.Len {
+		panic(fmt.Errorf("slice offset (%d) is farther than data length (%d)", intoffset, datap.Len))
+	}
+
+	if intoffset+hlp.Len > datap.Len {
+		panic(fmt.Errorf("slice ends (%d+%d) is farther than data length (%d)", intoffset, hlp.Len, datap.Len))
+	}
+
+	return intoffset
+}
+
+func BytesRange(start []byte, end []byte) []byte {
+	if start == nil || end == nil {
+		panic("cannot call BytesRange with nil")
+	}
+	startp := (*reflect.SliceHeader)(unsafe.Pointer(&start))
+	endp := (*reflect.SliceHeader)(unsafe.Pointer(&end))
+
+	if startp.Data > endp.Data {
+		panic(fmt.Errorf("start pointer address (%d) is after end pointer address (%d)", startp.Data, endp.Data))
+	}
+
+	l := startp.Len
+	endLen := int(endp.Data-startp.Data) + endp.Len
+	if endLen > l {
+		l = endLen
+	}
+
+	if l > startp.Cap {
+		panic(fmt.Errorf("range length is larger than capacity"))
+	}
+
+	return start[:l]
+}
+
+func Stride(ptr unsafe.Pointer, size uintptr, offset int) unsafe.Pointer {
+	// TODO: replace with unsafe.Add when Go 1.17 is released
+	//   https://github.com/golang/go/issues/40481
+	return unsafe.Pointer(uintptr(ptr) + uintptr(int(size)*offset))
+}
+
+type Slice struct {
+	Data unsafe.Pointer
+	Len  int
+	Cap  int
+}
+
+type iface struct {
+	typ unsafe.Pointer
+	ptr unsafe.Pointer
+}

internal/danger/danger_test.go

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

internal/danger/slices.go

@@ -0,0 +1,20 @@
+//go:build go1.18
+// +build go1.18
+
+package danger
+
+import (
+	"reflect"
+	"unsafe"
+)
+
+func ExtendSlice(t reflect.Type, s *Slice, n int) Slice {
+	arrayType := reflect.ArrayOf(n, t.Elem())
+	arrayData := reflect.New(arrayType)
+	reflect.Copy(arrayData.Elem(), reflect.NewAt(t, unsafe.Pointer(s)).Elem())
+	return Slice{
+		Data: unsafe.Pointer(arrayData.Pointer()),
+		Len:  s.Len,
+		Cap:  n,
+	}
+}

internal/danger/slices_optimized.go

@@ -0,0 +1,30 @@
+//go:build !go1.18
+// +build !go1.18
+
+package danger
+
+import (
+	"reflect"
+	"unsafe"
+)
+
+//go:linkname unsafe_NewArray reflect.unsafe_NewArray
+func unsafe_NewArray(rtype unsafe.Pointer, length int) unsafe.Pointer
+
+//go:linkname typedslicecopy reflect.typedslicecopy
+//go:noescape
+func typedslicecopy(elemType unsafe.Pointer, dst, src Slice) int
+
+func ExtendSlice(t reflect.Type, s *Slice, n int) Slice {
+	elemTypeRef := t.Elem()
+	elemTypePtr := ((*iface)(unsafe.Pointer(&elemTypeRef))).ptr
+
+	d := Slice{
+		Data: unsafe_NewArray(elemTypePtr, n),
+		Len:  s.Len,
+		Cap:  n,
+	}
+
+	typedslicecopy(elemTypePtr, d, *s)
+	return d
+}

internal/danger/typeid.go

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

internal/imported_tests/marshal_imported_test.go

@@ -0,0 +1,198 @@
+package imported_tests
+
+// Those tests have been imported from v1, but adjust to match the new
+// defaults of v2.
+
+import (
+	"fmt"
+	"testing"
+	"time"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+func TestDocMarshal(t *testing.T) {
+	type testDoc struct {
+		Title       string            `toml:"title"`
+		BasicLists  testDocBasicLists `toml:"basic_lists"`
+		SubDocPtrs  []*testSubDoc     `toml:"subdocptrs"`
+		BasicMap    map[string]string `toml:"basic_map"`
+		Subdocs     testDocSubs       `toml:"subdoc"`
+		Basics      testDocBasics     `toml:"basic"`
+		SubDocList  []testSubDoc      `toml:"subdoclist"`
+		err         int               `toml:"shouldntBeHere"`
+		unexported  int               `toml:"shouldntBeHere"`
+		Unexported2 int               `toml:"-"`
+	}
+
+	var docData = testDoc{
+		Title:       "TOML Marshal Testing",
+		unexported:  0,
+		Unexported2: 0,
+		Basics: testDocBasics{
+			Bool:       true,
+			Date:       time.Date(1979, 5, 27, 7, 32, 0, 0, time.UTC),
+			Float32:    123.4,
+			Float64:    123.456782132399,
+			Int:        5000,
+			Uint:       5001,
+			String:     &biteMe,
+			unexported: 0,
+		},
+		BasicLists: testDocBasicLists{
+			Floats: []*float32{&float1, &float2, &float3},
+			Bools:  []bool{true, false, true},
+			Dates: []time.Time{
+				time.Date(1979, 5, 27, 7, 32, 0, 0, time.UTC),
+				time.Date(1980, 5, 27, 7, 32, 0, 0, time.UTC),
+			},
+			Ints:    []int{8001, 8001, 8002},
+			Strings: []string{"One", "Two", "Three"},
+			UInts:   []uint{5002, 5003},
+		},
+		BasicMap: map[string]string{
+			"one": "one",
+			"two": "two",
+		},
+		Subdocs: testDocSubs{
+			First:  testSubDoc{"First", 0},
+			Second: &subdoc,
+		},
+		SubDocList: []testSubDoc{
+			{"List.First", 0},
+			{"List.Second", 0},
+		},
+		SubDocPtrs: []*testSubDoc{&subdoc},
+	}
+
+	marshalTestToml := `title = 'TOML Marshal Testing'
+[basic_lists]
+floats = [12.3, 45.6, 78.9]
+bools = [true, false, true]
+dates = [1979-05-27T07:32:00Z, 1980-05-27T07:32:00Z]
+ints = [8001, 8001, 8002]
+uints = [5002, 5003]
+strings = ['One', 'Two', 'Three']
+
+[[subdocptrs]]
+name = 'Second'
+
+[basic_map]
+one = 'one'
+two = 'two'
+
+[subdoc]
+[subdoc.second]
+name = 'Second'
+
+[subdoc.first]
+name = 'First'
+
+
+[basic]
+uint = 5001
+bool = true
+float = 123.4
+float64 = 123.456782132399
+int = 5000
+string = 'Bite me'
+date = 1979-05-27T07:32:00Z
+
+[[subdoclist]]
+name = 'List.First'
+[[subdoclist]]
+name = 'List.Second'
+
+`
+
+	result, err := toml.Marshal(docData)
+	require.NoError(t, err)
+	require.Equal(t, marshalTestToml, string(result))
+}
+
+func TestBasicMarshalQuotedKey(t *testing.T) {
+	result, err := toml.Marshal(quotedKeyMarshalTestData)
+	require.NoError(t, err)
+
+	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))
+
+}
+
+func TestEmptyMarshal(t *testing.T) {
+	type emptyMarshalTestStruct struct {
+		Title      string                  `toml:"title"`
+		Bool       bool                    `toml:"bool"`
+		Int        int                     `toml:"int"`
+		String     string                  `toml:"string"`
+		StringList []string                `toml:"stringlist"`
+		Ptr        *basicMarshalTestStruct `toml:"ptr"`
+		Map        map[string]string       `toml:"map"`
+	}
+
+	doc := emptyMarshalTestStruct{
+		Title:      "Placeholder",
+		Bool:       false,
+		Int:        0,
+		String:     "",
+		StringList: []string{},
+		Ptr:        nil,
+		Map:        map[string]string{},
+	}
+	result, err := toml.Marshal(doc)
+	require.NoError(t, err)
+
+	expected := `title = 'Placeholder'
+bool = false
+int = 0
+string = ''
+stringlist = []
+[map]
+
+`
+
+	require.Equal(t, string(expected), string(result))
+}
+
+type textMarshaler struct {
+	FirstName string
+	LastName  string
+}
+
+func (m textMarshaler) MarshalText() ([]byte, error) {
+	fullName := fmt.Sprintf("%s %s", m.FirstName, m.LastName)
+	return []byte(fullName), nil
+}
+
+func TestTextMarshaler(t *testing.T) {
+	type wrap struct {
+		TM textMarshaler
+	}
+
+	m := textMarshaler{FirstName: "Sally", LastName: "Fields"}
+
+	t.Run("at root", func(t *testing.T) {
+		_, err := toml.Marshal(m)
+		// in v2 we do not allow TextMarshaler at root
+		require.Error(t, err)
+	})
+
+	t.Run("leaf", func(t *testing.T) {
+		res, err := toml.Marshal(wrap{m})
+		require.NoError(t, err)
+
+		require.Equal(t, "TM = 'Sally Fields'\n", string(res))
+	})
+}

internal/tracker/key.go

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

internal/tracker/seen.go

@@ -0,0 +1,304 @@
+package tracker
+
+import (
+	"bytes"
+	"fmt"
+
+	"github.com/pelletier/go-toml/v2/internal/ast"
+)
+
+type keyKind uint8
+
+const (
+	invalidKind keyKind = iota
+	valueKind
+	tableKind
+	arrayTableKind
+)
+
+func (k keyKind) String() string {
+	switch k {
+	case invalidKind:
+		return "invalid"
+	case valueKind:
+		return "value"
+	case tableKind:
+		return "table"
+	case arrayTableKind:
+		return "array table"
+	}
+	panic("missing keyKind string mapping")
+}
+
+// SeenTracker tracks which keys have been seen with which TOML type to flag
+// duplicates and mismatches according to the spec.
+//
+// Each node in the visited tree is represented by an entry. Each entry has an
+// identifier, which is provided by a counter. Entries are stored in the array
+// entries. As new nodes are discovered (referenced for the first time in the
+// TOML document), entries are created and appended to the array. An entry
+// points to its parent using its id.
+//
+// To find whether a given key (sequence of []byte) has already been visited,
+// the entries are linearly searched, looking for one with the right name and
+// parent id.
+//
+// Given that all keys appear in the document after their parent, it is
+// guaranteed that all descendants of a node are stored after the node, this
+// speeds up the search process.
+//
+// When encountering [[array tables]], the descendants of that node are removed
+// to allow that branch of the tree to be "rediscovered". To maintain the
+// invariant above, the deletion process needs to keep the order of entries.
+// This results in more copies in that case.
+type SeenTracker struct {
+	entries    []entry
+	currentIdx int
+	nextID     int
+}
+
+type entry struct {
+	id       int
+	parent   int
+	name     []byte
+	kind     keyKind
+	explicit bool
+}
+
+// Remove all descendants of node at position idx.
+func (s *SeenTracker) clear(idx int) {
+	p := s.entries[idx].id
+	rest := clear(p, s.entries[idx+1:])
+	s.entries = s.entries[:idx+1+len(rest)]
+}
+
+func clear(parentID int, entries []entry) []entry {
+	for i := 0; i < len(entries); {
+		if entries[i].parent == parentID {
+			id := entries[i].id
+			copy(entries[i:], entries[i+1:])
+			entries = entries[:len(entries)-1]
+			rest := clear(id, entries[i:])
+			entries = entries[:i+len(rest)]
+		} else {
+			i++
+		}
+	}
+	return entries
+}
+
+func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit bool) int {
+	parentID := s.id(parentIdx)
+
+	idx := len(s.entries)
+	s.entries = append(s.entries, entry{
+		id:       s.nextID,
+		parent:   parentID,
+		name:     name,
+		kind:     kind,
+		explicit: explicit,
+	})
+	s.nextID++
+	return idx
+}
+
+// CheckExpression takes a top-level node and checks that it does not contain
+// keys that have been seen in previous calls, and validates that types are
+// consistent.
+func (s *SeenTracker) CheckExpression(node *ast.Node) error {
+	if s.entries == nil {
+		// Skip ID = 0 to remove the confusion between nodes whose
+		// parent has id 0 and root nodes (parent id is 0 because it's
+		// the zero value).
+		s.nextID = 1
+		// Start unscoped, so idx is negative.
+		s.currentIdx = -1
+	}
+	switch node.Kind {
+	case ast.KeyValue:
+		return s.checkKeyValue(s.currentIdx, node)
+	case ast.Table:
+		return s.checkTable(node)
+	case ast.ArrayTable:
+		return s.checkArrayTable(node)
+	default:
+		panic(fmt.Errorf("this should not be a top level node type: %s", node.Kind))
+	}
+}
+
+func (s *SeenTracker) checkTable(node *ast.Node) error {
+	it := node.Key()
+
+	parentIdx := -1
+
+	// This code is duplicated in checkArrayTable. This is because factoring
+	// it in a function requires to copy the iterator, or allocate it to the
+	// heap, which is not cheap.
+	for it.Next() {
+		if it.IsLast() {
+			break
+		}
+
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false)
+		}
+		parentIdx = idx
+	}
+
+	k := it.Node().Data
+	idx := s.find(parentIdx, k)
+
+	if idx >= 0 {
+		kind := s.entries[idx].kind
+		if kind != tableKind {
+			return fmt.Errorf("toml: key %s should be a table, not a %s", string(k), kind)
+		}
+		if s.entries[idx].explicit {
+			return fmt.Errorf("toml: table %s already exists", string(k))
+		}
+		s.entries[idx].explicit = true
+	} else {
+		idx = s.create(parentIdx, k, tableKind, true)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
+	it := node.Key()
+
+	parentIdx := -1
+
+	for it.Next() {
+		if it.IsLast() {
+			break
+		}
+
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false)
+		}
+		parentIdx = idx
+	}
+
+	k := it.Node().Data
+	idx := s.find(parentIdx, k)
+
+	if idx >= 0 {
+		kind := s.entries[idx].kind
+		if kind != arrayTableKind {
+			return fmt.Errorf("toml: key %s already exists as a %s,  but should be an array table", kind, string(k))
+		}
+		s.clear(idx)
+	} else {
+		idx = s.create(parentIdx, k, arrayTableKind, true)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkKeyValue(parentIdx int, node *ast.Node) error {
+	it := node.Key()
+
+	for it.Next() {
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false)
+		} else {
+			entry := s.entries[idx]
+			if it.IsLast() {
+				return fmt.Errorf("toml: key %s is already defined", string(k))
+			} else if entry.kind != tableKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			} else if entry.explicit {
+				return fmt.Errorf("toml: cannot redefine table %s that has already been explicitly defined", string(k))
+			}
+		}
+
+		parentIdx = idx
+	}
+
+	s.entries[parentIdx].kind = valueKind
+
+	value := node.Value()
+
+	switch value.Kind {
+	case ast.InlineTable:
+		return s.checkInlineTable(parentIdx, value)
+	case ast.Array:
+		return s.checkArray(parentIdx, value)
+	}
+
+	return nil
+}
+
+func (s *SeenTracker) checkArray(parentIdx int, node *ast.Node) error {
+	set := false
+	it := node.Children()
+	for it.Next() {
+		if set {
+			s.clear(parentIdx)
+		}
+		n := it.Node()
+		switch n.Kind {
+		case ast.InlineTable:
+			err := s.checkInlineTable(parentIdx, n)
+			if err != nil {
+				return err
+			}
+			set = true
+		case ast.Array:
+			err := s.checkArray(parentIdx, n)
+			if err != nil {
+				return err
+			}
+			set = true
+		}
+	}
+	return nil
+}
+
+func (s *SeenTracker) checkInlineTable(parentIdx int, node *ast.Node) error {
+	it := node.Children()
+	for it.Next() {
+		n := it.Node()
+		err := s.checkKeyValue(parentIdx, n)
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (s *SeenTracker) id(idx int) int {
+	if idx >= 0 {
+		return s.entries[idx].id
+	}
+	return 0
+}
+
+func (s *SeenTracker) find(parentIdx int, k []byte) int {
+	parentID := s.id(parentIdx)
+
+	for i := parentIdx + 1; i < len(s.entries); i++ {
+		if s.entries[i].parent == parentID && bytes.Equal(s.entries[i].name, k) {
+			return i
+		}
+	}
+
+	return -1
+}

internal/tracker/tracker.go

@@ -0,0 +1 @@
+package tracker

lexer.go

@@ -1,436 +0,0 @@
-// TOML lexer.// Written using the principles developped by Rob Pike in
-// http://www.youtube.com/watch?v=HxaD_trXwRE
-
-package toml
-
-import (
-	"fmt"
-	"regexp"
-	"strconv"
-	"strings"
-	"unicode"
-	"unicode/utf8"
-)
-
-var dateRegexp *regexp.Regexp
-
-// Define tokens
-type tokenType int
-
-const (
-	eof = -(iota + 1)
-)
-
-const (
-	tokenError tokenType = iota
-	tokenEOF
-	tokenComment
-	tokenKey
-	tokenEqual
-	tokenString
-	tokenInteger
-	tokenTrue
-	tokenFalse
-	tokenFloat
-	tokenLeftBracket
-	tokenRightBracket
-	tokenDate
-	tokenKeyGroup
-	tokenComma
-	tokenEOL
-)
-
-type token struct {
-	typ tokenType
-	val string
-}
-
-func (i token) String() string {
-	switch i.typ {
-	case tokenEOF:
-		return "EOF"
-	case tokenError:
-		return i.val
-	}
-
-	if len(i.val) > 10 {
-		return fmt.Sprintf("%.10q...", i.val)
-	}
-	return fmt.Sprintf("%q", i.val)
-}
-
-func isSpace(r rune) bool {
-	return r == ' ' || r == '\t'
-}
-
-func isAlphanumeric(r rune) bool {
-	return unicode.IsLetter(r) || r == '_'
-}
-
-func isDigit(r rune) bool {
-	return unicode.IsNumber(r)
-}
-
-func isHexDigit(r rune) bool {
-	return isDigit(r) ||
-		r == 'A' || r == 'B' || r == 'C' || r == 'D' || r == 'E' || r == 'F'
-}
-
-// Define lexer
-type lexer struct {
-	input  string
-	start  int
-	pos    int
-	width  int
-	tokens chan token
-	depth  int
-}
-
-func (l *lexer) run() {
-	for state := lexVoid; state != nil; {
-		state = state(l)
-	}
-	close(l.tokens)
-}
-
-func (l *lexer) emit(t tokenType) {
-	l.tokens <- token{t, l.input[l.start:l.pos]}
-	l.start = l.pos
-}
-
-func (l *lexer) emitWithValue(t tokenType, value string) {
-	l.tokens <- token{t, value}
-	l.start = l.pos
-}
-
-func (l *lexer) next() rune {
-	if l.pos >= len(l.input) {
-		l.width = 0
-		return eof
-	}
-	var r rune
-	r, l.width = utf8.DecodeRuneInString(l.input[l.pos:])
-	l.pos += l.width
-	return r
-}
-
-func (l *lexer) ignore() {
-	l.start = l.pos
-}
-
-func (l *lexer) backup() {
-	l.pos -= l.width
-}
-
-func (l *lexer) errorf(format string, args ...interface{}) stateFn {
-	l.tokens <- token{
-		tokenError,
-		fmt.Sprintf(format, args...),
-	}
-	return nil
-}
-
-func (l *lexer) peek() rune {
-	r := l.next()
-	l.backup()
-	return r
-}
-
-func (l *lexer) accept(valid string) bool {
-	if strings.IndexRune(valid, l.next()) >= 0 {
-		return true
-	}
-	l.backup()
-	return false
-}
-
-func (l *lexer) follow(next string) bool {
-	return strings.HasPrefix(l.input[l.pos:], next)
-}
-
-// Define state functions
-type stateFn func(*lexer) stateFn
-
-func lexVoid(l *lexer) stateFn {
-	for {
-		next := l.peek()
-		switch next {
-		case '[':
-			return lexKeyGroup
-		case '#':
-			return lexComment
-		case '=':
-			return lexEqual
-		}
-
-		if isAlphanumeric(next) {
-			return lexKey
-		}
-
-		if isSpace(next) {
-			l.ignore()
-		}
-
-		if l.next() == eof {
-			break
-		}
-	}
-
-	l.emit(tokenEOF)
-	return nil
-}
-
-func lexRvalue(l *lexer) stateFn {
-	for {
-		next := l.peek()
-		switch next {
-		case '[':
-			l.depth += 1
-			return lexLeftBracket
-		case ']':
-			l.depth -= 1
-			return lexRightBracket
-		case '#':
-			return lexComment
-		case '"':
-			return lexString
-		case ',':
-			return lexComma
-		case '\n':
-			l.ignore()
-			l.pos += 1
-			if l.depth == 0 {
-				return lexVoid
-			} else {
-				return lexRvalue
-			}
-		}
-
-		if l.follow("true") {
-			return lexTrue
-		}
-
-		if l.follow("false") {
-			return lexFalse
-		}
-
-		if isAlphanumeric(next) {
-			return lexKey
-		}
-
-		if dateRegexp.FindString(l.input[l.pos:]) != "" {
-			return lexDate
-		}
-
-		if next == '+' || next == '-' || isDigit(next) {
-			return lexNumber
-		}
-
-		if isSpace(next) {
-			l.ignore()
-		}
-
-		if l.next() == eof {
-			break
-		}
-	}
-
-	l.emit(tokenEOF)
-	return nil
-}
-
-func lexDate(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 20 // Fixed size of a date in TOML
-	l.emit(tokenDate)
-	return lexRvalue
-}
-
-func lexTrue(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 4
-	l.emit(tokenTrue)
-	return lexRvalue
-}
-
-func lexFalse(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 5
-	l.emit(tokenFalse)
-	return lexRvalue
-}
-
-func lexEqual(l *lexer) stateFn {
-	l.ignore()
-	l.accept("=")
-	l.emit(tokenEqual)
-	return lexRvalue
-}
-
-func lexComma(l *lexer) stateFn {
-	l.ignore()
-	l.accept(",")
-	l.emit(tokenComma)
-	return lexRvalue
-}
-
-func lexKey(l *lexer) stateFn {
-	l.ignore()
-	for isAlphanumeric(l.next()) {
-	}
-	l.backup()
-	l.emit(tokenKey)
-	return lexVoid
-}
-
-func lexComment(l *lexer) stateFn {
-	for {
-		next := l.next()
-		if next == '\n' || next == eof {
-			break
-		}
-	}
-	l.ignore()
-	return lexVoid
-}
-
-func lexLeftBracket(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 1
-	l.emit(tokenLeftBracket)
-	return lexRvalue
-}
-
-func lexString(l *lexer) stateFn {
-	l.pos += 1
-	l.ignore()
-	growing_string := ""
-
-	for {
-		if l.peek() == '"' {
-			l.emitWithValue(tokenString, growing_string)
-			l.pos += 1
-			l.ignore()
-			return lexRvalue
-		}
-
-		if l.follow("\\\"") {
-			l.pos += 1
-			growing_string += "\""
-		} else if l.follow("\\n") {
-			l.pos += 1
-			growing_string += "\n"
-		} else if l.follow("\\t") {
-			l.pos += 1
-			growing_string += "\t"
-		} else if l.follow("\\r") {
-			l.pos += 1
-			growing_string += "\r"
-		} else if l.follow("\\\\") {
-			l.pos += 1
-			growing_string += "\\"
-		} else if l.follow("\\u") {
-			l.pos += 2
-			code := ""
-			for i := 0; i < 4; i++ {
-				c := l.peek()
-				l.pos += 1
-				if !isHexDigit(c) {
-					return l.errorf("unfinished unicode escape")
-				}
-				code = code + string(c)
-			}
-			l.pos -= 1
-			intcode, err := strconv.ParseInt(code, 16, 32)
-			if err != nil {
-				return l.errorf("invalid unicode escape: \\u" + code)
-			}
-			growing_string += string(rune(intcode))
-		} else {
-			growing_string += string(l.peek())
-		}
-
-		if l.next() == eof {
-			break
-		}
-	}
-
-	return l.errorf("unclosed string")
-}
-
-func lexKeyGroup(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 1
-	l.emit(tokenLeftBracket)
-	return lexInsideKeyGroup
-}
-
-func lexInsideKeyGroup(l *lexer) stateFn {
-	for {
-		if l.peek() == ']' {
-			if l.pos > l.start {
-				l.emit(tokenKeyGroup)
-			}
-			l.ignore()
-			l.pos += 1
-			l.emit(tokenRightBracket)
-			return lexVoid
-		}
-
-		if l.next() == eof {
-			break
-		}
-	}
-	return l.errorf("unclosed key group")
-}
-
-func lexRightBracket(l *lexer) stateFn {
-	l.ignore()
-	l.pos += 1
-	l.emit(tokenRightBracket)
-	return lexRvalue
-}
-
-func lexNumber(l *lexer) stateFn {
-	l.ignore()
-	if !l.accept("+") {
-		l.accept("-")
-	}
-	point_seen := false
-	digit_seen := false
-	for {
-		next := l.next()
-		if next == '.' {
-			point_seen = true
-		} else if isDigit(next) {
-			digit_seen = true
-		} else {
-			l.backup()
-			break
-		}
-	}
-
-	if !digit_seen {
-		return l.errorf("no digit in that number")
-	}
-	if point_seen {
-		l.emit(tokenFloat)
-	} else {
-		l.emit(tokenInteger)
-	}
-	return lexRvalue
-}
-
-func init() {
-	dateRegexp = regexp.MustCompile("^\\d{1,4}-\\d{2}-\\d{2}T\\d{2}:\\d{2}:\\d{2}Z")
-}
-
-// Entry point
-func lex(input string) (*lexer, chan token) {
-	l := &lexer{
-		input:  input,
-		tokens: make(chan token),
-	}
-	go l.run()
-	return l, l.tokens
-}

lexer_test.go

@@ -1,320 +0,0 @@
-package toml
-
-import "testing"
-
-func testFlow(t *testing.T, input string, expectedFlow []token) {
-	_, ch := lex(input)
-	for _, expected := range expectedFlow {
-		token := <-ch
-		if token != expected {
-			t.Log("compared", token, "to", expected)
-			t.Log(token.val, "<->", expected.val)
-			t.Log(token.typ, "<->", expected.typ)
-			t.FailNow()
-		}
-	}
-
-	tok, ok := <-ch
-	if ok {
-		t.Log("channel is not closed!")
-		t.Log(len(ch)+1, "tokens remaining:")
-
-		t.Log("token ->", tok)
-		for token := range ch {
-			t.Log("token ->", token)
-		}
-		t.FailNow()
-	}
-}
-
-func TestValidKeyGroup(t *testing.T) {
-	testFlow(t, "[hello world]", []token{
-		token{tokenLeftBracket, "["},
-		token{tokenKeyGroup, "hello world"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestUnclosedKeyGroup(t *testing.T) {
-	testFlow(t, "[hello world", []token{
-		token{tokenLeftBracket, "["},
-		token{tokenError, "unclosed key group"},
-	})
-}
-
-func TestComment(t *testing.T) {
-	testFlow(t, "# blahblah", []token{
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyGroupComment(t *testing.T) {
-	testFlow(t, "[hello world] # blahblah", []token{
-		token{tokenLeftBracket, "["},
-		token{tokenKeyGroup, "hello world"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestMultipleKeyGroupsComment(t *testing.T) {
-	testFlow(t, "[hello world] # blahblah\n[test]", []token{
-		token{tokenLeftBracket, "["},
-		token{tokenKeyGroup, "hello world"},
-		token{tokenRightBracket, "]"},
-		token{tokenLeftBracket, "["},
-		token{tokenKeyGroup, "test"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestBasicKey(t *testing.T) {
-	testFlow(t, "hello", []token{
-		token{tokenKey, "hello"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestBasicKeyWithUnderscore(t *testing.T) {
-	testFlow(t, "hello_hello", []token{
-		token{tokenKey, "hello_hello"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestBasicKeyWithUppercaseMix(t *testing.T) {
-	testFlow(t, "helloHELLOHello", []token{
-		token{tokenKey, "helloHELLOHello"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestBasicKeyWithInternationalCharacters(t *testing.T) {
-	testFlow(t, "héllÖ", []token{
-		token{tokenKey, "héllÖ"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestBasicKeyAndEqual(t *testing.T) {
-	testFlow(t, "hello =", []token{
-		token{tokenKey, "hello"},
-		token{tokenEqual, "="},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualStringEscape(t *testing.T) {
-	testFlow(t, "foo = \"hello\\\"\"", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenString, "hello\""},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualStringUnfinished(t *testing.T) {
-	testFlow(t, "foo = \"bar", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenError, "unclosed string"},
-	})
-}
-
-func TestKeyEqualString(t *testing.T) {
-	testFlow(t, "foo = \"bar\"", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenString, "bar"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualTrue(t *testing.T) {
-	testFlow(t, "foo = true", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenTrue, "true"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualFalse(t *testing.T) {
-	testFlow(t, "foo = false", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenFalse, "false"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestArrayNestedString(t *testing.T) {
-	testFlow(t, "a = [ [\"hello\", \"world\"] ]", []token{
-		token{tokenKey, "a"},
-		token{tokenEqual, "="},
-		token{tokenLeftBracket, "["},
-		token{tokenLeftBracket, "["},
-		token{tokenString, "hello"},
-		token{tokenComma, ","},
-		token{tokenString, "world"},
-		token{tokenRightBracket, "]"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestArrayNestedInts(t *testing.T) {
-	testFlow(t, "a = [ [42, 21], [10] ]", []token{
-		token{tokenKey, "a"},
-		token{tokenEqual, "="},
-		token{tokenLeftBracket, "["},
-		token{tokenLeftBracket, "["},
-		token{tokenInteger, "42"},
-		token{tokenComma, ","},
-		token{tokenInteger, "21"},
-		token{tokenRightBracket, "]"},
-		token{tokenComma, ","},
-		token{tokenLeftBracket, "["},
-		token{tokenInteger, "10"},
-		token{tokenRightBracket, "]"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestArrayInts(t *testing.T) {
-	testFlow(t, "a = [ 42, 21, 10, ]", []token{
-		token{tokenKey, "a"},
-		token{tokenEqual, "="},
-		token{tokenLeftBracket, "["},
-		token{tokenInteger, "42"},
-		token{tokenComma, ","},
-		token{tokenInteger, "21"},
-		token{tokenComma, ","},
-		token{tokenInteger, "10"},
-		token{tokenComma, ","},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualArrayBools(t *testing.T) {
-	testFlow(t, "foo = [true, false, true]", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenLeftBracket, "["},
-		token{tokenTrue, "true"},
-		token{tokenComma, ","},
-		token{tokenFalse, "false"},
-		token{tokenComma, ","},
-		token{tokenTrue, "true"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualArrayBoolsWithComments(t *testing.T) {
-	testFlow(t, "foo = [true, false, true] # YEAH", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenLeftBracket, "["},
-		token{tokenTrue, "true"},
-		token{tokenComma, ","},
-		token{tokenFalse, "false"},
-		token{tokenComma, ","},
-		token{tokenTrue, "true"},
-		token{tokenRightBracket, "]"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestDateRegexp(t *testing.T) {
-	if dateRegexp.FindString("1979-05-27T07:32:00Z") == "" {
-		t.Fail()
-	}
-}
-
-func TestKeyEqualDate(t *testing.T) {
-	testFlow(t, "foo = 1979-05-27T07:32:00Z", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenDate, "1979-05-27T07:32:00Z"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualNumber(t *testing.T) {
-	testFlow(t, "foo = 42", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenInteger, "42"},
-		token{tokenEOF, ""},
-	})
-
-	testFlow(t, "foo = +42", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenInteger, "+42"},
-		token{tokenEOF, ""},
-	})
-
-	testFlow(t, "foo = -42", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenInteger, "-42"},
-		token{tokenEOF, ""},
-	})
-
-	testFlow(t, "foo = 4.2", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenFloat, "4.2"},
-		token{tokenEOF, ""},
-	})
-
-	testFlow(t, "foo = +4.2", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenFloat, "+4.2"},
-		token{tokenEOF, ""},
-	})
-
-	testFlow(t, "foo = -4.2", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenFloat, "-4.2"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestMultiline(t *testing.T) {
-	testFlow(t, "foo = 42\nbar=21", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenInteger, "42"},
-		token{tokenKey, "bar"},
-		token{tokenEqual, "="},
-		token{tokenInteger, "21"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestKeyEqualStringUnicodeEscape(t *testing.T) {
-	testFlow(t, "foo = \"hello \\u2665\"", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenString, "hello ♥"},
-		token{tokenEOF, ""},
-	})
-}
-
-func TestUnicodeString(t *testing.T) {
-	testFlow(t, "foo = \"hello ♥ world\"", []token{
-		token{tokenKey, "foo"},
-		token{tokenEqual, "="},
-		token{tokenString, "hello ♥ world"},
-		token{tokenEOF, ""},
-	})
-}

localtime.go

@@ -0,0 +1,120 @@
+package toml
+
+import (
+	"fmt"
+	"strings"
+	"time"
+)
+
+// LocalDate represents a calendar day in no specific timezone.
+type LocalDate struct {
+	Year  int
+	Month int
+	Day   int
+}
+
+// AsTime converts d into a specific time instance at midnight in zone.
+func (d LocalDate) AsTime(zone *time.Location) time.Time {
+	return time.Date(d.Year, time.Month(d.Month), d.Day, 0, 0, 0, 0, zone)
+}
+
+// String returns RFC 3339 representation of d.
+func (d LocalDate) String() string {
+	return fmt.Sprintf("%04d-%02d-%02d", d.Year, d.Month, d.Day)
+}
+
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalDate) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalDate) UnmarshalText(b []byte) error {
+	res, err := parseLocalDate(b)
+	if err != nil {
+		return err
+	}
+	*d = res
+	return nil
+}
+
+// LocalTime represents a time of day of no specific day in no specific
+// timezone.
+type LocalTime struct {
+	Hour       int // Hour of the day: [0; 24[
+	Minute     int // Minute of the hour: [0; 60[
+	Second     int // Second of the minute: [0; 60[
+	Nanosecond int // Nanoseconds within the second:  [0, 1000000000[
+	Precision  int // Number of digits to display for Nanosecond.
+}
+
+// String returns RFC 3339 representation of d.
+// If d.Nanosecond and d.Precision are zero, the time won't have a nanosecond
+// component. If d.Nanosecond > 0 but d.Precision = 0, then the minimum number
+// of digits for nanoseconds is provided.
+func (d LocalTime) String() string {
+	s := fmt.Sprintf("%02d:%02d:%02d", d.Hour, d.Minute, d.Second)
+
+	if d.Precision > 0 {
+		s += fmt.Sprintf(".%09d", d.Nanosecond)[:d.Precision+1]
+	} else if d.Nanosecond > 0 {
+		// Nanoseconds are specified, but precision is not provided. Use the
+		// minimum.
+		s += strings.Trim(fmt.Sprintf(".%09d", d.Nanosecond), "0")
+	}
+
+	return s
+}
+
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalTime) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalTime) UnmarshalText(b []byte) error {
+	res, left, err := parseLocalTime(b)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
+	}
+	if err != nil {
+		return err
+	}
+	*d = res
+	return nil
+}
+
+// LocalDateTime represents a time of a specific day in no specific timezone.
+type LocalDateTime struct {
+	LocalDate
+	LocalTime
+}
+
+// AsTime converts d into a specific time instance in zone.
+func (d LocalDateTime) AsTime(zone *time.Location) time.Time {
+	return time.Date(d.Year, time.Month(d.Month), d.Day, d.Hour, d.Minute, d.Second, d.Nanosecond, zone)
+}
+
+// String returns RFC 3339 representation of d.
+func (d LocalDateTime) String() string {
+	return d.LocalDate.String() + "T" + d.LocalTime.String()
+}
+
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalDateTime) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalDateTime) UnmarshalText(data []byte) error {
+	res, left, err := parseLocalDateTime(data)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
+	}
+	if err != nil {
+		return err
+	}
+
+	*d = res
+	return nil
+}

localtime_test.go

@@ -0,0 +1,118 @@
+package toml_test
+
+import (
+	"testing"
+	"time"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
+)
+
+func TestLocalDate_AsTime(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 0, 0, 0, 0, time.UTC), cast)
+}
+
+func TestLocalDate_String(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	require.Equal(t, "2021-06-08", d.String())
+}
+
+func TestLocalDate_MarshalText(t *testing.T) {
+	d := toml.LocalDate{2021, 6, 8}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("2021-06-08"), b)
+}
+
+func TestLocalDate_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalDate{}
+	err := d.UnmarshalText([]byte("2021-06-08"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalDate{2021, 6, 8}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
+}
+
+func TestLocalTime_String(t *testing.T) {
+	d := toml.LocalTime{20, 12, 1, 2, 9}
+	require.Equal(t, "20:12:01.000000002", d.String())
+	d = toml.LocalTime{20, 12, 1, 0, 0}
+	require.Equal(t, "20:12:01", d.String())
+	d = toml.LocalTime{20, 12, 1, 0, 9}
+	require.Equal(t, "20:12:01.000000000", d.String())
+	d = toml.LocalTime{20, 12, 1, 100, 0}
+	require.Equal(t, "20:12:01.0000001", d.String())
+}
+
+func TestLocalTime_MarshalText(t *testing.T) {
+	d := toml.LocalTime{20, 12, 1, 2, 9}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("20:12:01.000000002"), b)
+}
+
+func TestLocalTime_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalTime{}
+	err := d.UnmarshalText([]byte("20:12:01.000000002"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalTime{20, 12, 1, 2, 9}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
+
+	err = d.UnmarshalText([]byte("20:12:01.000000002 bad"))
+	require.Error(t, err)
+}
+
+func TestLocalTime_RoundTrip(t *testing.T) {
+	var d struct{ A toml.LocalTime }
+	err := toml.Unmarshal([]byte("a=20:12:01.500"), &d)
+	require.NoError(t, err)
+	require.Equal(t, "20:12:01.500", d.A.String())
+}
+
+func TestLocalDateTime_AsTime(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
+	}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 20, 12, 1, 2, time.UTC), cast)
+}
+
+func TestLocalDateTime_String(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
+	}
+	require.Equal(t, "2021-06-08T20:12:01.000000002", d.String())
+}
+
+func TestLocalDateTime_MarshalText(t *testing.T) {
+	d := toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
+	}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("2021-06-08T20:12:01.000000002"), b)
+}
+
+func TestLocalDateTime_UnmarshalMarshalText(t *testing.T) {
+	d := toml.LocalDateTime{}
+	err := d.UnmarshalText([]byte("2021-06-08 20:12:01.000000002"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalDateTime{
+		toml.LocalDate{2021, 6, 8},
+		toml.LocalTime{20, 12, 1, 2, 9},
+	}, d)
+
+	err = d.UnmarshalText([]byte("what"))
+	require.Error(t, err)
+
+	err = d.UnmarshalText([]byte("2021-06-08 20:12:01.000000002 bad"))
+	require.Error(t, err)
+}

marshaler.go

@@ -0,0 +1,810 @@
+package toml
+
+import (
+	"bytes"
+	"encoding"
+	"fmt"
+	"io"
+	"math"
+	"reflect"
+	"sort"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// Marshal serializes a Go value as a TOML document.
+//
+// It is a shortcut for Encoder.Encode() with the default options.
+func Marshal(v interface{}) ([]byte, error) {
+	var buf bytes.Buffer
+	enc := NewEncoder(&buf)
+
+	err := enc.Encode(v)
+	if err != nil {
+		return nil, err
+	}
+
+	return buf.Bytes(), nil
+}
+
+// Encoder writes a TOML document to an output stream.
+type Encoder struct {
+	// output
+	w io.Writer
+
+	// global settings
+	tablesInline    bool
+	arraysMultiline bool
+	indentSymbol    string
+	indentTables    bool
+}
+
+// NewEncoder returns a new Encoder that writes to w.
+func NewEncoder(w io.Writer) *Encoder {
+	return &Encoder{
+		w:            w,
+		indentSymbol: "  ",
+	}
+}
+
+// SetTablesInline forces the encoder to emit all tables inline.
+//
+// This behavior can be controlled on an individual struct field basis with the
+// inline tag:
+//
+//   MyField `inline:"true"`
+func (enc *Encoder) SetTablesInline(inline bool) *Encoder {
+	enc.tablesInline = inline
+	return enc
+}
+
+// SetArraysMultiline forces the encoder to emit all arrays with one element per
+// line.
+//
+// This behavior can be controlled on an individual struct field basis with the multiline tag:
+//
+//   MyField `multiline:"true"`
+func (enc *Encoder) SetArraysMultiline(multiline bool) *Encoder {
+	enc.arraysMultiline = multiline
+	return enc
+}
+
+// SetIndentSymbol defines the string that should be used for indentation. The
+// provided string is repeated for each indentation level. Defaults to two
+// spaces.
+func (enc *Encoder) SetIndentSymbol(s string) *Encoder {
+	enc.indentSymbol = s
+	return enc
+}
+
+// SetIndentTables forces the encoder to intent tables and array tables.
+func (enc *Encoder) SetIndentTables(indent bool) *Encoder {
+	enc.indentTables = indent
+	return enc
+}
+
+// Encode writes a TOML representation of v to the stream.
+//
+// If v cannot be represented to TOML it returns an error.
+//
+// Encoding rules
+//
+// A top level slice containing only maps or structs is encoded as [[table
+// array]].
+//
+// All slices not matching rule 1 are encoded as [array]. As a result, any map
+// or struct they contain is encoded as an {inline table}.
+//
+// Nil interfaces and nil pointers are not supported.
+//
+// Keys in key-values always have one part.
+//
+// Intermediate tables are always printed.
+//
+// By default, strings are encoded as literal string, unless they contain either
+// a newline character or a single quote. In that case they are emitted as quoted
+// strings.
+//
+// When encoding structs, fields are encoded in order of definition, with their
+// exact name.
+//
+// Struct tags
+//
+// The following struct tags are available to tweak encoding on a per-field
+// basis:
+//
+//   toml:"foo"
+//     Changes the name of the key to use for the field to foo.
+//
+//   multiline:"true"
+//     When the field contains a string, it will be emitted as a quoted
+//     multi-line TOML string.
+//
+//   inline:"true"
+//     When the field would normally be encoded as a table, it is instead
+//     encoded as an inline table.
+func (enc *Encoder) Encode(v interface{}) error {
+	var (
+		b   []byte
+		ctx encoderCtx
+	)
+
+	ctx.inline = enc.tablesInline
+
+	if v == nil {
+		return fmt.Errorf("toml: cannot encode a nil interface")
+	}
+
+	b, err := enc.encode(b, ctx, reflect.ValueOf(v))
+	if err != nil {
+		return err
+	}
+
+	_, err = enc.w.Write(b)
+	if err != nil {
+		return fmt.Errorf("toml: cannot write: %w", err)
+	}
+
+	return nil
+}
+
+type valueOptions struct {
+	multiline bool
+}
+
+type encoderCtx struct {
+	// Current top-level key.
+	parentKey []string
+
+	// Key that should be used for a KV.
+	key string
+	// Extra flag to account for the empty string
+	hasKey bool
+
+	// Set to true to indicate that the encoder is inside a KV, so that all
+	// tables need to be inlined.
+	insideKv bool
+
+	// Set to true to skip the first table header in an array table.
+	skipTableHeader bool
+
+	// Should the next table be encoded as inline
+	inline bool
+
+	// Indentation level
+	indent int
+
+	// Options coming from struct tags
+	options valueOptions
+}
+
+func (ctx *encoderCtx) shiftKey() {
+	if ctx.hasKey {
+		ctx.parentKey = append(ctx.parentKey, ctx.key)
+		ctx.clearKey()
+	}
+}
+
+func (ctx *encoderCtx) setKey(k string) {
+	ctx.key = k
+	ctx.hasKey = true
+}
+
+func (ctx *encoderCtx) clearKey() {
+	ctx.key = ""
+	ctx.hasKey = false
+}
+
+func (ctx *encoderCtx) isRoot() bool {
+	return len(ctx.parentKey) == 0 && !ctx.hasKey
+}
+
+//nolint:cyclop,funlen
+func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	if !v.IsZero() {
+		i, ok := v.Interface().(time.Time)
+		if ok {
+			return i.AppendFormat(b, time.RFC3339), nil
+		}
+	}
+
+	if v.Type().Implements(textMarshalerType) {
+		if ctx.isRoot() {
+			return nil, fmt.Errorf("toml: type %s implementing the TextMarshaler interface cannot be a root element", v.Type())
+		}
+
+		text, err := v.Interface().(encoding.TextMarshaler).MarshalText()
+		if err != nil {
+			return nil, err
+		}
+
+		b = enc.encodeString(b, string(text), ctx.options)
+
+		return b, nil
+	}
+
+	switch v.Kind() {
+	// containers
+	case reflect.Map:
+		return enc.encodeMap(b, ctx, v)
+	case reflect.Struct:
+		return enc.encodeStruct(b, ctx, v)
+	case reflect.Slice:
+		return enc.encodeSlice(b, ctx, v)
+	case reflect.Interface:
+		if v.IsNil() {
+			return nil, fmt.Errorf("toml: encoding a nil interface is not supported")
+		}
+
+		return enc.encode(b, ctx, v.Elem())
+	case reflect.Ptr:
+		if v.IsNil() {
+			return enc.encode(b, ctx, reflect.Zero(v.Type().Elem()))
+		}
+
+		return enc.encode(b, ctx, v.Elem())
+
+	// values
+	case reflect.String:
+		b = enc.encodeString(b, v.String(), ctx.options)
+	case reflect.Float32:
+		if math.Trunc(v.Float()) == v.Float() {
+			b = strconv.AppendFloat(b, v.Float(), 'f', 1, 32)
+		} else {
+			b = strconv.AppendFloat(b, v.Float(), 'f', -1, 32)
+		}
+	case reflect.Float64:
+		if math.Trunc(v.Float()) == v.Float() {
+			b = strconv.AppendFloat(b, v.Float(), 'f', 1, 64)
+		} else {
+			b = strconv.AppendFloat(b, v.Float(), 'f', -1, 64)
+		}
+	case reflect.Bool:
+		if v.Bool() {
+			b = append(b, "true"...)
+		} else {
+			b = append(b, "false"...)
+		}
+	case reflect.Uint64, reflect.Uint32, reflect.Uint16, reflect.Uint8, reflect.Uint:
+		b = strconv.AppendUint(b, v.Uint(), 10)
+	case reflect.Int64, reflect.Int32, reflect.Int16, reflect.Int8, reflect.Int:
+		b = strconv.AppendInt(b, v.Int(), 10)
+	default:
+		return nil, fmt.Errorf("toml: cannot encode value of type %s", v.Kind())
+	}
+
+	return b, nil
+}
+
+func isNil(v reflect.Value) bool {
+	switch v.Kind() {
+	case reflect.Ptr, reflect.Interface, reflect.Map:
+		return v.IsNil()
+	default:
+		return false
+	}
+}
+
+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")
+	}
+	b = enc.indent(ctx.indent, b)
+
+	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
+	// modify the global context.
+	subctx := ctx
+	subctx.insideKv = true
+	subctx.shiftKey()
+	subctx.options = options
+
+	b, err = enc.encode(b, subctx, v)
+	if err != nil {
+		return nil, err
+	}
+
+	return b, nil
+}
+
+const literalQuote = '\''
+
+func (enc *Encoder) encodeString(b []byte, v string, options valueOptions) []byte {
+	if needsQuoting(v) {
+		return enc.encodeQuotedString(options.multiline, b, v)
+	}
+
+	return enc.encodeLiteralString(b, v)
+}
+
+func needsQuoting(v string) bool {
+	return strings.ContainsAny(v, "'\b\f\n\r\t")
+}
+
+// caller should have checked that the string does not contain new lines or ' .
+func (enc *Encoder) encodeLiteralString(b []byte, v string) []byte {
+	b = append(b, literalQuote)
+	b = append(b, v...)
+	b = append(b, literalQuote)
+
+	return b
+}
+
+//nolint:cyclop
+func (enc *Encoder) encodeQuotedString(multiline bool, b []byte, v string) []byte {
+	stringQuote := `"`
+
+	if multiline {
+		stringQuote = `"""`
+	}
+
+	b = append(b, stringQuote...)
+	if multiline {
+		b = append(b, '\n')
+	}
+
+	const (
+		hextable = "0123456789ABCDEF"
+		// U+0000 to U+0008, U+000A to U+001F, U+007F
+		nul = 0x0
+		bs  = 0x8
+		lf  = 0xa
+		us  = 0x1f
+		del = 0x7f
+	)
+
+	for _, r := range []byte(v) {
+		switch r {
+		case '\\':
+			b = append(b, `\\`...)
+		case '"':
+			b = append(b, `\"`...)
+		case '\b':
+			b = append(b, `\b`...)
+		case '\f':
+			b = append(b, `\f`...)
+		case '\n':
+			if multiline {
+				b = append(b, r)
+			} else {
+				b = append(b, `\n`...)
+			}
+		case '\r':
+			b = append(b, `\r`...)
+		case '\t':
+			b = append(b, `\t`...)
+		default:
+			switch {
+			case r >= nul && r <= bs, r >= lf && r <= us, r == del:
+				b = append(b, `\u00`...)
+				b = append(b, hextable[r>>4])
+				b = append(b, hextable[r&0x0f])
+			default:
+				b = append(b, r)
+			}
+		}
+	}
+
+	b = append(b, stringQuote...)
+
+	return b
+}
+
+// called 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...)
+}
+
+func (enc *Encoder) encodeTableHeader(ctx encoderCtx, b []byte) ([]byte, error) {
+	if len(ctx.parentKey) == 0 {
+		return b, nil
+	}
+
+	b = enc.indent(ctx.indent, b)
+
+	b = append(b, '[')
+
+	var err error
+
+	b, err = enc.encodeKey(b, ctx.parentKey[0])
+	if err != nil {
+		return nil, err
+	}
+
+	for _, k := range ctx.parentKey[1:] {
+		b = append(b, '.')
+
+		b, err = enc.encodeKey(b, k)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	b = append(b, "]\n"...)
+
+	return b, nil
+}
+
+//nolint:cyclop
+func (enc *Encoder) encodeKey(b []byte, k string) ([]byte, error) {
+	needsQuotation := false
+	cannotUseLiteral := false
+
+	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
+		}
+
+		needsQuotation = true
+	}
+
+	switch {
+	case cannotUseLiteral:
+		return enc.encodeQuotedString(false, b, k), nil
+	case needsQuotation:
+		return enc.encodeLiteralString(b, k), nil
+	default:
+		return enc.encodeUnquotedKey(b, k), nil
+	}
+}
+
+func (enc *Encoder) encodeMap(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	if v.Type().Key().Kind() != reflect.String {
+		return nil, fmt.Errorf("toml: type %s is not supported as a map key", v.Type().Key().Kind())
+	}
+
+	var (
+		t                 table
+		emptyValueOptions valueOptions
+	)
+
+	iter := v.MapRange()
+	for iter.Next() {
+		k := iter.Key().String()
+		v := iter.Value()
+
+		if isNil(v) {
+			continue
+		}
+
+		if willConvertToTableOrArrayTable(ctx, v) {
+			t.pushTable(k, v, emptyValueOptions)
+		} else {
+			t.pushKV(k, v, emptyValueOptions)
+		}
+	}
+
+	sortEntriesByKey(t.kvs)
+	sortEntriesByKey(t.tables)
+
+	return enc.encodeTable(b, ctx, t)
+}
+
+func sortEntriesByKey(e []entry) {
+	sort.Slice(e, func(i, j int) bool {
+		return e[i].Key < e[j].Key
+	})
+}
+
+type entry struct {
+	Key     string
+	Value   reflect.Value
+	Options valueOptions
+}
+
+type table struct {
+	kvs    []entry
+	tables []entry
+}
+
+func (t *table) pushKV(k string, v reflect.Value, options valueOptions) {
+	t.kvs = append(t.kvs, entry{Key: k, Value: v, Options: options})
+}
+
+func (t *table) pushTable(k string, v reflect.Value, options valueOptions) {
+	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) {
+	var t table
+
+	//nolint:godox
+	// TODO: cache this?
+	typ := v.Type()
+	for i := 0; i < typ.NumField(); i++ {
+		fieldType := typ.Field(i)
+
+		// only consider exported fields
+		if fieldType.PkgPath != "" {
+			continue
+		}
+
+		k, ok := fieldType.Tag.Lookup("toml")
+		if !ok {
+			k = fieldType.Name
+		}
+
+		// special field name to skip field
+		if k == "-" {
+			continue
+		}
+
+		f := v.Field(i)
+
+		if isNil(f) {
+			continue
+		}
+
+		options := valueOptions{
+			multiline: fieldBoolTag(fieldType, "multiline"),
+		}
+
+		inline := fieldBoolTag(fieldType, "inline")
+
+		if inline || !willConvertToTableOrArrayTable(ctx, f) {
+			t.pushKV(k, f, options)
+		} else {
+			t.pushTable(k, f, options)
+		}
+	}
+
+	return enc.encodeTable(b, ctx, t)
+}
+
+func fieldBoolTag(field reflect.StructField, tag string) bool {
+	x, ok := field.Tag.Lookup(tag)
+
+	return ok && x == "true"
+}
+
+//nolint:cyclop
+func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, error) {
+	var err error
+
+	ctx.shiftKey()
+
+	if ctx.insideKv || (ctx.inline && !ctx.isRoot()) {
+		return enc.encodeTableInline(b, ctx, t)
+	}
+
+	if !ctx.skipTableHeader {
+		b, err = enc.encodeTableHeader(ctx, b)
+		if err != nil {
+			return nil, err
+		}
+
+		if enc.indentTables && len(ctx.parentKey) > 0 {
+			ctx.indent++
+		}
+	}
+	ctx.skipTableHeader = false
+
+	for _, kv := range t.kvs {
+		ctx.setKey(kv.Key)
+
+		b, err = enc.encodeKv(b, ctx, kv.Options, kv.Value)
+		if err != nil {
+			return nil, err
+		}
+
+		b = append(b, '\n')
+	}
+
+	for _, table := range t.tables {
+		ctx.setKey(table.Key)
+
+		ctx.options = table.Options
+
+		b, err = enc.encode(b, ctx, table.Value)
+		if err != nil {
+			return nil, err
+		}
+
+		b = append(b, '\n')
+	}
+
+	return b, nil
+}
+
+func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte, error) {
+	var err error
+
+	b = append(b, '{')
+
+	first := true
+	for _, kv := range t.kvs {
+		if first {
+			first = false
+		} else {
+			b = append(b, `, `...)
+		}
+
+		ctx.setKey(kv.Key)
+
+		b, err = enc.encodeKv(b, ctx, kv.Options, kv.Value)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	if len(t.tables) > 0 {
+		panic("inline table cannot contain nested tables, online key-values")
+	}
+
+	b = append(b, "}"...)
+
+	return b, nil
+}
+
+func willConvertToTable(ctx encoderCtx, v reflect.Value) bool {
+	if !v.IsValid() {
+		return false
+	}
+	if v.Type() == timeType || v.Type().Implements(textMarshalerType) {
+		return false
+	}
+
+	t := v.Type()
+	switch t.Kind() {
+	case reflect.Map, reflect.Struct:
+		return !ctx.inline
+	case reflect.Interface:
+		return willConvertToTable(ctx, v.Elem())
+	case reflect.Ptr:
+		if v.IsNil() {
+			return false
+		}
+
+		return willConvertToTable(ctx, v.Elem())
+	default:
+		return false
+	}
+}
+
+func willConvertToTableOrArrayTable(ctx encoderCtx, v reflect.Value) bool {
+	t := v.Type()
+
+	if t.Kind() == reflect.Interface {
+		return willConvertToTableOrArrayTable(ctx, v.Elem())
+	}
+
+	if t.Kind() == reflect.Slice {
+		if v.Len() == 0 {
+			// An empty slice should be a kv = [].
+			return false
+		}
+
+		for i := 0; i < v.Len(); i++ {
+			t := willConvertToTable(ctx, v.Index(i))
+
+			if !t {
+				return false
+			}
+		}
+
+		return true
+	}
+
+	return willConvertToTable(ctx, v)
+}
+
+func (enc *Encoder) encodeSlice(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	if v.Len() == 0 {
+		b = append(b, "[]"...)
+
+		return b, nil
+	}
+
+	if willConvertToTableOrArrayTable(ctx, v) {
+		return enc.encodeSliceAsArrayTable(b, ctx, v)
+	}
+
+	return enc.encodeSliceAsArray(b, ctx, v)
+}
+
+// caller should have checked that v is a slice that only contains values that
+// encode into tables.
+func (enc *Encoder) encodeSliceAsArrayTable(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	ctx.shiftKey()
+
+	var err error
+	scratch := make([]byte, 0, 64)
+	scratch = append(scratch, "[["...)
+
+	for i, k := range ctx.parentKey {
+		if i > 0 {
+			scratch = append(scratch, '.')
+		}
+
+		scratch, err = enc.encodeKey(scratch, k)
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	scratch = append(scratch, "]]\n"...)
+	ctx.skipTableHeader = true
+
+	for i := 0; i < v.Len(); i++ {
+		b = append(b, scratch...)
+
+		b, err = enc.encode(b, ctx, v.Index(i))
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	return b, nil
+}
+
+func (enc *Encoder) encodeSliceAsArray(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	multiline := ctx.options.multiline || enc.arraysMultiline
+	separator := ", "
+
+	b = append(b, '[')
+
+	subCtx := ctx
+	subCtx.options = valueOptions{}
+
+	if multiline {
+		separator = ",\n"
+
+		b = append(b, '\n')
+
+		subCtx.indent++
+	}
+
+	var err error
+	first := true
+
+	for i := 0; i < v.Len(); i++ {
+		if first {
+			first = false
+		} else {
+			b = append(b, separator...)
+		}
+
+		if multiline {
+			b = enc.indent(subCtx.indent, b)
+		}
+
+		b, err = enc.encode(b, subCtx, v.Index(i))
+		if err != nil {
+			return nil, err
+		}
+	}
+
+	if multiline {
+		b = append(b, '\n')
+		b = enc.indent(ctx.indent, b)
+	}
+
+	b = append(b, ']')
+
+	return b, nil
+}
+
+func (enc *Encoder) indent(level int, b []byte) []byte {
+	for i := 0; i < level; i++ {
+		b = append(b, enc.indentSymbol...)
+	}
+
+	return b
+}

marshaler_test.go

@@ -0,0 +1,847 @@
+package toml_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"strings"
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+//nolint:funlen
+func TestMarshal(t *testing.T) {
+	someInt := 42
+
+	type structInline struct {
+		A interface{} `inline:"true"`
+	}
+
+	examples := []struct {
+		desc     string
+		v        interface{}
+		expected string
+		err      bool
+	}{
+		{
+			desc: "simple map and string",
+			v: map[string]string{
+				"hello": "world",
+			},
+			expected: "hello = 'world'",
+		},
+		{
+			desc: "map with new line in key",
+			v: map[string]string{
+				"hel\nlo": "world",
+			},
+			err: true,
+		},
+		{
+			desc: `map with " in key`,
+			v: map[string]string{
+				`hel"lo`: "world",
+			},
+			expected: `'hel"lo' = 'world'`,
+		},
+		{
+			desc: "map in map and string",
+			v: map[string]map[string]string{
+				"table": {
+					"hello": "world",
+				},
+			},
+			expected: `
+[table]
+hello = 'world'`,
+		},
+		{
+			desc: "map in map in map and string",
+			v: map[string]map[string]map[string]string{
+				"this": {
+					"is": {
+						"a": "test",
+					},
+				},
+			},
+			expected: `
+[this]
+[this.is]
+a = 'test'`,
+		},
+		{
+			desc: "map in map in map and string with values",
+			v: map[string]interface{}{
+				"this": map[string]interface{}{
+					"is": map[string]string{
+						"a": "test",
+					},
+					"also": "that",
+				},
+			},
+			expected: `
+[this]
+also = 'that'
+[this.is]
+a = 'test'`,
+		},
+		{
+			desc: "simple string array",
+			v: map[string][]string{
+				"array": {"one", "two", "three"},
+			},
+			expected: `array = ['one', 'two', 'three']`,
+		},
+		{
+			desc:     "empty string array",
+			v:        map[string][]string{},
+			expected: ``,
+		},
+		{
+			desc:     "map",
+			v:        map[string][]string{},
+			expected: ``,
+		},
+		{
+			desc: "nested string arrays",
+			v: map[string][][]string{
+				"array": {{"one", "two"}, {"three"}},
+			},
+			expected: `array = [['one', 'two'], ['three']]`,
+		},
+		{
+			desc: "mixed strings and nested string arrays",
+			v: map[string][]interface{}{
+				"array": {"a string", []string{"one", "two"}, "last"},
+			},
+			expected: `array = ['a string', ['one', 'two'], 'last']`,
+		},
+		{
+			desc: "array of maps",
+			v: map[string][]map[string]string{
+				"top": {
+					{"map1.1": "v1.1"},
+					{"map2.1": "v2.1"},
+				},
+			},
+			expected: `
+[[top]]
+'map1.1' = 'v1.1'
+[[top]]
+'map2.1' = 'v2.1'
+`,
+		},
+		{
+			desc: "map with two keys",
+			v: map[string]string{
+				"key1": "value1",
+				"key2": "value2",
+			},
+			expected: `
+key1 = 'value1'
+key2 = 'value2'`,
+		},
+		{
+			desc: "simple struct",
+			v: struct {
+				A string
+			}{
+				A: "foo",
+			},
+			expected: `A = 'foo'`,
+		},
+		{
+			desc: "one level of structs within structs",
+			v: struct {
+				A interface{}
+			}{
+				A: struct {
+					K1 string
+					K2 string
+				}{
+					K1: "v1",
+					K2: "v2",
+				},
+			},
+			expected: `
+[A]
+K1 = 'v1'
+K2 = 'v2'
+`,
+		},
+		{
+			desc: "structs in array with interfaces",
+			v: map[string]interface{}{
+				"root": map[string]interface{}{
+					"nested": []interface{}{
+						map[string]interface{}{"name": "Bob"},
+						map[string]interface{}{"name": "Alice"},
+					},
+				},
+			},
+			expected: `
+[root]
+[[root.nested]]
+name = 'Bob'
+[[root.nested]]
+name = 'Alice'
+`,
+		},
+		{
+			desc: "string escapes",
+			v: map[string]interface{}{
+				"a": `'"\`,
+			},
+			expected: `a = "'\"\\"`,
+		},
+		{
+			desc: "string utf8 low",
+			v: map[string]interface{}{
+				"a": "'Ę",
+			},
+			expected: `a = "'Ę"`,
+		},
+		{
+			desc: "string utf8 low 2",
+			v: map[string]interface{}{
+				"a": "'\u10A85",
+			},
+			expected: "a = \"'\u10A85\"",
+		},
+		{
+			desc: "string utf8 low 2",
+			v: map[string]interface{}{
+				"a": "'\u10A85",
+			},
+			expected: "a = \"'\u10A85\"",
+		},
+		{
+			desc: "emoji",
+			v: map[string]interface{}{
+				"a": "'😀",
+			},
+			expected: "a = \"'😀\"",
+		},
+		{
+			desc: "control char",
+			v: map[string]interface{}{
+				"a": "'\u001A",
+			},
+			expected: `a = "'\u001A"`,
+		},
+		{
+			desc: "multi-line string",
+			v: map[string]interface{}{
+				"a": "hello\nworld",
+			},
+			expected: `a = "hello\nworld"`,
+		},
+		{
+			desc: "multi-line forced",
+			v: struct {
+				A string `multiline:"true"`
+			}{
+				A: "hello\nworld",
+			},
+			expected: `A = """
+hello
+world"""`,
+		},
+		{
+			desc: "inline field",
+			v: struct {
+				A map[string]string `inline:"true"`
+				B map[string]string
+			}{
+				A: map[string]string{
+					"isinline": "yes",
+				},
+				B: map[string]string{
+					"isinline": "no",
+				},
+			},
+			expected: `
+A = {isinline = 'yes'}
+[B]
+isinline = 'no'
+`,
+		},
+		{
+			desc: "mutiline array int",
+			v: struct {
+				A []int `multiline:"true"`
+				B []int
+			}{
+				A: []int{1, 2, 3, 4},
+				B: []int{1, 2, 3, 4},
+			},
+			expected: `
+A = [
+  1,
+  2,
+  3,
+  4
+]
+B = [1, 2, 3, 4]
+`,
+		},
+		{
+			desc: "mutiline array in array",
+			v: struct {
+				A [][]int `multiline:"true"`
+			}{
+				A: [][]int{{1, 2}, {3, 4}},
+			},
+			expected: `
+A = [
+  [1, 2],
+  [3, 4]
+]
+`,
+		},
+		{
+			desc: "nil interface not supported at root",
+			v:    nil,
+			err:  true,
+		},
+		{
+			desc: "nil interface not supported in slice",
+			v: map[string]interface{}{
+				"a": []interface{}{"a", nil, 2},
+			},
+			err: true,
+		},
+		{
+			desc: "nil pointer in slice uses zero value",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{nil},
+			},
+			expected: `A = [0]`,
+		},
+		{
+			desc: "nil pointer in slice uses zero value",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{nil},
+			},
+			expected: `A = [0]`,
+		},
+		{
+			desc: "pointer in slice",
+			v: struct {
+				A []*int
+			}{
+				A: []*int{&someInt},
+			},
+			expected: `A = [42]`,
+		},
+		{
+			desc: "inline table in inline table",
+			v: structInline{
+				A: structInline{
+					A: structInline{
+						A: "hello",
+					},
+				},
+			},
+			expected: `A = {A = {A = 'hello'}}`,
+		},
+		{
+			desc: "empty slice in map",
+			v: map[string][]string{
+				"a": {},
+			},
+			expected: `a = []`,
+		},
+		{
+			desc: "map in slice",
+			v: map[string][]map[string]string{
+				"a": {{"hello": "world"}},
+			},
+			expected: `
+[[a]]
+hello = 'world'`,
+		},
+		{
+			desc: "newline in map in slice",
+			v: map[string][]map[string]string{
+				"a\n": {{"hello": "world"}},
+			},
+			err: true,
+		},
+		{
+			desc: "newline in map in slice",
+			v: map[string][]map[string]*customTextMarshaler{
+				"a": {{"hello": &customTextMarshaler{1}}},
+			},
+			err: true,
+		},
+		{
+			desc: "empty slice of empty struct",
+			v: struct {
+				A []struct{}
+			}{
+				A: []struct{}{},
+			},
+			expected: `A = []`,
+		},
+		{
+			desc: "nil field is ignored",
+			v: struct {
+				A interface{}
+			}{
+				A: nil,
+			},
+			expected: ``,
+		},
+		{
+			desc: "private fields are ignored",
+			v: struct {
+				Public  string
+				private string
+			}{
+				Public:  "shown",
+				private: "hidden",
+			},
+			expected: `Public = 'shown'`,
+		},
+		{
+			desc: "fields tagged - are ignored",
+			v: struct {
+				Public  string `toml:"-"`
+				private string
+			}{
+				Public: "hidden",
+			},
+			expected: ``,
+		},
+		{
+			desc: "nil value in map is ignored",
+			v: map[string]interface{}{
+				"A": nil,
+			},
+			expected: ``,
+		},
+		{
+			desc: "new line in table key",
+			v: map[string]interface{}{
+				"hello\nworld": 42,
+			},
+			err: true,
+		},
+		{
+			desc: "new line in parent of nested table key",
+			v: map[string]interface{}{
+				"hello\nworld": map[string]interface{}{
+					"inner": 42,
+				},
+			},
+			err: true,
+		},
+		{
+			desc: "new line in nested table key",
+			v: map[string]interface{}{
+				"parent": map[string]interface{}{
+					"in\ner": map[string]interface{}{
+						"foo": 42,
+					},
+				},
+			},
+			err: true,
+		},
+		{
+			desc: "invalid map key",
+			v:    map[int]interface{}{},
+			err:  true,
+		},
+		{
+			desc: "unhandled type",
+			v: struct {
+				A chan int
+			}{
+				A: make(chan int),
+			},
+			err: true,
+		},
+		{
+			desc: "numbers",
+			v: struct {
+				A float32
+				B uint64
+				C uint32
+				D uint16
+				E uint8
+				F uint
+				G int64
+				H int32
+				I int16
+				J int8
+				K int
+			}{
+				A: 1.1,
+				B: 42,
+				C: 42,
+				D: 42,
+				E: 42,
+				F: 42,
+				G: 42,
+				H: 42,
+				I: 42,
+				J: 42,
+				K: 42,
+			},
+			expected: `
+A = 1.1
+B = 42
+C = 42
+D = 42
+E = 42
+F = 42
+G = 42
+H = 42
+I = 42
+J = 42
+K = 42`,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			b, err := toml.Marshal(e.v)
+			if e.err {
+				require.Error(t, err)
+
+				return
+			}
+
+			require.NoError(t, err)
+			equalStringsIgnoreNewlines(t, e.expected, string(b))
+
+			// make sure the output is always valid TOML
+			defaultMap := map[string]interface{}{}
+			err = toml.Unmarshal(b, &defaultMap)
+			require.NoError(t, err)
+
+			testWithAllFlags(t, func(t *testing.T, flags int) {
+				t.Helper()
+
+				var buf bytes.Buffer
+				enc := toml.NewEncoder(&buf)
+				setFlags(enc, flags)
+
+				err := enc.Encode(e.v)
+				require.NoError(t, err)
+
+				inlineMap := map[string]interface{}{}
+				err = toml.Unmarshal(buf.Bytes(), &inlineMap)
+				require.NoError(t, err)
+
+				require.Equal(t, defaultMap, inlineMap)
+			})
+		})
+	}
+}
+
+type flagsSetters []struct {
+	name string
+	f    func(enc *toml.Encoder, flag bool) *toml.Encoder
+}
+
+var allFlags = flagsSetters{
+	{"arrays-multiline", (*toml.Encoder).SetArraysMultiline},
+	{"tables-inline", (*toml.Encoder).SetTablesInline},
+	{"indent-tables", (*toml.Encoder).SetIndentTables},
+}
+
+func setFlags(enc *toml.Encoder, flags int) {
+	for i := 0; i < len(allFlags); i++ {
+		enabled := flags&1 > 0
+		allFlags[i].f(enc, enabled)
+	}
+}
+
+func testWithAllFlags(t *testing.T, testfn func(t *testing.T, flags int)) {
+	t.Helper()
+	testWithFlags(t, 0, allFlags, testfn)
+}
+
+func testWithFlags(t *testing.T, flags int, setters flagsSetters, testfn func(t *testing.T, flags int)) {
+	t.Helper()
+
+	if len(setters) == 0 {
+		testfn(t, flags)
+
+		return
+	}
+
+	s := setters[0]
+
+	for _, enabled := range []bool{false, true} {
+		name := fmt.Sprintf("%s=%t", s.name, enabled)
+		newFlags := flags << 1
+
+		if enabled {
+			newFlags++
+		}
+
+		t.Run(name, func(t *testing.T) {
+			testWithFlags(t, newFlags, setters[1:], testfn)
+		})
+	}
+}
+
+func equalStringsIgnoreNewlines(t *testing.T, expected string, actual string) {
+	t.Helper()
+	cutset := "\n"
+	assert.Equal(t, strings.Trim(expected, cutset), strings.Trim(actual, cutset))
+}
+
+//nolint:funlen
+func TestMarshalIndentTables(t *testing.T) {
+	examples := []struct {
+		desc     string
+		v        interface{}
+		expected string
+	}{
+		{
+			desc: "one kv",
+			v: map[string]interface{}{
+				"foo": "bar",
+			},
+			expected: `foo = 'bar'`,
+		},
+		{
+			desc: "one level table",
+			v: map[string]map[string]string{
+				"foo": {
+					"one": "value1",
+					"two": "value2",
+				},
+			},
+			expected: `
+[foo]
+  one = 'value1'
+  two = 'value2'
+`,
+		},
+		{
+			desc: "two levels table",
+			v: map[string]interface{}{
+				"root": "value0",
+				"level1": map[string]interface{}{
+					"one": "value1",
+					"level2": map[string]interface{}{
+						"two": "value2",
+					},
+				},
+			},
+			expected: `
+root = 'value0'
+[level1]
+  one = 'value1'
+  [level1.level2]
+    two = 'value2'
+`,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			var buf strings.Builder
+			enc := toml.NewEncoder(&buf)
+			enc.SetIndentTables(true)
+			err := enc.Encode(e.v)
+			require.NoError(t, err)
+			equalStringsIgnoreNewlines(t, e.expected, buf.String())
+		})
+	}
+}
+
+type customTextMarshaler struct {
+	value int64
+}
+
+func (c *customTextMarshaler) MarshalText() ([]byte, error) {
+	if c.value == 1 {
+		return nil, fmt.Errorf("cannot represent 1 because this is a silly test")
+	}
+	return []byte(fmt.Sprintf("::%d", c.value)), nil
+}
+
+func TestMarshalTextMarshaler_NoRoot(t *testing.T) {
+	c := customTextMarshaler{}
+	_, err := toml.Marshal(&c)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler_Error(t *testing.T) {
+	m := map[string]interface{}{"a": &customTextMarshaler{value: 1}}
+	_, err := toml.Marshal(m)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler_ErrorInline(t *testing.T) {
+	type s struct {
+		A map[string]interface{} `inline:"true"`
+	}
+
+	d := s{
+		A: map[string]interface{}{"a": &customTextMarshaler{value: 1}},
+	}
+
+	_, err := toml.Marshal(d)
+	require.Error(t, err)
+}
+
+func TestMarshalTextMarshaler(t *testing.T) {
+	m := map[string]interface{}{"a": &customTextMarshaler{value: 2}}
+	r, err := toml.Marshal(m)
+	require.NoError(t, err)
+	equalStringsIgnoreNewlines(t, "a = '::2'", string(r))
+}
+
+type brokenWriter struct{}
+
+func (b *brokenWriter) Write([]byte) (int, error) {
+	return 0, fmt.Errorf("dead")
+}
+
+func TestEncodeToBrokenWriter(t *testing.T) {
+	w := brokenWriter{}
+	enc := toml.NewEncoder(&w)
+	err := enc.Encode(map[string]string{"hello": "world"})
+	require.Error(t, err)
+}
+
+func TestEncoderSetIndentSymbol(t *testing.T) {
+	var w strings.Builder
+	enc := toml.NewEncoder(&w)
+	enc.SetIndentTables(true)
+	enc.SetIndentSymbol(">>>")
+	err := enc.Encode(map[string]map[string]string{"parent": {"hello": "world"}})
+	require.NoError(t, err)
+	expected := `
+[parent]
+>>>hello = 'world'`
+	equalStringsIgnoreNewlines(t, expected, w.String())
+}
+
+func TestIssue436(t *testing.T) {
+	data := []byte(`{"a": [ { "b": { "c": "d" } } ]}`)
+
+	var v interface{}
+	err := json.Unmarshal(data, &v)
+	require.NoError(t, err)
+
+	var buf bytes.Buffer
+	err = toml.NewEncoder(&buf).Encode(v)
+	require.NoError(t, err)
+
+	expected := `
+[[a]]
+[a.b]
+c = 'd'
+`
+	equalStringsIgnoreNewlines(t, expected, buf.String())
+}
+
+func TestIssue424(t *testing.T) {
+	type Message1 struct {
+		Text string
+	}
+
+	type Message2 struct {
+		Text string `multiline:"true"`
+	}
+
+	msg1 := Message1{"Hello\\World"}
+	msg2 := Message2{"Hello\\World"}
+
+	toml1, err := toml.Marshal(msg1)
+	require.NoError(t, err)
+
+	toml2, err := toml.Marshal(msg2)
+	require.NoError(t, err)
+
+	msg1parsed := Message1{}
+	err = toml.Unmarshal(toml1, &msg1parsed)
+	require.NoError(t, err)
+	require.Equal(t, msg1, msg1parsed)
+
+	msg2parsed := Message2{}
+	err = toml.Unmarshal(toml2, &msg2parsed)
+	require.NoError(t, err)
+	require.Equal(t, msg2, msg2parsed)
+}
+
+func TestIssue567(t *testing.T) {
+	var m map[string]interface{}
+	err := toml.Unmarshal([]byte("A = 12:08:05"), &m)
+	require.NoError(t, err)
+	require.IsType(t, m["A"], toml.LocalTime{})
+}
+
+func TestIssue590(t *testing.T) {
+	type CustomType int
+	var cfg struct {
+		Option CustomType `toml:"option"`
+	}
+	err := toml.Unmarshal([]byte("option = 42"), &cfg)
+	require.NoError(t, err)
+}
+
+func ExampleMarshal() {
+	type MyConfig struct {
+		Version int
+		Name    string
+		Tags    []string
+	}
+
+	cfg := MyConfig{
+		Version: 2,
+		Name:    "go-toml",
+		Tags:    []string{"go", "toml"},
+	}
+
+	b, err := toml.Marshal(cfg)
+	if err != nil {
+		panic(err)
+	}
+	fmt.Println(string(b))
+
+	// Output:
+	// Version = 2
+	// Name = 'go-toml'
+	// Tags = ['go', 'toml']
+}
+
+func TestIssue571(t *testing.T) {
+	type Foo struct {
+		Float32 float32
+		Float64 float64
+	}
+
+	const closeEnough = 1e-9
+
+	foo := Foo{
+		Float32: 42,
+		Float64: 43,
+	}
+	b, err := toml.Marshal(foo)
+	require.NoError(t, err)
+
+	var foo2 Foo
+	err = toml.Unmarshal(b, &foo2)
+	require.NoError(t, err)
+
+	assert.InDelta(t, 42, foo2.Float32, closeEnough)
+	assert.InDelta(t, 43, foo2.Float64, closeEnough)
+}

parser_test.go

@@ -1,199 +1,450 @@
 package toml
 
 import (
-	"fmt"
+	"strconv"
+	"strings"
 	"testing"
-	"time"
+
+	"github.com/pelletier/go-toml/v2/internal/ast"
+	"github.com/stretchr/testify/require"
 )
 
-func assertTree(t *testing.T, tree *TomlTree, err error, ref map[string]interface{}) {
-	if err != nil {
-		t.Error("Non-nil error:", err.Error())
-		return
+//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 k, v := range ref {
-		if fmt.Sprintf("%v", tree.Get(k)) != fmt.Sprintf("%v", v) {
-			t.Log("was expecting", v, "at", k, "but got", tree.Get(k))
-			t.Error()
+
+	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))
 	}
 }
 
-func TestCreateSubTree(t *testing.T) {
-	tree := make(TomlTree)
-	tree.createSubTree("a.b.c")
-	tree.Set("a.b.c", 42)
-	if tree.Get("a.b.c") != 42 {
-		t.Fail()
+//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 TestSimpleKV(t *testing.T) {
-	tree, err := Load("a = 42")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": int64(42),
-	})
+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)))
 
-	tree, _ = Load("a = 42\nb = 21")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": int64(42),
-		"b": int64(21),
+		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 TestSimpleNumbers(t *testing.T) {
-	tree, err := Load("a = +42\nb = -21\nc = +4.2\nd = -2.1")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": int64(42),
-		"b": int64(-21),
-		"c": float64(4.2),
-		"d": float64(-2.1),
-	})
-}
+func BenchmarkParseBasicStringsEasy(b *testing.B) {
+	p := &parser{}
 
-func TestSimpleDate(t *testing.T) {
-	tree, err := Load("a = 1979-05-27T07:32:00Z")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": time.Date(1979, time.May, 27, 7, 32, 0, 0, time.UTC),
-	})
-}
+	for _, size := range []int{1, 4, 8, 16, 21} {
+		b.Run(strconv.Itoa(size), func(b *testing.B) {
+			input := []byte(`"` + strings.Repeat("A", size) + `"`)
 
-func TestSimpleString(t *testing.T) {
-	tree, err := Load("a = \"hello world\"")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": "hello world",
-	})
-}
+			b.ReportAllocs()
+			b.SetBytes(int64(len(input)))
 
-func TestStringEscapables(t *testing.T) {
-	tree, err := Load("a = \"a \\n b\"")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": "a \n b",
-	})
-
-	tree, err = Load("a = \"a \\t b\"")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": "a \t b",
-	})
-
-	tree, err = Load("a = \"a \\r b\"")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": "a \r b",
-	})
-
-	tree, err = Load("a = \"a \\\\ b\"")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": "a \\ b",
-	})
-}
-
-func TestBools(t *testing.T) {
-	tree, err := Load("a = true\nb = false")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": true,
-		"b": false,
-	})
-}
-
-func TestNestedKeys(t *testing.T) {
-	tree, err := Load("[a.b.c]\nd = 42")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a.b.c.d": int64(42),
-	})
-}
-
-func TestArrayOne(t *testing.T) {
-	tree, err := Load("a = [1]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []int64{int64(1)},
-	})
-}
-
-func TestArrayZero(t *testing.T) {
-	tree, err := Load("a = []")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []interface{}{},
-	})
-}
-
-func TestArraySimple(t *testing.T) {
-	tree, err := Load("a = [42, 21, 10]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []int64{int64(42), int64(21), int64(10)},
-	})
-
-	tree, _ = Load("a = [42, 21, 10,]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []int64{int64(42), int64(21), int64(10)},
-	})
-}
-
-func TestArrayMultiline(t *testing.T) {
-	tree, err := Load("a = [42,\n21, 10,]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []int64{int64(42), int64(21), int64(10)},
-	})
-}
-
-func TestArrayNested(t *testing.T) {
-	tree, err := Load("a = [[42, 21], [10]]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": [][]int64{[]int64{int64(42), int64(21)}, []int64{int64(10)}},
-	})
-}
-
-func TestArrayNestedStrings(t *testing.T) {
-	tree, err := Load("data = [ [\"gamma\", \"delta\"], [\"Foo\"] ]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"data": [][]string{[]string{"gamma", "delta"}, []string{"Foo"}},
-	})
-}
-
-func TestMissingValue(t *testing.T) {
-	_, err := Load("a = ")
-	if err.Error() != "expecting a value" {
-		t.Error("Bad error message:", err.Error())
+			for i := 0; i < b.N; i++ {
+				p.parseBasicString(input)
+			}
+		})
 	}
 }
 
-func TestUnterminatedArray(t *testing.T) {
-	_, err := Load("a = [1,")
-	if err.Error() != "unterminated array" {
-		t.Error("Bad error message:", err.Error())
+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())
+			}
+		})
 	}
 }
-
-func TestNewlinesInArrays(t *testing.T) {
-	tree, err := Load("a = [1,\n2,\n3]")
-	assertTree(t, tree, err, map[string]interface{}{
-		"a": []int64{int64(1), int64(2), int64(3)},
-	})
-}
-
-func TestMissingFile(t *testing.T) {
-	_, err := LoadFile("foo.toml")
-	if err.Error() != "open foo.toml: no such file or directory" {
-		t.Error("Bad error message:", err.Error())
-	}
-}
-
-func TestParseFile(t *testing.T) {
-	tree, err := LoadFile("example.toml")
-
-	assertTree(t, tree, err, map[string]interface{}{
-		"title":                   "TOML Example",
-		"owner.name":              "Tom Preston-Werner",
-		"owner.organization":      "GitHub",
-		"owner.bio":               "GitHub Cofounder & CEO\nLikes tater tots and beer.",
-		"owner.dob":               time.Date(1979, time.May, 27, 7, 32, 0, 0, time.UTC),
-		"database.server":         "192.168.1.1",
-		"database.ports":          []int64{8001, 8001, 8002},
-		"database.connection_max": 5000,
-		"database.enabled":        true,
-		"servers.alpha.ip":        "10.0.0.1",
-		"servers.alpha.dc":        "eqdc10",
-		"servers.beta.ip":         "10.0.0.2",
-		"servers.beta.dc":         "eqdc10",
-		"clients.data":            []interface{}{[]string{"gamma", "delta"}, []int64{1, 2}},
-	})
-}

scanner.go

@@ -0,0 +1,249 @@
+package toml
+
+func scanFollows(b []byte, pattern string) bool {
+	n := len(pattern)
+
+	return len(b) >= n && string(b[:n]) == pattern
+}
+
+func scanFollowsMultilineBasicStringDelimiter(b []byte) bool {
+	return scanFollows(b, `"""`)
+}
+
+func scanFollowsMultilineLiteralStringDelimiter(b []byte) bool {
+	return scanFollows(b, `'''`)
+}
+
+func scanFollowsTrue(b []byte) bool {
+	return scanFollows(b, `true`)
+}
+
+func scanFollowsFalse(b []byte) bool {
+	return scanFollows(b, `false`)
+}
+
+func scanFollowsInf(b []byte) bool {
+	return scanFollows(b, `inf`)
+}
+
+func scanFollowsNan(b []byte) bool {
+	return scanFollows(b, `nan`)
+}
+
+func scanUnquotedKey(b []byte) ([]byte, []byte) {
+	// unquoted-key = 1*( ALPHA / DIGIT / %x2D / %x5F ) ; A-Z / a-z / 0-9 / - / _
+	for i := 0; i < len(b); i++ {
+		if !isUnquotedKeyChar(b[i]) {
+			return b[:i], b[i:]
+		}
+	}
+
+	return b, b[len(b):]
+}
+
+func isUnquotedKeyChar(r byte) bool {
+	return (r >= 'A' && r <= 'Z') || (r >= 'a' && r <= 'z') || (r >= '0' && r <= '9') || r == '-' || r == '_'
+}
+
+func scanLiteralString(b []byte) ([]byte, []byte, error) {
+	// literal-string = apostrophe *literal-char apostrophe
+	// apostrophe = %x27 ; ' apostrophe
+	// literal-char = %x09 / %x20-26 / %x28-7E / non-ascii
+	for i := 1; i < len(b); {
+		switch b[i] {
+		case '\'':
+			return b[:i+1], b[i+1:], nil
+		case '\n':
+			return nil, nil, newDecodeError(b[i:i+1], "literal strings cannot have new lines")
+		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
+	}
+
+	return nil, nil, newDecodeError(b[len(b):], "unterminated literal string")
+}
+
+func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
+	// ml-literal-string = ml-literal-string-delim [ newline ] ml-literal-body
+	// ml-literal-string-delim
+	// ml-literal-string-delim = 3apostrophe
+	// ml-literal-body = *mll-content *( mll-quotes 1*mll-content ) [ mll-quotes ]
+	//
+	// mll-content = mll-char / newline
+	// mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
+	// mll-quotes = 1*2apostrophe
+	for i := 3; i < len(b); {
+		if scanFollowsMultilineLiteralStringDelimiter(b[i:]) {
+			i += 3
+
+			// At that point we found 3 apostrophe, and i is the
+			// index of the byte after the third one. The scanner
+			// needs to be eager, because there can be an extra 2
+			// apostrophe that can be accepted at the end of the
+			// string.
+
+			if i >= len(b) || b[i] != '\'' {
+				return b[:i], b[i:], nil
+			}
+			i++
+
+			if i >= len(b) || b[i] != '\'' {
+				return b[:i], b[i:], nil
+			}
+			i++
+
+			if i < len(b) && b[i] == '\'' {
+				return nil, nil, newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+			}
+
+			return b[:i], b[i:], nil
+		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
+	}
+
+	return nil, nil, newDecodeError(b[len(b):], `multiline literal string not terminated by '''`)
+}
+
+func scanWindowsNewline(b []byte) ([]byte, []byte, error) {
+	const lenCRLF = 2
+	if len(b) < lenCRLF {
+		return nil, nil, newDecodeError(b, "windows new line expected")
+	}
+
+	if b[1] != '\n' {
+		return nil, nil, newDecodeError(b, `windows new line should be \r\n`)
+	}
+
+	return b[:lenCRLF], b[lenCRLF:], nil
+}
+
+func scanWhitespace(b []byte) ([]byte, []byte) {
+	for i := 0; i < len(b); i++ {
+		switch b[i] {
+		case ' ', '\t':
+			continue
+		default:
+			return b[:i], b[i:]
+		}
+	}
+
+	return b, b[len(b):]
+}
+
+//nolint:unparam
+func scanComment(b []byte) ([]byte, []byte, error) {
+	// comment-start-symbol = %x23 ; #
+	// non-ascii = %x80-D7FF / %xE000-10FFFF
+	// non-eol = %x09 / %x20-7F / non-ascii
+	//
+	// comment = comment-start-symbol *non-eol
+
+	for i := 1; i < len(b); {
+		if b[i] == '\n' {
+			return b[:i], b[i:], nil
+		}
+		if b[i] == '\r' {
+			if i+1 < len(b) && b[i+1] == '\n' {
+				return b[:i+1], b[i+1:], nil
+			}
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character in comment")
+		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, nil, newDecodeError(b[i:i+1], "invalid character in comment")
+		}
+
+		i += size
+	}
+
+	return b, b[len(b):], nil
+}
+
+func scanBasicString(b []byte) ([]byte, bool, []byte, error) {
+	// basic-string = quotation-mark *basic-char quotation-mark
+	// quotation-mark = %x22            ; "
+	// basic-char = basic-unescaped / escaped
+	// basic-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+	// escaped = escape escape-seq-char
+	escaped := false
+	i := 1
+
+	for ; i < len(b); i++ {
+		switch b[i] {
+		case '"':
+			return b[:i+1], escaped, b[i+1:], nil
+		case '\n', '\r':
+			return nil, escaped, nil, newDecodeError(b[i:i+1], "basic strings cannot have new lines")
+		case '\\':
+			if len(b) < i+2 {
+				return nil, escaped, nil, newDecodeError(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 "`)
+}
+
+func scanMultilineBasicString(b []byte) ([]byte, bool, []byte, error) {
+	// ml-basic-string = ml-basic-string-delim [ newline ] ml-basic-body
+	// ml-basic-string-delim
+	// ml-basic-string-delim = 3quotation-mark
+	// ml-basic-body = *mlb-content *( mlb-quotes 1*mlb-content ) [ mlb-quotes ]
+	//
+	// mlb-content = mlb-char / newline / mlb-escaped-nl
+	// mlb-char = mlb-unescaped / escaped
+	// mlb-quotes = 1*2quotation-mark
+	// mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+	// mlb-escaped-nl = escape ws newline *( wschar / newline )
+
+	escaped := false
+	i := 3
+
+	for ; i < len(b); i++ {
+		switch b[i] {
+		case '"':
+			if scanFollowsMultilineBasicStringDelimiter(b[i:]) {
+				i += 3
+
+				// At that point we found 3 apostrophe, and i is the
+				// index of the byte after the third one. The scanner
+				// needs to be eager, because there can be an extra 2
+				// apostrophe that can be accepted at the end of the
+				// string.
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i < len(b) && b[i] == '"' {
+					return nil, escaped, nil, newDecodeError(b[i-3:i+1], `""" not allowed in multiline basic string`)
+				}
+
+				return b[:i], escaped, b[i:], nil
+			}
+		case '\\':
+			if len(b) < i+2 {
+				return nil, escaped, nil, newDecodeError(b[len(b):], "need a character after \\")
+			}
+			escaped = true
+			i++ // skip the next character
+		}
+	}
+
+	return nil, escaped, nil, newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
+}

strict.go

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

testsuite/add.go

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

testsuite/json.go

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

testsuite/parser.go

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

testsuite/rm.go

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

testsuite/testsuite.go

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

toml.abnf

@@ -0,0 +1,243 @@
+;; This document describes TOML's syntax, using the ABNF format (defined in
+;; RFC 5234 -- https://www.ietf.org/rfc/rfc5234.txt).
+;;
+;; All valid TOML documents will match this description, however certain
+;; invalid documents would need to be rejected as per the semantics described
+;; in the supporting text description.
+
+;; It is possible to try this grammar interactively, using instaparse.
+;;     http://instaparse.mojombo.com/
+;;
+;; To do so, in the lower right, click on Options and change `:input-format` to
+;; ':abnf'. Then paste this entire ABNF document into the grammar entry box
+;; (above the options). Then you can type or paste a sample TOML document into
+;; the beige box on the left. Tada!
+
+;; Overall Structure
+
+toml = expression *( newline expression )
+
+expression =  ws [ comment ]
+expression =/ ws keyval ws [ comment ]
+expression =/ ws table ws [ comment ]
+
+;; Whitespace
+
+ws = *wschar
+wschar =  %x20  ; Space
+wschar =/ %x09  ; Horizontal tab
+
+;; Newline
+
+newline =  %x0A     ; LF
+newline =/ %x0D.0A  ; CRLF
+
+;; Comment
+
+comment-start-symbol = %x23 ; #
+non-ascii = %x80-D7FF / %xE000-10FFFF
+non-eol = %x09 / %x20-7F / non-ascii
+
+comment = comment-start-symbol *non-eol
+
+;; Key-Value pairs
+
+keyval = key keyval-sep val
+
+key = simple-key / dotted-key
+simple-key = quoted-key / unquoted-key
+
+unquoted-key = 1*( ALPHA / DIGIT / %x2D / %x5F ) ; A-Z / a-z / 0-9 / - / _
+quoted-key = basic-string / literal-string
+dotted-key = simple-key 1*( dot-sep simple-key )
+
+dot-sep   = ws %x2E ws  ; . Period
+keyval-sep = ws %x3D ws ; =
+
+val = string / boolean / array / inline-table / date-time / float / integer
+
+;; String
+
+string = ml-basic-string / basic-string / ml-literal-string / literal-string
+
+;; Basic String
+
+basic-string = quotation-mark *basic-char quotation-mark
+
+quotation-mark = %x22            ; "
+
+basic-char = basic-unescaped / escaped
+basic-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+escaped = escape escape-seq-char
+
+escape = %x5C                   ; \
+escape-seq-char =  %x22         ; "    quotation mark  U+0022
+escape-seq-char =/ %x5C         ; \    reverse solidus U+005C
+escape-seq-char =/ %x62         ; b    backspace       U+0008
+escape-seq-char =/ %x66         ; f    form feed       U+000C
+escape-seq-char =/ %x6E         ; n    line feed       U+000A
+escape-seq-char =/ %x72         ; r    carriage return U+000D
+escape-seq-char =/ %x74         ; t    tab             U+0009
+escape-seq-char =/ %x75 4HEXDIG ; uXXXX                U+XXXX
+escape-seq-char =/ %x55 8HEXDIG ; UXXXXXXXX            U+XXXXXXXX
+
+;; Multiline Basic String
+
+ml-basic-string = ml-basic-string-delim [ newline ] ml-basic-body
+                  ml-basic-string-delim
+ml-basic-string-delim = 3quotation-mark
+ml-basic-body = *mlb-content *( mlb-quotes 1*mlb-content ) [ mlb-quotes ]
+
+mlb-content = mlb-char / newline / mlb-escaped-nl
+mlb-char = mlb-unescaped / escaped
+mlb-quotes = 1*2quotation-mark
+mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
+mlb-escaped-nl = escape ws newline *( wschar / newline )
+
+;; Literal String
+
+literal-string = apostrophe *literal-char apostrophe
+
+apostrophe = %x27 ; ' apostrophe
+
+literal-char = %x09 / %x20-26 / %x28-7E / non-ascii
+
+;; Multiline Literal String
+
+ml-literal-string = ml-literal-string-delim [ newline ] ml-literal-body
+                    ml-literal-string-delim
+ml-literal-string-delim = 3apostrophe
+ml-literal-body = *mll-content *( mll-quotes 1*mll-content ) [ mll-quotes ]
+
+mll-content = mll-char / newline
+mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
+mll-quotes = 1*2apostrophe
+
+;; Integer
+
+integer = dec-int / hex-int / oct-int / bin-int
+
+minus = %x2D                       ; -
+plus = %x2B                        ; +
+underscore = %x5F                  ; _
+digit1-9 = %x31-39                 ; 1-9
+digit0-7 = %x30-37                 ; 0-7
+digit0-1 = %x30-31                 ; 0-1
+
+hex-prefix = %x30.78               ; 0x
+oct-prefix = %x30.6F               ; 0o
+bin-prefix = %x30.62               ; 0b
+
+dec-int = [ minus / plus ] unsigned-dec-int
+unsigned-dec-int = DIGIT / digit1-9 1*( DIGIT / underscore DIGIT )
+
+hex-int = hex-prefix HEXDIG *( HEXDIG / underscore HEXDIG )
+oct-int = oct-prefix digit0-7 *( digit0-7 / underscore digit0-7 )
+bin-int = bin-prefix digit0-1 *( digit0-1 / underscore digit0-1 )
+
+;; Float
+
+float = float-int-part ( exp / frac [ exp ] )
+float =/ special-float
+
+float-int-part = dec-int
+frac = decimal-point zero-prefixable-int
+decimal-point = %x2E               ; .
+zero-prefixable-int = DIGIT *( DIGIT / underscore DIGIT )
+
+exp = "e" float-exp-part
+float-exp-part = [ minus / plus ] zero-prefixable-int
+
+special-float = [ minus / plus ] ( inf / nan )
+inf = %x69.6e.66  ; inf
+nan = %x6e.61.6e  ; nan
+
+;; Boolean
+
+boolean = true / false
+
+true    = %x74.72.75.65     ; true
+false   = %x66.61.6C.73.65  ; false
+
+;; Date and Time (as defined in RFC 3339)
+
+date-time      = offset-date-time / local-date-time / local-date / local-time
+
+date-fullyear  = 4DIGIT
+date-month     = 2DIGIT  ; 01-12
+date-mday      = 2DIGIT  ; 01-28, 01-29, 01-30, 01-31 based on month/year
+time-delim     = "T" / %x20 ; T, t, or space
+time-hour      = 2DIGIT  ; 00-23
+time-minute    = 2DIGIT  ; 00-59
+time-second    = 2DIGIT  ; 00-58, 00-59, 00-60 based on leap second rules
+time-secfrac   = "." 1*DIGIT
+time-numoffset = ( "+" / "-" ) time-hour ":" time-minute
+time-offset    = "Z" / time-numoffset
+
+partial-time   = time-hour ":" time-minute ":" time-second [ time-secfrac ]
+full-date      = date-fullyear "-" date-month "-" date-mday
+full-time      = partial-time time-offset
+
+;; Offset Date-Time
+
+offset-date-time = full-date time-delim full-time
+
+;; Local Date-Time
+
+local-date-time = full-date time-delim partial-time
+
+;; Local Date
+
+local-date = full-date
+
+;; Local Time
+
+local-time = partial-time
+
+;; Array
+
+array = array-open [ array-values ] ws-comment-newline array-close
+
+array-open =  %x5B ; [
+array-close = %x5D ; ]
+
+array-values =  ws-comment-newline val ws-comment-newline array-sep array-values
+array-values =/ ws-comment-newline val ws-comment-newline [ array-sep ]
+
+array-sep = %x2C  ; , Comma
+
+ws-comment-newline = *( wschar / [ comment ] newline )
+
+;; Table
+
+table = std-table / array-table
+
+;; Standard Table
+
+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
+
+;; Inline Table
+
+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 ]
+
+;; Array Table
+
+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
+
+;; Built-in ABNF terms, reproduced here for clarity
+
+ALPHA = %x41-5A / %x61-7A ; A-Z / a-z
+DIGIT = %x30-39 ; 0-9
+HEXDIG = DIGIT / "A" / "B" / "C" / "D" / "E" / "F"

toml.go

@@ -1,114 +0,0 @@
-// TOML markup language parser.
-//
-// This version supports the specification as described in
-// https://github.com/mojombo/toml/tree/e3656ad493400895f4460f1244a25f8f8e31a32a
-package toml
-
-import (
-	"errors"
-	"io/ioutil"
-	"runtime"
-	"strings"
-)
-
-// Definition of a TomlTree.
-// This is the result of the parsing of a TOML file.
-type TomlTree map[string]interface{}
-
-// Keys returns the keys of the toplevel tree.
-// Warning: this is a costly operation.
-func (t *TomlTree) Keys() []string {
-	keys := make([]string, 0)
-	mp := (map[string]interface{})(*t)
-	for k, _ := range mp {
-		keys = append(keys, k)
-	}
-	return keys
-}
-
-// Get the value at key in the TomlTree.
-// Key is a dot-separated path (e.g. a.b.c).
-// Returns nil if the path does not exist in the tree.
-func (t *TomlTree) Get(key string) interface{} {
-	subtree := t
-	keys := strings.Split(key, ".")
-	for _, intermediate_key := range keys[:len(keys)-1] {
-		_, exists := (*subtree)[intermediate_key]
-		if !exists {
-			return nil
-		}
-		subtree = (*subtree)[intermediate_key].(*TomlTree)
-	}
-	return (*subtree)[keys[len(keys)-1]]
-}
-
-// Same as Get but with a default value
-func (t *TomlTree) GetDefault(key string, def interface{}) interface{} {
-    val := t.Get(key)
-    if val == nil {
-        return def
-    }
-    return val;
-}
-
-// Set an element in the tree.
-// Key is a dot-separated path (e.g. a.b.c).
-// Creates all necessary intermediates trees, if needed.
-func (t *TomlTree) Set(key string, value interface{}) {
-	subtree := t
-	keys := strings.Split(key, ".")
-	for _, intermediate_key := range keys[:len(keys)-1] {
-		_, exists := (*subtree)[intermediate_key]
-		if !exists {
-			var new_tree TomlTree = make(TomlTree)
-			(*subtree)[intermediate_key] = &new_tree
-		}
-		subtree = (*subtree)[intermediate_key].(*TomlTree)
-	}
-	(*subtree)[keys[len(keys)-1]] = value
-}
-
-// createSubTree takes a tree and a key and create the necessary intermediate
-// subtrees to create a subtree at that point. In-place.
-//
-// e.g. passing a.b.c will create (assuming tree is empty) tree[a], tree[a][b]
-// and tree[a][b][c]
-func (t *TomlTree) createSubTree(key string) {
-	subtree := t
-	for _, intermediate_key := range strings.Split(key, ".") {
-		_, exists := (*subtree)[intermediate_key]
-		if !exists {
-			var new_tree TomlTree = make(TomlTree)
-			(*subtree)[intermediate_key] = &new_tree
-		}
-		subtree = ((*subtree)[intermediate_key]).(*TomlTree)
-	}
-}
-
-// Create a TomlTree from a string.
-func Load(content string) (tree *TomlTree, err error) {
-	defer func() {
-		if r := recover(); r != nil {
-			if _, ok := r.(runtime.Error); ok {
-				panic(r)
-			}
-			err = errors.New(r.(string))
-		}
-	}()
-	_, flow := lex(content)
-	tree = parse(flow)
-	return
-}
-
-// Create a TomlTree from a file.
-func LoadFile(path string) (tree *TomlTree, err error) {
-	buff, ferr := ioutil.ReadFile(path)
-	if ferr != nil {
-		err = ferr
-	} else {
-		s := string(buff)
-		tree, err = Load(s)
-	}
-
-	return
-}

toml_test.go

@@ -1 +0,0 @@
-package toml

toml_testgen_support_test.go

@@ -0,0 +1,58 @@
+//go:generate go run ./cmd/tomltestgen/main.go -o toml_testgen_test.go
+
+// This is a support file for toml_testgen_test.go
+package toml_test
+
+import (
+	"encoding/json"
+	"testing"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/testsuite"
+	"github.com/stretchr/testify/require"
+)
+
+func testgenInvalid(t *testing.T, input string) {
+	t.Helper()
+	t.Logf("Input TOML:\n%s", input)
+
+	doc := map[string]interface{}{}
+	err := testsuite.Unmarshal([]byte(input), &doc)
+
+	if err == nil {
+		out, err := json.Marshal(doc)
+		if err != nil {
+			panic("could not marshal map to json")
+		}
+		t.Log("JSON output from unmarshal:", string(out))
+		t.Fatalf("test did not fail")
+	}
+}
+
+func testgenValid(t *testing.T, input string, jsonRef string) {
+	t.Helper()
+	t.Logf("Input TOML:\n%s", input)
+
+	// TODO: change this to interface{}
+	var doc map[string]interface{}
+
+	err := testsuite.Unmarshal([]byte(input), &doc)
+	if err != nil {
+		if de, ok := err.(*toml.DecodeError); ok {
+			t.Logf("%s\n%s", err, de)
+		}
+		t.Fatalf("failed parsing toml: %s", err)
+	}
+	j, err := testsuite.ValueToTaggedJSON(doc)
+	require.NoError(t, err)
+
+	var ref interface{}
+	err = json.Unmarshal([]byte(jsonRef), &ref)
+	require.NoError(t, err)
+
+	var actual interface{}
+	err = json.Unmarshal([]byte(j), &actual)
+	require.NoError(t, err)
+
+	testsuite.CmpJSON(t, "", ref, actual)
+}

types.go

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

utf8.go

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