Make all nodes contain Raw

Also has the perks of reducing the overhead of FindByIndex:

```
name                                old time/op    new time/op    delta
UnmarshalDataset/config-32            17.0ms ± 1%    17.0ms ± 1%    ~     (p=1.000 n=5+5)
UnmarshalDataset/canada-32            71.6ms ± 1%    71.4ms ± 1%    ~     (p=1.000 n=5+5)
UnmarshalDataset/citm_catalog-32      24.2ms ± 3%    23.5ms ± 2%  -3.03%  (p=0.032 n=5+5)
UnmarshalDataset/twitter-32           9.37ms ± 1%    9.09ms ± 2%  -2.97%  (p=0.032 n=5+5)
UnmarshalDataset/code-32              75.4ms ± 2%    74.9ms ± 0%    ~     (p=0.222 n=5+5)
UnmarshalDataset/example-32            147µs ±10%     136µs ± 1%  -7.14%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/struct-32     512ns ± 2%     500ns ± 0%  -2.35%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/map-32        721ns ± 2%     702ns ± 1%  -2.68%  (p=0.008 n=5+5)
Unmarshal/ReferenceFile/struct-32     40.1µs ± 0%    39.6µs ± 0%  -1.30%  (p=0.008 n=5+5)
Unmarshal/ReferenceFile/map-32        62.3µs ± 1%    60.6µs ± 0%  -2.83%  (p=0.008 n=5+5)
Unmarshal/HugoFrontMatter-32          10.8µs ± 1%    10.5µs ± 1%  -2.86%  (p=0.008 n=5+5)

name                                old speed      new speed      delta
UnmarshalDataset/config-32          61.8MB/s ± 1%  61.8MB/s ± 1%    ~     (p=1.000 n=5+5)
UnmarshalDataset/canada-32          30.8MB/s ± 1%  30.8MB/s ± 1%    ~     (p=1.000 n=5+5)
UnmarshalDataset/citm_catalog-32    23.0MB/s ± 3%  23.8MB/s ± 2%  +3.09%  (p=0.032 n=5+5)
UnmarshalDataset/twitter-32         47.2MB/s ± 1%  48.6MB/s ± 2%  +3.09%  (p=0.032 n=5+5)
UnmarshalDataset/code-32            35.6MB/s ± 2%  35.9MB/s ± 0%    ~     (p=0.222 n=5+5)
UnmarshalDataset/example-32         55.3MB/s ±10%  59.4MB/s ± 1%  +7.36%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/struct-32  21.5MB/s ± 2%  22.0MB/s ± 0%  +2.41%  (p=0.008 n=5+5)
Unmarshal/SimpleDocument/map-32     15.2MB/s ± 2%  15.7MB/s ± 1%  +2.74%  (p=0.008 n=5+5)
Unmarshal/ReferenceFile/struct-32    131MB/s ± 0%   132MB/s ± 0%  +1.31%  (p=0.008 n=5+5)
Unmarshal/ReferenceFile/map-32      84.1MB/s ± 1%  86.6MB/s ± 0%  +2.91%  (p=0.008 n=5+5)
Unmarshal/HugoFrontMatter-32        50.6MB/s ± 1%  52.1MB/s ± 1%  +2.93%  (p=0.008 n=5+5)

name                                old alloc/op   new alloc/op   delta
UnmarshalDataset/config-32            5.86MB ± 0%    5.86MB ± 0%    ~     (p=0.579 n=5+5)
UnmarshalDataset/canada-32            83.0MB ± 0%    83.0MB ± 0%    ~     (p=0.651 n=5+5)
UnmarshalDataset/citm_catalog-32      34.7MB ± 0%    34.7MB ± 0%    ~     (p=0.548 n=5+5)
UnmarshalDataset/twitter-32           12.7MB ± 0%    12.7MB ± 0%    ~     (p=1.000 n=5+5)
UnmarshalDataset/code-32              22.2MB ± 0%    22.2MB ± 0%    ~     (p=0.841 n=5+5)
UnmarshalDataset/example-32            186kB ± 0%     186kB ± 0%    ~     (p=0.111 n=5+5)
Unmarshal/SimpleDocument/struct-32      805B ± 0%      805B ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/map-32       1.13kB ± 0%    1.13kB ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/struct-32     20.9kB ± 0%    20.9kB ± 0%    ~     (p=0.643 n=5+5)
Unmarshal/ReferenceFile/map-32        38.3kB ± 0%    38.3kB ± 0%    ~     (p=0.397 n=5+5)
Unmarshal/HugoFrontMatter-32          7.44kB ± 0%    7.44kB ± 0%    ~     (all equal)

name                                old allocs/op  new allocs/op  delta
UnmarshalDataset/config-32              227k ± 0%      227k ± 0%    ~     (p=1.000 n=5+5)
UnmarshalDataset/canada-32              782k ± 0%      782k ± 0%    ~     (all equal)
UnmarshalDataset/citm_catalog-32        192k ± 0%      192k ± 0%    ~     (p=0.968 n=4+5)
UnmarshalDataset/twitter-32            56.9k ± 0%     56.9k ± 0%    ~     (p=0.429 n=4+5)
UnmarshalDataset/code-32               1.05M ± 0%     1.05M ± 0%    ~     (p=0.556 n=4+5)
UnmarshalDataset/example-32            1.36k ± 0%     1.36k ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/struct-32      9.00 ± 0%      9.00 ± 0%    ~     (all equal)
Unmarshal/SimpleDocument/map-32         13.0 ± 0%      13.0 ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/struct-32        183 ± 0%       183 ± 0%    ~     (all equal)
Unmarshal/ReferenceFile/map-32           642 ± 0%       642 ± 0%    ~     (all equal)
Unmarshal/HugoFrontMatter-32             141 ± 0%       141 ± 0%    ~     (all equal)
```

Fixes #807

.gitattributes

@@ -0,0 +1,4 @@
+* text=auto
+
+benchmark/benchmark.toml text eol=lf
+testdata/** text eol=lf

.github/dependabot.yml

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

.github/release.yml

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

.github/workflows/codeql-analysis.yml

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

.github/workflows/coverage.yml

@@ -1,23 +1,20 @@
 name: coverage
 on:
-  push:
-    branches:
-      - v2
   pull_request:
     branches:
       - v2
 
 jobs:
   report:
-    runs-on: 'ubuntu-latest'
+    runs-on: "ubuntu-latest"
     name: report
     steps:
-      - uses: actions/checkout@master
+      - uses: actions/checkout@v3
         with:
           fetch-depth: 0
       - name: Setup go
         uses: actions/setup-go@master
         with:
-          go-version: 1.16
+          go-version: 1.19
       - name: Run tests with coverage
         run: ./ci.sh coverage -d "${GITHUB_BASE_REF-HEAD}"

.github/workflows/release.yml

@@ -0,0 +1,39 @@
+name: release
+on:
+  push:
+    tags:
+      - "v2.*"
+  workflow_call:
+    inputs:
+      args:
+        description: "Extra arguments to pass goreleaser"
+        default: ""
+        required: false
+        type: string
+
+jobs:
+  release:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+      - name: Set up Go
+        uses: actions/setup-go@v2
+        with:
+          go-version: 1.19
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v2
+        with:
+          registry: ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Run GoReleaser
+        uses: goreleaser/goreleaser-action@v3
+        with:
+          distribution: goreleaser
+          version: latest
+          args: release ${{ inputs.args }} --rm-dist
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/workflow.yml

@@ -12,14 +12,21 @@ jobs:
     strategy:
       matrix:
         os: [ 'ubuntu-latest', 'windows-latest', 'macos-latest']
-        go: [ '1.15', '1.16' ]
+        go: [ '1.18', '1.19' ]
     runs-on: ${{ matrix.os }}
     name: ${{ matrix.go }}/${{ matrix.os }}
     steps:
-      - uses: actions/checkout@master
+      - uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
       - name: Setup go ${{ matrix.go }}
         uses: actions/setup-go@master
         with:
           go-version: ${{ matrix.go }}
       - name: Run unit tests
         run: go test -race ./...
+  release-check:
+    if: ${{ github.ref != 'refs/heads/v2' }}
+    uses: pelletier/go-toml/.github/workflows/release.yml@v2
+    with:
+      args: --snapshot

.gitignore

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

.golangci.toml

@@ -60,7 +60,7 @@ enable = [
     # "nlreturn",
     "noctx",
     "nolintlint",
-    "paralleltest",
+    #"paralleltest",
     "prealloc",
     "predeclared",
     "revive",

.goreleaser.yaml

@@ -0,0 +1,123 @@
+before:
+  hooks:
+    - go mod tidy
+    - go fmt ./...
+    - go test ./...
+builds:
+  - id: tomll
+    main: ./cmd/tomll
+    binary: tomll
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - linux_arm64
+      - linux_arm
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - darwin_amd64
+      - darwin_arm64
+  - id: tomljson
+    main: ./cmd/tomljson
+    binary: tomljson
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - linux_arm64
+      - linux_arm
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - darwin_amd64
+      - darwin_arm64
+  - id: jsontoml
+    main: ./cmd/jsontoml
+    binary: jsontoml
+    env:
+      - CGO_ENABLED=0
+    flags:
+      - -trimpath
+    ldflags:
+      - -X main.version={{.Version}} -X main.commit={{.Commit}} -X main.date={{.CommitDate}}
+    mod_timestamp: '{{ .CommitTimestamp }}'
+    targets:
+      - linux_amd64
+      - linux_arm64
+      - linux_arm
+      - windows_amd64
+      - windows_arm64
+      - windows_arm
+      - darwin_amd64
+      - darwin_arm64
+universal_binaries:
+  - id: tomll
+    replace: true
+    name_template: tomll
+  - id: tomljson
+    replace: true
+    name_template: tomljson
+  - id: jsontoml
+    replace: true
+    name_template: jsontoml
+archives:
+- id: jsontoml
+  format: tar.xz
+  builds:
+    - jsontoml
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+- id: tomljson
+  format: tar.xz
+  builds:
+    - tomljson
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+- id: tomll
+  format: tar.xz
+  builds:
+    - tomll
+  files:
+  - none*
+  name_template: "{{ .Binary }}_{{.Version}}_{{ .Os }}_{{ .Arch }}"
+dockers:
+  - id: tools
+    goos: linux
+    goarch: amd64
+    ids:
+      - jsontoml
+      - tomljson
+      - tomll
+    image_templates:
+      - "ghcr.io/pelletier/go-toml:latest"
+      - "ghcr.io/pelletier/go-toml:{{ .Tag }}"
+      - "ghcr.io/pelletier/go-toml:v{{ .Major }}"
+    skip_push: false
+checksum:
+  name_template: 'sha256sums.txt'
+snapshot:
+  name_template: "{{ incpatch .Version }}-next"
+release:
+  github:
+    owner: pelletier
+    name: go-toml
+  draft: true
+  prerelease: auto
+  mode: replace
+changelog:
+  use: github-native
+announce:
+  skip: true

CONTRIBUTING.md

@@ -155,6 +155,8 @@ Checklist:
 - Does not introduce backward-incompatible changes (unless discussed).
 - Has relevant doc changes.
 - Benchstat does not show performance regression.
+- Pull request is [labeled appropriately][pr-labels].
+- Title will be understandable in the changelog.
 
 1. Merge using "squash and merge".
 2. Make sure to edit the commit message to keep all the useful information
@@ -163,13 +165,25 @@ Checklist:
 
 ### New release
 
-1. Go to [releases][releases]. Click on "X commits to master since this
+1. Decide on the next version number. Use semver.
-   release".
+2. Generate release notes using [`gh`][gh]. Example:
-2. Make note of all the changes. Look for backward incompatible changes,
+```
-   new features, and bug fixes.
+$ gh api -X POST \
-3. Pick the new version using the above and semver.
+  -F tag_name='v2.0.0-beta.5' \
-4. Create a [new release][new-release].
+  -F target_commitish='v2' \
-5. Follow the same format as [1.1.0][release-110].
+  -F previous_tag_name='v2.0.0-beta.4' \
+  --jq '.body' \
+  repos/pelletier/go-toml/releases/generate-notes
+```
+3. Look for "Other changes". That would indicate a pull request not labeled
+   properly. Tweak labels and pull request titles until changelog looks good for
+   users.
+4. [Draft new release][new-release].
+5. Fill tag and target with the same value used to generate the changelog.
+6. Set title to the new tag value.
+7. Paste the generated changelog.
+8. Check "create discussion", in the "Releases" category.
+9. Check pre-release if new version is an alpha or beta.
 
 [issues-tracker]: https://github.com/pelletier/go-toml/issues
 [bug-report]: https://github.com/pelletier/go-toml/issues/new?template=bug_report.md
@@ -177,6 +191,6 @@ Checklist:
 [readme]: ./README.md
 [fork]: https://help.github.com/articles/fork-a-repo
 [pull-request]: https://help.github.com/en/articles/creating-a-pull-request
-[releases]: https://github.com/pelletier/go-toml/releases
 [new-release]: https://github.com/pelletier/go-toml/releases/new
-[release-110]: https://github.com/pelletier/go-toml/releases/tag/v1.1.0
+[gh]: https://github.com/cli/cli
+[pr-labels]: https://github.com/pelletier/go-toml/blob/v2/.github/release.yml

Dockerfile

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

LICENSE

@@ -1,6 +1,6 @@
 The MIT License (MIT)
 
-Copyright (c) 2013 - 2021 Thomas Pelletier, Eric Anderton
+Copyright (c) 2013 - 2022 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

README.md

@@ -4,32 +4,25 @@ Go library for the [TOML](https://toml.io/en/) format.
 
 This library supports [TOML v1.0.0](https://toml.io/en/v1.0.0).
 
+[🐞 Bug Reports](https://github.com/pelletier/go-toml/issues)
 
-## Development status
+[💬 Anything else](https://github.com/pelletier/go-toml/discussions)
-
-This is the upcoming major version of go-toml. It is currently in active
-development. As of release v2.0.0-beta.1, the library has reached feature parity
-with v1, and fixes a lot known bugs and performance issues along the way.
-
-If you do not need the advanced document editing features of v1, you are
-encouraged to try out this version.
-
-👉 [Roadmap for v2](https://github.com/pelletier/go-toml/discussions/506).
-
 
 ## Documentation
 
-Full API, examples, and implementation notes are available in the Go documentation.
+Full API, examples, and implementation notes are available in the Go
+documentation.
 
 [![Go Reference](https://pkg.go.dev/badge/github.com/pelletier/go-toml/v2.svg)](https://pkg.go.dev/github.com/pelletier/go-toml/v2)
 
-
 ## Import
 
 ```go
 import "github.com/pelletier/go-toml/v2"
 ```
 
+See [Modules](#Modules).
+
 ## Features
 
 ### Stdlib behavior
@@ -40,20 +33,21 @@ 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.
+operations should not be shockingly slow. See [benchmarks](#benchmarks).
 
 ### Strict mode
 
 `Decoder` can be set to "strict mode", which makes it error when some parts of
-the TOML document was not prevent in the target structure. This is a great way
+the TOML document was not present in the target structure. This is a great way
 to check for typos. [See example in the documentation][strict].
 
-[strict]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#example-Decoder.SetStrict
+[strict]: https://pkg.go.dev/github.com/pelletier/go-toml/v2#example-Decoder.DisallowUnknownFields
 
 ### Contextualized errors
 
-When decoding errors occur, go-toml returns [`DecodeError`][decode-err]), which
+When most decoding errors occur, go-toml returns [`DecodeError`][decode-err]),
-contains a human readable contextualized version of the error. For example:
+which contains a human readable contextualized version of the error. For
+example:
 
 ```
 2| key1 = "value1"
@@ -146,6 +140,101 @@ fmt.Println(string(b))
 
 [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>1.8x</td></tr>
+        <tr><td>Marshal/ReferenceFile/struct-2</td><td>2.2x</td><td>2.5x</td></tr>
+        <tr><td>Unmarshal/HugoFrontMatter-2</td><td>2.9x</td><td>2.9x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/map-2</td><td>2.6x</td><td>2.9x</td></tr>
+        <tr><td>Unmarshal/ReferenceFile/struct-2</td><td>4.4x</td><td>5.3x</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>1.8x</td><td>2.9x</td></tr>
+        <tr><td>Marshal/SimpleDocument/struct-2</td><td>2.7x</td><td>4.2x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/map-2</td><td>4.5x</td><td>3.1x</td></tr>
+        <tr><td>Unmarshal/SimpleDocument/struct-2</td><td>6.2x</td><td>3.9x</td></tr>
+        <tr><td>UnmarshalDataset/example-2</td><td>3.1x</td><td>3.5x</td></tr>
+        <tr><td>UnmarshalDataset/code-2</td><td>2.3x</td><td>3.1x</td></tr>
+        <tr><td>UnmarshalDataset/twitter-2</td><td>2.5x</td><td>2.6x</td></tr>
+        <tr><td>UnmarshalDataset/citm_catalog-2</td><td>2.1x</td><td>2.2x</td></tr>
+        <tr><td>UnmarshalDataset/canada-2</td><td>1.6x</td><td>1.3x</td></tr>
+        <tr><td>UnmarshalDataset/config-2</td><td>4.3x</td><td>3.2x</td></tr>
+        <tr><td>[Geo mean]</td><td>2.7x</td><td>2.8x</td></tr>
+     </tbody>
+</table>
+<p>This table can be generated with <code>./ci.sh benchmark -a -html</code>.</p>
+</details>
+
+## Modules
+
+go-toml uses Go's standard modules system.
+
+Installation instructions:
+
+- Go ≥ 1.16: Nothing to do. Use the import in your code. The `go` command deals
+  with it automatically.
+- Go ≥ 1.13: `GO111MODULE=on go get github.com/pelletier/go-toml/v2`.
+
+In case of trouble: [Go Modules FAQ][mod-faq].
+
+[mod-faq]: https://github.com/golang/go/wiki/Modules#why-does-installing-a-tool-via-go-get-fail-with-error-cannot-find-main-module
+
+## Tools
+
+Go-toml provides three handy command line tools:
+
+ * `tomljson`: Reads a TOML file and outputs its JSON representation.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/tomljson@latest
+    $ tomljson --help
+    ```
+
+ * `jsontoml`: Reads a JSON file and outputs a TOML representation.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/jsontoml@latest
+    $ jsontoml --help
+    ```
+
+ * `tomll`: Lints and reformats a TOML file.
+
+    ```
+    $ go install github.com/pelletier/go-toml/v2/cmd/tomll@latest
+    $ tomll --help
+    ```
+
+### Docker image
+
+Those tools are also available as a [Docker image][docker]. For example, to use
+`tomljson`:
+
+```
+docker run -i ghcr.io/pelletier/go-toml:v2 tomljson < example.toml
+```
+
+Multiple versions are availble on [ghcr.io][docker].
+
+[docker]: https://github.com/pelletier/go-toml/pkgs/container/go-toml
+
 ## Migrating from v1
 
 This section describes the differences between v1 and v2, with some pointers on
@@ -187,7 +276,7 @@ d := doc{
 }
 
 data := `
-[A]                                                                                                                           
+[A]
 B = "After"
 `
 
@@ -248,6 +337,45 @@ This method was not widely used, poorly defined, and added a lot of complexity.
 A similar effect can be achieved by implementing the `encoding.TextUnmarshaler`
 interface and use strings.
 
+#### Support for `default` struct tag has been dropped
+
+This feature adds complexity and a poorly defined API for an effect that can be
+accomplished outside of the library.
+
+It does not seem like other format parsers in Go support that feature (the
+project referenced in the original ticket #202 has not been updated since 2017).
+Given that go-toml v2 should not touch values not in the document, the same
+effect can be achieved by pre-filling the struct with defaults (libraries like
+[go-defaults][go-defaults] can help). Also, string representation is not well
+defined for all types: it creates issues like #278.
+
+The recommended replacement is pre-filling the struct before unmarshaling.
+
+[go-defaults]: https://github.com/mcuadros/go-defaults
+
+#### `toml.Tree` replacement
+
+This structure was the initial attempt at providing a document model for
+go-toml. It allows manipulating the structure of any document, encoding and
+decoding from their TOML representation. While a more robust feature was
+initially planned in go-toml v2, this has been ultimately [removed from
+scope][nodoc] of this library, with no plan to add it back at the moment. The
+closest equivalent at the moment would be to unmarshal into an `interface{}` and
+use type assertions and/or reflection to manipulate the arbitrary
+structure. However this would fall short of providing all of the TOML features
+such as adding comments and be specific about whitespace.
+
+
+#### `toml.Position` are not retrievable anymore
+
+The API for retrieving the position (line, column) of a specific TOML element do
+not exist anymore. This was done to minimize the amount of concepts introduced
+by the library (query path), and avoid the performance hit related to storing
+positions in the absence of a document model, for a feature that seemed to have
+little use. Errors however have gained more detailed position
+information. Position retrieval seems better fitted for a document model, which
+has been [removed from the scope][nodoc] of go-toml v2 at the moment.
+
 ### Encoding / Marshal
 
 #### Default struct fields order
@@ -283,14 +411,14 @@ fmt.Println("v2:\n" + string(b))
 ```
 
 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.
+manually sort the fields alphabetically in the struct definition, or generate
+struct types using `reflect.StructOf`.
 
 #### No indentation by default
 
 V1 automatically indents content of tables by default. V2 does not. However the
 same behavior can be obtained using [`Encoder.SetIndentTables`][sit]. For example:
 
-
 ```go
 data := map[string]interface{}{
 	"table": map[string]string{
@@ -312,15 +440,15 @@ fmt.Println("v2 Encoder:\n" + string(buf.Bytes()))
 
 // Output:
 // v1:
-// 
+//
 // [table]
 //   key = "value"
-// 
+//
 // v2:
 // [table]
 // key = 'value'
-// 
+//
-// 
+//
 // v2 Encoder:
 // [table]
 //   key = 'value'
@@ -332,7 +460,9 @@ fmt.Println("v2 Encoder:\n" + string(buf.Bytes()))
 
 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.
+unless a character cannot be represented, then falls back to double quotes. As a
+result of this change, `Encoder.QuoteMapKeys` has been removed, as it is not
+useful anymore.
 
 There is no way to make v2 encoder behave like v1.
 
@@ -347,6 +477,76 @@ There is no way to make v2 encoder behave like v1.
 
 [tm]: https://golang.org/pkg/encoding/#TextMarshaler
 
+#### `Encoder.CompactComments` has been removed
+
+Emitting compact comments is now the default behavior of go-toml. This option
+is not necessary anymore.
+
+#### Struct tags have been merged
+
+V1 used to provide multiple struct tags: `comment`, `commented`, `multiline`,
+`toml`, and `omitempty`. To behave more like the standard library, v2 has merged
+`toml`, `multiline`, and `omitempty`. For example:
+
+```go
+type doc struct {
+	// v1
+	F string `toml:"field" multiline:"true" omitempty:"true"`
+	// v2
+	F string `toml:"field,multiline,omitempty"`
+}
+```
+
+Has a result, the `Encoder.SetTag*` methods have been removed, as there is just
+one tag now.
+
+
+#### `commented` tag has been removed
+
+There is no replacement for the `commented` tag. This feature would be better
+suited in a proper document model for go-toml v2, which has been [cut from
+scope][nodoc] at the moment.
+
+#### `Encoder.ArraysWithOneElementPerLine` has been renamed
+
+The new name is `Encoder.SetArraysMultiline`. The behavior should be the same.
+
+#### `Encoder.Indentation` has been renamed
+
+The new name is `Encoder.SetIndentSymbol`. The behavior should be the same.
+
+
+#### Embedded structs behave like stdlib
+
+V1 defaults to merging embedded struct fields into the embedding struct. This
+behavior was unexpected because it does not follow the standard library. To
+avoid breaking backward compatibility, the `Encoder.PromoteAnonymous` method was
+added to make the encoder behave correctly. Given backward compatibility is not
+a problem anymore, v2 does the right thing by default: it follows the behavior
+of `encoding/json`. `Encoder.PromoteAnonymous` has been removed.
+
+[nodoc]: https://github.com/pelletier/go-toml/discussions/506#discussioncomment-1526038
+
+### `query`
+
+go-toml v1 provided the [`go-toml/query`][query] package. It allowed to run
+JSONPath-style queries on TOML files. This feature is not available in v2. For a
+replacement, check out [dasel][dasel].
+
+This package has been removed because it was essentially not supported anymore
+(last commit May 2020), increased the complexity of the code base, and more
+complete solutions exist out there.
+
+[query]: https://github.com/pelletier/go-toml/tree/f99d6bbca119636aeafcf351ee52b3d202782627/query
+[dasel]: https://github.com/TomWright/dasel
+
+## Versioning
+
+Go-toml follows [Semantic Versioning](http://semver.org/). The supported version
+of [TOML](https://github.com/toml-lang/toml) is indicated at the beginning of
+this document. The last two major versions of Go are supported
+(see [Go Release Policy](https://golang.org/doc/devel/release.html#policy)).
+
 ## License
 
 The MIT License (MIT). Read [LICENSE](LICENSE).

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

@@ -31,13 +31,14 @@ var bench_inputs = []struct {
 
 func TestUnmarshalDatasetCode(t *testing.T) {
 	for _, tc := range bench_inputs {
-		buf := fixture(t, tc.name)
 		t.Run(tc.name, func(t *testing.T) {
+			buf := fixture(t, tc.name)
+
 			var v interface{}
-			check(t, toml.Unmarshal(buf, &v))
+			require.NoError(t, toml.Unmarshal(buf, &v))
 
 			b, err := json.Marshal(v)
-			check(t, err)
+			require.NoError(t, err)
 			require.Equal(t, len(b), tc.jsonLen)
 		})
 	}
@@ -45,14 +46,14 @@ func TestUnmarshalDatasetCode(t *testing.T) {
 
 func BenchmarkUnmarshalDataset(b *testing.B) {
 	for _, tc := range bench_inputs {
-		buf := fixture(b, tc.name)
 		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{}
-				check(b, toml.Unmarshal(buf, &v))
+				require.NoError(b, toml.Unmarshal(buf, &v))
 			}
 		})
 	}
@@ -60,22 +61,20 @@ func BenchmarkUnmarshalDataset(b *testing.B) {
 
 // fixture returns the uncompressed contents of path.
 func fixture(tb testing.TB, path string) []byte {
-	f, err := os.Open(filepath.Join("testdata", path+".toml.gz"))
+	tb.Helper()
-	check(tb, err)
+
+	file := path + ".toml.gz"
+	f, err := os.Open(filepath.Join("testdata", file))
+	if os.IsNotExist(err) {
+		tb.Skip("benchmark fixture not found:", file)
+	}
+	require.NoError(tb, err)
 	defer f.Close()
 
 	gz, err := gzip.NewReader(f)
-	check(tb, err)
+	require.NoError(tb, err)
 
 	buf, err := ioutil.ReadAll(gz)
-	check(tb, err)
+	require.NoError(tb, err)
-
 	return buf
 }
-
-func check(tb testing.TB, err error) {
-	if err != nil {
-		tb.Helper()
-		tb.Fatal(err)
-	}
-}

benchmark/benchmark.toml

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

benchmark/benchmark_test.go

@@ -1,6 +1,7 @@
 package benchmark_test
 
 import (
+	"bytes"
 	"io/ioutil"
 	"testing"
 	"time"
@@ -9,17 +10,230 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-func BenchmarkUnmarshalSimple(b *testing.B) {
+func TestUnmarshalSimple(t *testing.T) {
+	doc := []byte(`A = "hello"`)
 	d := struct {
 		A string
 	}{}
-	doc := []byte(`A = "hello"`)
+
-	for i := 0; i < b.N; i++ {
+	err := toml.Unmarshal(doc, &d)
-		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 {
@@ -35,7 +249,7 @@ type benchmarkDoc struct {
 			}
 			Point struct {
 				X int64
-				U int64
+				Y int64
 			}
 		}
 	}
@@ -108,6 +322,7 @@ type benchmarkDoc struct {
 		Key2 []string
 		Key3 [][]int64
 		// TODO: Key4 not supported by go-toml's Unmarshal
+		Key4 []interface{}
 		Key5 []int64
 		Key6 []int64
 	}
@@ -119,36 +334,321 @@ type benchmarkDoc struct {
 	Fruit []struct {
 		Name     string
 		Physical struct {
-			Color   string
+			Color string
-			Shape   string
+			Shape string
-			Variety []struct {
+		}
-				Name string
+		Variety []struct {
-			}
+			Name string
 		}
 	}
 }
 
-func BenchmarkReferenceFile(b *testing.B) {
+func TestUnmarshalReferenceFile(t *testing.T) {
-	bytes, err := ioutil.ReadFile("benchmark.toml")
-	if err != nil {
-		b.Fatal(err)
-	}
-	b.SetBytes(int64(len(bytes)))
-	b.ReportAllocs()
-	b.ResetTimer()
-	for i := 0; i < b.N; i++ {
-		d := benchmarkDoc{}
-		err := toml.Unmarshal(bytes, &d)
-		if err != nil {
-			panic(err)
-		}
-	}
-}
-
-func TestReferenceFile(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

@@ -39,6 +39,12 @@ benchmark [OPTIONS...] [BRANCH]
         -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.
@@ -70,7 +76,8 @@ cover() {
     fi
 
     pushd "$dir"
-    go test -covermode=atomic -coverprofile=coverage.out ./...
+    go test -covermode=atomic  -coverpkg=./... -coverprofile=coverage.out.tmp ./...
+    cat coverage.out.tmp | grep -v testsuite | grep -v tomltestgen | grep -v gotoml-test-decoder > coverage.out
     go tool cover -func=coverage.out
     popd
 
@@ -97,16 +104,23 @@ coverage() {
 
 	    echo ""
 
-	    target_pct="$(cat ${target_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    target_pct="$(tail -n2 ${target_out} | head -n1 | sed -E 's/.*total.*\t([0-9.]+)%.*/\1/')"
-	    head_pct="$(cat ${head_out} |sed -E 's/.*total.*\t([0-9.]+)%/\1/;t;d')"
+	    head_pct="$(tail -n2 ${head_out} | head -n1 | sed -E 's/.*total.*\t([0-9.]+)%/\1/')"
 	    echo "Results: ${target} ${target_pct}% HEAD ${head_pct}%"
 
 	    delta_pct=$(echo "$head_pct - $target_pct" | bc -l)
 	    echo "Delta: ${delta_pct}"
 
 	    if [[ $delta_pct = \-* ]]; then
-		echo "Regression!";
+		    echo "Regression!";
-		return 1
+
+            target_diff="${output_dir}/target.diff.txt"
+            head_diff="${output_dir}/head.diff.txt"
+            cat "${target_out}" | grep -E '^github.com/pelletier/go-toml' | tr -s "\t " | cut -f 2,3 | sort > "${target_diff}"
+            cat "${head_out}" | grep -E '^github.com/pelletier/go-toml' | tr -s "\t " | cut -f 2,3 | sort > "${head_diff}"
+
+            diff --side-by-side --suppress-common-lines "${target_diff}" "${head_diff}"
+		    return 1
 	    fi
 	    return 0
 	    ;;
@@ -118,6 +132,7 @@ coverage() {
 bench() {
     branch="${1}"
     out="${2}"
+    replace="${3}"
     dir="$(mktemp -d)"
 
     stderr "Executing benchmark for ${branch} at ${dir}"
@@ -129,7 +144,14 @@ bench() {
     fi
 
     pushd "$dir"
-    go test -bench=. -count=10 ./... | tee "${out}"
+
+    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
@@ -137,19 +159,114 @@ bench() {
     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=`mktemp`
+
+        old=`fmktemp ${target}`
         bench "${target}" "${old}"
 
-        new=`mktemp`
+        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`

cmd/gotoml-test-decoder/main.go

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

cmd/jsontoml/main.go

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

cmd/jsontoml/main_test.go

@@ -0,0 +1,48 @@
+package main
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConvert(t *testing.T) {
+	examples := []struct {
+		name     string
+		input    string
+		expected string
+		errors   bool
+	}{
+		{
+			name: "valid json",
+			input: `
+{
+  "mytoml": {
+    "a": 42
+  }
+}`,
+			expected: `[mytoml]
+a = 42.0
+`,
+		},
+		{
+			name:   "invalid json",
+			input:  `{ foo`,
+			errors: true,
+		},
+	}
+
+	for _, e := range examples {
+		b := new(bytes.Buffer)
+		err := convert(strings.NewReader(e.input), b)
+		if e.errors {
+			require.Error(t, err)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, e.expected, b.String())
+		}
+	}
+}

cmd/tomljson/main.go

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

cmd/tomljson/main_test.go

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

cmd/tomll/main.go

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

cmd/tomll/main_test.go

@@ -0,0 +1,45 @@
+package main
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConvert(t *testing.T) {
+	examples := []struct {
+		name     string
+		input    string
+		expected string
+		errors   bool
+	}{
+		{
+			name: "valid toml",
+			input: `
+mytoml.a = 42.0
+`,
+			expected: `[mytoml]
+a = 42.0
+`,
+		},
+		{
+			name:   "invalid toml",
+			input:  `[what`,
+			errors: true,
+		},
+	}
+
+	for _, e := range examples {
+		b := new(bytes.Buffer)
+		err := convert(strings.NewReader(e.input), b)
+		if e.errors {
+			require.Error(t, err)
+		} else {
+			assert.NoError(t, err)
+			assert.Equal(t, e.expected, b.String())
+		}
+	}
+}

cmd/tomltestgen/main.go

@@ -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

@@ -35,26 +35,42 @@ func parseLocalDate(b []byte) (LocalDate, error) {
 		return date, newDecodeError(b, "dates are expected to have the format YYYY-MM-DD")
 	}
 
-	date.Year = parseDecimalDigits(b[0:4])
+	var err error
 
-	v := parseDecimalDigits(b[5:7])
+	date.Year, err = parseDecimalDigits(b[0:4])
+	if err != nil {
+		return LocalDate{}, err
+	}
 
-	date.Month = time.Month(v)
+	date.Month, err = parseDecimalDigits(b[5:7])
+	if err != nil {
+		return LocalDate{}, err
+	}
 
-	date.Day = parseDecimalDigits(b[8:10])
+	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 {
+func parseDecimalDigits(b []byte) (int, error) {
 	v := 0
 
-	for _, c := range b {
+	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
+	return v, nil
 }
 
 func parseDateTime(b []byte) (time.Time, error) {
@@ -75,7 +91,7 @@ func parseDateTime(b []byte) (time.Time, error) {
 		panic("date time should have a timezone")
 	}
 
-	if b[0] == 'Z' {
+	if b[0] == 'Z' || b[0] == 'z' {
 		b = b[1:]
 		zone = time.UTC
 	} else {
@@ -83,15 +99,42 @@ func parseDateTime(b []byte) (time.Time, error) {
 		if len(b) != dateTimeByteLen {
 			return time.Time{}, newDecodeError(b, "invalid date-time timezone")
 		}
-		direction := 1
+		var direction int
-		if b[0] == '-' {
+		switch b[0] {
+		case '-':
 			direction = -1
+		case '+':
+			direction = +1
+		default:
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset character")
+		}
+
+		if b[3] != ':' {
+			return time.Time{}, newDecodeError(b[3:4], "expected a : separator")
+		}
+
+		hours, err := parseDecimalDigits(b[1:3])
+		if err != nil {
+			return time.Time{}, err
+		}
+		if hours > 23 {
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset hours")
+		}
+
+		minutes, err := parseDecimalDigits(b[4:6])
+		if err != nil {
+			return time.Time{}, err
+		}
+		if minutes > 59 {
+			return time.Time{}, newDecodeError(b[:1], "invalid timezone offset minutes")
 		}
 
-		hours := digitsToInt(b[1:3])
-		minutes := digitsToInt(b[4:6])
 		seconds := direction * (hours*3600 + minutes*60)
-		zone = time.FixedZone("", seconds)
+		if seconds == 0 {
+			zone = time.UTC
+		} else {
+			zone = time.FixedZone("", seconds)
+		}
 		b = b[dateTimeByteLen:]
 	}
 
@@ -100,13 +143,13 @@ func parseDateTime(b []byte) (time.Time, error) {
 	}
 
 	t := time.Date(
-		dt.Date.Year,
+		dt.Year,
-		dt.Date.Month,
+		time.Month(dt.Month),
-		dt.Date.Day,
+		dt.Day,
-		dt.Time.Hour,
+		dt.Hour,
-		dt.Time.Minute,
+		dt.Minute,
-		dt.Time.Second,
+		dt.Second,
-		dt.Time.Nanosecond,
+		dt.Nanosecond,
 		zone)
 
 	return t, nil
@@ -124,10 +167,10 @@ func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
 	if err != nil {
 		return dt, nil, err
 	}
-	dt.Date = date
+	dt.LocalDate = date
 
 	sep := b[10]
-	if sep != 'T' && sep != ' ' {
+	if sep != 'T' && sep != ' ' && sep != 't' {
 		return dt, nil, newDecodeError(b[10:11], "datetime separator is expected to be T or a space")
 	}
 
@@ -135,7 +178,7 @@ func parseLocalDateTime(b []byte) (LocalDateTime, []byte, error) {
 	if err != nil {
 		return dt, nil, err
 	}
-	dt.Time = t
+	dt.LocalTime = t
 
 	return dt, rest, nil
 }
@@ -149,45 +192,89 @@ func parseLocalTime(b []byte) (LocalTime, []byte, error) {
 		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]")
 	}
 
-	t.Hour = parseDecimalDigits(b[0:2])
+	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 = parseDecimalDigits(b[3:5])
+	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 = parseDecimalDigits(b[6:8])
+	t.Second, err = parseDecimalDigits(b[6:8])
+	if err != nil {
+		return t, nil, err
+	}
 
-	const minLengthWithFrac = 9
+	if t.Second > 60 {
-	if len(b) >= minLengthWithFrac && b[minLengthWithFrac-1] == '.' {
+		return t, nil, newDecodeError(b[6:8], "seconds cannot be greater 60")
+	}
+
+	b = b[8:]
+
+	if len(b) >= 1 && b[0] == '.' {
 		frac := 0
+		precision := 0
 		digits := 0
 
-		for i, c := range b[minLengthWithFrac:] {
+		for i, c := range b[1:] {
+			if !isDigit(c) {
+				if i == 0 {
+					return t, nil, newDecodeError(b[0:1], "need at least one digit after fraction point")
+				}
+				break
+			}
+			digits++
+
 			const maxFracPrecision = 9
 			if i >= maxFracPrecision {
-				return t, nil, newDecodeError(b[i:i+1], "maximum precision for date time is nanosecond")
+				// go-toml allows decoding fractional seconds
+				// beyond the supported precision of 9
+				// digits. It truncates the fractional component
+				// to the supported precision and ignores the
+				// remaining digits.
+				//
+				// https://github.com/pelletier/go-toml/discussions/707
+				continue
 			}
 
 			frac *= 10
 			frac += int(c - '0')
-			digits++
+			precision++
 		}
 
-		t.Nanosecond = frac * nspow[digits]
+		if precision == 0 {
+			return t, nil, newDecodeError(b[:1], "nanoseconds need at least one digit")
+		}
 
-		return t, b[9+digits:], nil
+		t.Nanosecond = frac * nspow[precision]
+		t.Precision = precision
+
+		return t, b[1+digits:], nil
 	}
-
+	return t, b, nil
-	return t, b[8:], nil
 }
 
 //nolint:cyclop
@@ -196,7 +283,7 @@ func parseFloat(b []byte) (float64, error) {
 		return math.NaN(), nil
 	}
 
-	cleaned, err := checkAndRemoveUnderscores(b)
+	cleaned, err := checkAndRemoveUnderscoresFloats(b)
 	if err != nil {
 		return 0, err
 	}
@@ -209,6 +296,30 @@ func parseFloat(b []byte) (float64, error) {
 		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 cleaned[0] == '+' || cleaned[0] == '-' {
+		start = 1
+	}
+	if cleaned[start] == '0' && isDigit(cleaned[start+1]) {
+		return 0, newDecodeError(b, "float integer part cannot have leading zeroes")
+	}
+
 	f, err := strconv.ParseFloat(string(cleaned), 64)
 	if err != nil {
 		return 0, newDecodeError(b, "unable to parse float: %w", err)
@@ -218,7 +329,7 @@ func parseFloat(b []byte) (float64, error) {
 }
 
 func parseIntHex(b []byte) (int64, error) {
-	cleaned, err := checkAndRemoveUnderscores(b[2:])
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
@@ -232,7 +343,7 @@ func parseIntHex(b []byte) (int64, error) {
 }
 
 func parseIntOct(b []byte) (int64, error) {
-	cleaned, err := checkAndRemoveUnderscores(b[2:])
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
@@ -246,7 +357,7 @@ func parseIntOct(b []byte) (int64, error) {
 }
 
 func parseIntBin(b []byte) (int64, error) {
-	cleaned, err := checkAndRemoveUnderscores(b[2:])
+	cleaned, err := checkAndRemoveUnderscoresIntegers(b[2:])
 	if err != nil {
 		return 0, err
 	}
@@ -259,12 +370,26 @@ func parseIntBin(b []byte) (int64, error) {
 	return i, nil
 }
 
+func isSign(b byte) bool {
+	return b == '+' || b == '-'
+}
+
 func parseIntDec(b []byte) (int64, error) {
-	cleaned, err := checkAndRemoveUnderscores(b)
+	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)
@@ -273,9 +398,18 @@ func parseIntDec(b []byte) (int64, error) {
 	return i, nil
 }
 
-func checkAndRemoveUnderscores(b []byte) ([]byte, error) {
+func checkAndRemoveUnderscoresIntegers(b []byte) ([]byte, error) {
-	if b[0] == '_' {
+	start := 0
-		return nil, newDecodeError(b[0:1], "number cannot start with underscore")
+	if b[start] == '+' || b[start] == '-' {
+		start++
+	}
+
+	if len(b) == start {
+		return b, nil
+	}
+
+	if b[start] == '_' {
+		return nil, newDecodeError(b[start:start+1], "number cannot start with underscore")
 	}
 
 	if b[len(b)-1] == '_' {
@@ -312,3 +446,99 @@ func checkAndRemoveUnderscores(b []byte) ([]byte, error) {
 
 	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 '+', '-':
+			// signed exponents
+			cleaned = append(cleaned, c)
+			before = false
+		case 'e', 'E':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after exponent")
+			}
+			cleaned = append(cleaned, c)
+		case '.':
+			if i < len(b)-1 && b[i+1] == '_' {
+				return nil, newDecodeError(b[i+1:i+2], "cannot have underscore after decimal point")
+			}
+			if i > 0 && b[i-1] == '_' {
+				return nil, newDecodeError(b[i-1:i], "cannot have underscore before decimal point")
+			}
+			cleaned = append(cleaned, c)
+		default:
+			before = true
+			cleaned = append(cleaned, c)
+		}
+	}
+
+	return cleaned, nil
+}
+
+// isValidDate checks if a provided date is a date that exists.
+func isValidDate(year int, month int, day int) bool {
+	return month > 0 && month < 13 && day > 0 && day <= daysIn(month, year)
+}
+
+// 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)
+}

errors.go

@@ -5,7 +5,7 @@ import (
 	"strconv"
 	"strings"
 
-	"github.com/pelletier/go-toml/v2/internal/unsafe"
+	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
 // DecodeError represents an error encountered during the parsing or decoding
@@ -27,7 +27,7 @@ type DecodeError struct {
 // corresponding field in the target value. It contains all the missing fields
 // in Errors.
 //
-// Emitted by Decoder when SetStrict(true) was called.
+// Emitted by Decoder when DisallowUnknownFields() was called.
 type StrictMissingError struct {
 	// One error per field that could not be found.
 	Errors []DecodeError
@@ -103,9 +103,10 @@ func (e *DecodeError) Key() Key {
 //
 // The function copies all bytes used in DecodeError, so that document and
 // highlight can be freely deallocated.
+//
 //nolint:funlen
 func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
-	offset := unsafe.SubsliceOffset(document, de.highlight)
+	offset := danger.SubsliceOffset(document, de.highlight)
 
 	errMessage := de.Error()
 	errLine, errColumn := positionAtEnd(document[:offset])
@@ -116,6 +117,7 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 	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))
@@ -129,6 +131,8 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 		buf.WriteRune('\n')
 	}
 
+	// Write the document line that contains the error.
+
 	buf.WriteString(formatLineNumber(errLine, lineColumnWidth))
 	buf.WriteString("| ")
 
@@ -143,6 +147,10 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 	}
 
 	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("| ")
 
@@ -157,6 +165,8 @@ func wrapDecodeError(document []byte, de *decodeError) *DecodeError {
 		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
@@ -230,7 +240,7 @@ forward:
 			rest = rest[o+1:]
 			o = 0
 
-		case o == len(rest)-1 && o > 0:
+		case o == len(rest)-1:
 			// add last line only if it's non-empty
 			afterLines = append(afterLines, rest)
 

errors_test.go

@@ -12,7 +12,6 @@ import (
 
 //nolint:funlen
 func TestDecodeError(t *testing.T) {
-	t.Parallel()
 
 	examples := []struct {
 		desc     string
@@ -149,12 +148,19 @@ line 5`,
 6|
 7| line 4`,
 		},
+		{
+			desc: "handle remainder of the error line when there is only one line",
+			doc:  [3]string{`P=`, `[`, `#`},
+			msg:  "array is incomplete",
+			expected: `1| P=[#
+ |   ~ array is incomplete`,
+		},
 	}
 
 	for _, e := range examples {
 		e := e
 		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
+
 			b := bytes.Buffer{}
 			b.Write([]byte(e.doc[0]))
 			start := b.Len()
@@ -182,7 +188,6 @@ line 5`,
 }
 
 func TestDecodeError_Accessors(t *testing.T) {
-	t.Parallel()
 
 	e := DecodeError{
 		message: "foo",
@@ -207,12 +212,12 @@ func ExampleDecodeError() {
 
 	fmt.Println(err)
 
-	//nolint:errorlint
+	var derr *DecodeError
-	de := err.(*DecodeError)
+	if errors.As(err, &derr) {
-	fmt.Println(de.String())
+		fmt.Println(derr.String())
-
+		row, col := derr.Position()
-	row, col := de.Position()
+		fmt.Println("error occurred at row", row, "column", col)
-	fmt.Println("error occurred at row", row, "column", col)
+	}
 	// Output:
 	// toml: number must have at least one digit between underscores
 	// 1| name = 123__456

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)
+}

fuzz_test.go

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

go.mod

@@ -1,5 +1,5 @@
 module github.com/pelletier/go-toml/v2
 
-go 1.15
+go 1.16
 
-require github.com/stretchr/testify v1.7.0
+require github.com/stretchr/testify v1.8.0

go.sum

@@ -1,11 +1,15 @@
-github.com/davecgh/go-spew v1.1.0 h1:ZDRjVQ15GmhC3fiQ8ni8+OwkZQO4DARzQgrnXU1Liz8=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
+github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
+github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
 github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/testify v1.7.0 h1:nwc3DEeHmmLAfoZucVR881uASk0Mfjw8xYJ99tb5CcY=
+github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
-github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
+github.com/stretchr/testify v1.8.0 h1:pSgiaMZlXftHpm5L7V1+rVB+AZJydKsMxsQBIJw4PKk=
+github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO+kdMU+MU=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405 h1:yhCVgyC4o1eVCa2tZl7eS0r+SDo693bJlVdllGtEeKM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
-gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c h1:dUUwHk2QECo/6vqA44rthZ8ie2QXMNeKRTHCNY2nXvo=
 gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
+gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
+gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

internal/ast/ast.go

@@ -2,23 +2,26 @@ 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()
+//	it := n.Children()
-// for it.Next() {
+//	for it.Next() {
-// 		it.Node()
+//			it.Node()
-// }
+//	}
 type Iterator struct {
 	started bool
-	node    Node
+	node    *Node
 }
 
-// Next moves the iterator forward and returns true if points to a node, false
+// Next moves the iterator forward and returns true if points to a
-// otherwise.
+// node, false otherwise.
 func (c *Iterator) Next() bool {
 	if !c.started {
 		c.started = true
@@ -28,8 +31,14 @@ func (c *Iterator) Next() bool {
 	return c.node.Valid()
 }
 
+// IsLast returns true if the current node of the iterator is the last
+// one.  Subsequent call to Next() will return false.
+func (c *Iterator) IsLast() bool {
+	return c.node.next == 0
+}
+
 // Node returns a copy of the node pointed at by the iterator.
-func (c *Iterator) Node() Node {
+func (c *Iterator) Node() *Node {
 	return c.node
 }
 
@@ -44,65 +53,69 @@ type Root struct {
 func (r *Root) Iterator() Iterator {
 	it := Iterator{}
 	if len(r.nodes) > 0 {
-		it.node = r.nodes[0]
+		it.node = &r.nodes[0]
 	}
 	return it
 }
 
-func (r *Root) at(idx int) Node {
+func (r *Root) at(idx Reference) *Node {
-	// TODO: unsafe to point to the node directly
+	return &r.nodes[idx]
-	return r.nodes[idx]
 }
 
-// Arrays have one child per element in the array.
+// Arrays have one child per element in the array.  InlineTables have
-// InlineTables have one child per key-value pair in the table.
+// one child per key-value pair in the table.  KeyValues have at least
-// KeyValues have at least two children. The first one is the value. The
+// two children. The first one is the value. The rest make a
-// rest make a potentially dotted key.
+// potentially dotted key.  Table and Array table have one child per
-// Table and Array table have one child per element of the key they
+// element of the key they represent (same as KeyValue, but without
-// represent (same as KeyValue, but without the last node being the value).
+// the last node being the value).
-// children []Node
 type Node struct {
 	Kind Kind
-	Data []byte // Raw bytes from the input
+	Raw  Range  // Raw bytes from the input.
+	Data []byte // Node value (either allocated or referencing the input).
 
-	// next idx (in the root array). 0 if last of the collection.
+	// References to other nodes, as offsets in the backing array
-	next int
+	// from this node. References can go backward, so those can be
-	// child idx (in the root array). 0 if no child.
+	// negative.
-	child int
+	next  int // 0 if last element
-	// pointer to the root array
+	child int // 0 if no child
-	root *Root
 }
 
-// Next returns a copy of the next node, or an invalid Node if there is no
+type Range struct {
-// next node.
+	Offset uint32
-func (n Node) Next() Node {
+	Length uint32
-	if n.next <= 0 {
-		return noNode
-	}
-	return n.root.at(n.next)
 }
 
-// Child returns a copy of the first child node of this node. Other children
+// Next returns a copy of the next node, or an invalid Node if there
-// can be accessed calling Next on the first child.
+// is no next node.
-// Returns an invalid Node if there is none.
+func (n *Node) Next() *Node {
-func (n Node) Child() Node {
+	if n.next == 0 {
-	if n.child <= 0 {
+		return nil
-		return noNode
 	}
-	return n.root.at(n.child)
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.next))
+}
+
+// Child returns a copy of the first child node of this node. Other
+// children can be accessed calling Next on the first child.  Returns
+// an invalid Node if there is none.
+func (n *Node) Child() *Node {
+	if n.child == 0 {
+		return nil
+	}
+	ptr := unsafe.Pointer(n)
+	size := unsafe.Sizeof(Node{})
+	return (*Node)(danger.Stride(ptr, size, n.child))
 }
 
 // Valid returns true if the node's kind is set (not to Invalid).
-func (n Node) Valid() bool {
+func (n *Node) Valid() bool {
-	return n.Kind != Invalid
+	return n != nil
 }
 
-var noNode = Node{}
+// Key returns the child nodes making the Key on a supported
-
+// node. Panics otherwise.  They are guaranteed to be all be of the
-// Key returns the child nodes making the Key on a supported node. Panics
+// Kind Key. A simple key would return just one element.
-// otherwise.
-// They are guaranteed to be all be of the Kind Key. A simple key would return
-// just one element.
 func (n *Node) Key() Iterator {
 	switch n.Kind {
 	case KeyValue:
@@ -119,20 +132,13 @@ func (n *Node) Key() Iterator {
 }
 
 // Value returns a pointer to the value node of a KeyValue.
-// Guaranteed to be non-nil.
+// Guaranteed to be non-nil.  Panics if not called on a KeyValue node,
-// Panics if not called on a KeyValue node, or if the Children are malformed.
+// or if the Children are malformed.
-func (n Node) Value() Node {
+func (n *Node) Value() *Node {
-	assertKind(KeyValue, n)
 	return n.Child()
 }
 
 // Children returns an iterator over a node's children.
-func (n Node) Children() Iterator {
+func (n *Node) Children() Iterator {
 	return Iterator{node: n.Child()}
 }
-
-func assertKind(k Kind, n Node) {
-	if n.Kind != k {
-		panic(fmt.Errorf("method was expecting a %s, not a %s", k, n.Kind))
-	}
-}

internal/ast/builder.go

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

internal/ast/kind.go

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

internal/cli/cli.go

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

internal/cli/cli_test.go

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

internal/danger/danger.go

@@ -1,4 +1,4 @@
-package unsafe
+package danger
 
 import (
 	"fmt"
@@ -57,3 +57,9 @@ func BytesRange(start []byte, end []byte) []byte {
 
 	return start[:l]
 }
+
+func Stride(ptr unsafe.Pointer, size uintptr, offset int) unsafe.Pointer {
+	// TODO: replace with unsafe.Add when Go 1.17 is released
+	//   https://github.com/golang/go/issues/40481
+	return unsafe.Pointer(uintptr(ptr) + uintptr(int(size)*offset))
+}

internal/danger/danger_test.go

@@ -1,15 +1,16 @@
-package unsafe_test
+package danger_test
 
 import (
 	"testing"
+	"unsafe"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
-	"github.com/pelletier/go-toml/v2/internal/unsafe"
+	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
-func TestUnsafeSubsliceOffsetValid(t *testing.T) {
+func TestSubsliceOffsetValid(t *testing.T) {
 	examples := []struct {
 		desc   string
 		test   func() ([]byte, []byte)
@@ -28,13 +29,13 @@ func TestUnsafeSubsliceOffsetValid(t *testing.T) {
 	for _, e := range examples {
 		t.Run(e.desc, func(t *testing.T) {
 			d, s := e.test()
-			offset := unsafe.SubsliceOffset(d, s)
+			offset := danger.SubsliceOffset(d, s)
 			assert.Equal(t, e.offset, offset)
 		})
 	}
 }
 
-func TestUnsafeSubsliceOffsetInvalid(t *testing.T) {
+func TestSubsliceOffsetInvalid(t *testing.T) {
 	examples := []struct {
 		desc string
 		test func() ([]byte, []byte)
@@ -72,13 +73,22 @@ func TestUnsafeSubsliceOffsetInvalid(t *testing.T) {
 		t.Run(e.desc, func(t *testing.T) {
 			d, s := e.test()
 			require.Panics(t, func() {
-				unsafe.SubsliceOffset(d, s)
+				danger.SubsliceOffset(d, s)
 			})
 		})
 	}
 }
 
-func TestUnsafeBytesRange(t *testing.T) {
+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
@@ -157,10 +167,10 @@ func TestUnsafeBytesRange(t *testing.T) {
 			start, end := e.test()
 			if e.expected == nil {
 				require.Panics(t, func() {
-					unsafe.BytesRange(start, end)
+					danger.BytesRange(start, end)
 				})
 			} else {
-				res := unsafe.BytesRange(start, end)
+				res := danger.BytesRange(start, end)
 				require.Equal(t, e.expected, res)
 			}
 		})

internal/danger/typeid.go

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

internal/imported_tests/marshal_imported_test.go

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

internal/imported_tests/unmarshal_imported_test.go

@@ -151,6 +151,7 @@ type quotedKeyMarshalTestStruct struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var quotedKeyMarshalTestData = quotedKeyMarshalTestStruct{
 	String:  "Hello",
@@ -160,6 +161,7 @@ var quotedKeyMarshalTestData = quotedKeyMarshalTestStruct{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var quotedKeyMarshalTestToml = []byte(`"Yfloat-𝟘" = 3.5
 "Z.string-àéù" = "Hello"
@@ -223,11 +225,13 @@ type testSubDoc struct {
 	unexported int    `toml:"shouldntBeHere"`
 }
 
-var biteMe = "Bite me"
+var (
-var float1 float32 = 12.3
+	biteMe         = "Bite me"
-var float2 float32 = 45.6
+	float1 float32 = 12.3
-var float3 float32 = 78.9
+	float2 float32 = 45.6
-var subdoc = testSubDoc{"Second", 0}
+	float3 float32 = 78.9
+	subdoc         = testSubDoc{"Second", 0}
+)
 
 var docData = testDoc{
 	Title:       "TOML Marshal Testing",
@@ -270,6 +274,7 @@ var docData = testDoc{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapTestDoc = testMapDoc{
 	Title: "TOML Marshal Testing",
@@ -382,7 +387,7 @@ var intErrTomls = []string{
 }
 
 func TestErrUnmarshal(t *testing.T) {
-	var errTomls = []string{
+	errTomls := []string{
 		"bool = truly\ndate = 1979-05-27T07:32:00Z\nfloat = 123.4\nint = 5000\nstring = \"Bite me\"",
 		"bool = true\ndate = 1979-05-27T07:3200Z\nfloat = 123.4\nint = 5000\nstring = \"Bite me\"",
 		"bool = true\ndate = 1979-05-27T07:32:00Z\nfloat = 123a4\nint = 5000\nstring = \"Bite me\"",
@@ -455,35 +460,6 @@ func TestEmptytomlUnmarshal(t *testing.T) {
 	assert.Equal(t, emptyTestData, result)
 }
 
-func TestEmptyUnmarshalOmit(t *testing.T) {
-	t.Skipf("Have not figured yet if omitempty is a good idea")
-
-	type emptyMarshalTestStruct2 struct {
-		Title      string                  `toml:"title"`
-		Bool       bool                    `toml:"bool,omitempty"`
-		Int        int                     `toml:"int, omitempty"`
-		String     string                  `toml:"string,omitempty "`
-		StringList []string                `toml:"stringlist,omitempty"`
-		Ptr        *basicMarshalTestStruct `toml:"ptr,omitempty"`
-		Map        map[string]string       `toml:"map,omitempty"`
-	}
-
-	var emptyTestData2 = emptyMarshalTestStruct2{
-		Title:      "Placeholder",
-		Bool:       false,
-		Int:        0,
-		String:     "",
-		StringList: []string{},
-		Ptr:        nil,
-		Map:        map[string]string{},
-	}
-
-	result := emptyMarshalTestStruct2{}
-	err := toml.Unmarshal(emptyTestToml, &result)
-	require.NoError(t, err)
-	assert.Equal(t, emptyTestData2, result)
-}
-
 type pointerMarshalTestStruct struct {
 	Str       *string
 	List      *[]string
@@ -496,21 +472,23 @@ type pointerMarshalTestStruct struct {
 	DblPtr    *[]*[]*string
 }
 
-var pointerStr = "Hello"
+var (
-var pointerList = []string{"Hello back"}
+	pointerStr      = "Hello"
-var pointerListPtr = []*string{&pointerStr}
+	pointerList     = []string{"Hello back"}
-var pointerMap = map[string]string{"response": "Goodbye"}
+	pointerListPtr  = []*string{&pointerStr}
-var pointerMapPtr = map[string]*string{"alternate": &pointerStr}
+	pointerMap      = map[string]string{"response": "Goodbye"}
-var pointerTestData = pointerMarshalTestStruct{
+	pointerMapPtr   = map[string]*string{"alternate": &pointerStr}
-	Str:       &pointerStr,
+	pointerTestData = pointerMarshalTestStruct{
-	List:      &pointerList,
+		Str:       &pointerStr,
-	ListPtr:   &pointerListPtr,
+		List:      &pointerList,
-	Map:       &pointerMap,
+		ListPtr:   &pointerListPtr,
-	MapPtr:    &pointerMapPtr,
+		Map:       &pointerMap,
-	EmptyStr:  nil,
+		MapPtr:    &pointerMapPtr,
-	EmptyList: nil,
+		EmptyStr:  nil,
-	EmptyMap:  nil,
+		EmptyList: nil,
-}
+		EmptyMap:  nil,
+	}
+)
 
 var pointerTestToml = []byte(`List = ["Hello back"]
 ListPtr = ["Hello"]
@@ -538,15 +516,17 @@ func TestUnmarshalTypeMismatch(t *testing.T) {
 
 type nestedMarshalTestStruct struct {
 	String [][]string
-	//Struct [][]basicMarshalTestSubStruct
+	// Struct [][]basicMarshalTestSubStruct
 	StringPtr *[]*[]*string
 	// StructPtr *[]*[]*basicMarshalTestSubStruct
 }
 
-var str1 = "Three"
+var (
-var str2 = "Four"
+	str1    = "Three"
-var strPtr = []*string{&str1, &str2}
+	str2    = "Four"
-var strPtr2 = []*[]*string{&strPtr}
+	strPtr  = []*string{&str1, &str2}
+	strPtr2 = []*[]*string{&strPtr}
+)
 
 var nestedTestData = nestedMarshalTestStruct{
 	String:    [][]string{{"Five", "Six"}, {"One", "Two"}},
@@ -582,10 +562,12 @@ func (c customMarshaler) MarshalTOML() ([]byte, error) {
 var customMarshalerData = customMarshaler{FirstName: "Sally", LastName: "Fields"}
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customMarshalerToml = []byte(`Sally Fields`)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var nestedCustomMarshalerData = customMarshalerParent{
 	Self:    customMarshaler{FirstName: "Maiku", LastName: "Suteda"},
@@ -593,10 +575,12 @@ var nestedCustomMarshalerData = customMarshalerParent{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var nestedCustomMarshalerToml = []byte(`friends = ["Sally Fields"]
 me = "Maiku Suteda"
 `)
+
 var nestedCustomMarshalerTomlForUnmarshal = []byte(`[friends]
 FirstName = "Sally"
 LastName = "Fields"`)
@@ -613,11 +597,11 @@ func (x *IntOrString) MarshalTOML() ([]byte, error) {
 }
 
 func TestUnmarshalTextMarshaler(t *testing.T) {
-	var nested = struct {
+	nested := struct {
 		Friends textMarshaler `toml:"friends"`
 	}{}
 
-	var expected = struct {
+	expected := struct {
 		Friends textMarshaler `toml:"friends"`
 	}{
 		Friends: textMarshaler{FirstName: "Sally", LastName: "Fields"},
@@ -633,6 +617,7 @@ func TestUnmarshalTextMarshaler(t *testing.T) {
 }
 
 // TODO: Remove nolint once type and methods are used by a test
+//
 //nolint:unused
 type precedentMarshaler struct {
 	FirstName string
@@ -651,6 +636,7 @@ func (m precedentMarshaler) MarshalTOML() ([]byte, error) {
 }
 
 // TODO: Remove nolint once type and method are used by a test
+//
 //nolint:unused
 type customPointerMarshaler struct {
 	FirstName string
@@ -663,6 +649,7 @@ func (m *customPointerMarshaler) MarshalTOML() ([]byte, error) {
 }
 
 // TODO: Remove nolint once type and method are used by a test
+//
 //nolint:unused
 type textPointerMarshaler struct {
 	FirstName string
@@ -675,6 +662,7 @@ func (m *textPointerMarshaler) MarshalText() ([]byte, error) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var commentTestToml = []byte(`
 # it's a comment on type
@@ -712,6 +700,7 @@ type mapsTestStruct struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapsTestData = mapsTestStruct{
 	Simple: map[string]string{
@@ -735,6 +724,7 @@ var mapsTestData = mapsTestStruct{
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var mapsTestToml = []byte(`
 [Other]
@@ -757,6 +747,7 @@ var mapsTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode,unused
 type structArrayNoTag struct {
 	A struct {
@@ -766,6 +757,7 @@ type structArrayNoTag struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customTagTestToml = []byte(`
 [postgres]
@@ -780,6 +772,7 @@ var customTagTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customCommentTagTestToml = []byte(`
 # db connection
@@ -793,6 +786,7 @@ var customCommentTagTestToml = []byte(`
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customCommentedTagTestToml = []byte(`
 [postgres]
@@ -847,6 +841,7 @@ func TestUnmarshalTabInStringAndQuotedKey(t *testing.T) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var customMultilineTagTestToml = []byte(`int_slice = [
   1,
@@ -856,6 +851,7 @@ var customMultilineTagTestToml = []byte(`int_slice = [
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDocBasicToml = []byte(`
 [document]
@@ -868,12 +864,14 @@ var testDocBasicToml = []byte(`
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode
 type testDocCustomTag struct {
 	Doc testDocBasicsCustomTag `file:"document"`
 }
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode
 type testDocBasicsCustomTag struct {
 	Bool       bool      `file:"bool_val"`
@@ -886,6 +884,7 @@ type testDocBasicsCustomTag struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,varcheck
 var testDocCustomTagData = testDocCustomTag{
 	Doc: testDocBasicsCustomTag{
@@ -949,6 +948,29 @@ func TestUnmarshalMapWithTypedKey(t *testing.T) {
 	}
 }
 
+func TestUnmarshalTypeTableHeader(t *testing.T) {
+	testToml := []byte(`
+		[test]
+		a = 1
+		`)
+
+	type header string
+	var result map[header]map[string]int
+	err := toml.Unmarshal(testToml, &result)
+	if err != nil {
+		t.Errorf("Received unexpected error: %s", err)
+		return
+	}
+
+	expected := map[header]map[string]int{
+		"test": map[string]int{"a": 1},
+	}
+
+	if !reflect.DeepEqual(result, expected) {
+		t.Errorf("Bad unmarshal: expected %v, got %v", expected, result)
+	}
+}
+
 func TestUnmarshalNonPointer(t *testing.T) {
 	a := 1
 	err := toml.Unmarshal([]byte{}, a)
@@ -965,6 +987,7 @@ func TestUnmarshalInvalidPointerKind(t *testing.T) {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused
 type testDuration struct {
 	Nanosec   time.Duration  `toml:"nanosec"`
@@ -979,6 +1002,7 @@ type testDuration struct {
 }
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDurationToml = []byte(`
 nanosec = "1ns"
@@ -993,6 +1017,7 @@ a_string = "15s"
 `)
 
 // TODO: Remove nolint once var is used by a test
+//
 //nolint:deadcode,unused,varcheck
 var testDurationToml2 = []byte(`a_string = "15s"
 hour = "1h0m0s"
@@ -1006,6 +1031,7 @@ sec = "1s"
 `)
 
 // TODO: Remove nolint once type is used by a test
+//
 //nolint:deadcode,unused
 type testBadDuration struct {
 	Val time.Duration `toml:"val"`
@@ -1360,7 +1386,6 @@ func TestUnmarshalPreservesUnexportedFields(t *testing.T) {
 	t.Run("unexported field should not be set from toml", func(t *testing.T) {
 		var actual unexportedFieldPreservationTest
 		err := toml.Unmarshal([]byte(doc), &actual)
-
 		if err != nil {
 			t.Fatal("did not expect an error")
 		}
@@ -1394,7 +1419,6 @@ func TestUnmarshalPreservesUnexportedFields(t *testing.T) {
 			Nested3:    &unexportedFieldPreservationTestNested{"baz", "bax"},
 		}
 		err := toml.Unmarshal([]byte(doc), &actual)
-
 		if err != nil {
 			t.Fatal("did not expect an error")
 		}
@@ -1431,7 +1455,6 @@ func TestUnmarshalLocalDate(t *testing.T) {
 		var obj dateStruct
 
 		err := toml.Unmarshal([]byte(doc), &obj)
-
 		if err != nil {
 			t.Fatal(err)
 		}
@@ -1457,7 +1480,6 @@ func TestUnmarshalLocalDate(t *testing.T) {
 		var obj dateStruct
 
 		err := toml.Unmarshal([]byte(doc), &obj)
-
 		if err != nil {
 			t.Fatal(err)
 		}
@@ -1484,32 +1506,34 @@ func TestUnmarshalLocalDateTime(t *testing.T) {
 			name: "normal",
 			in:   "1979-05-27T07:32:00",
 			out: toml.LocalDateTime{
-				Date: toml.LocalDate{
+				LocalDate: toml.LocalDate{
 					Year:  1979,
 					Month: 5,
 					Day:   27,
 				},
-				Time: toml.LocalTime{
+				LocalTime: toml.LocalTime{
 					Hour:       7,
 					Minute:     32,
 					Second:     0,
 					Nanosecond: 0,
 				},
-			}},
+			},
+		},
 		{
 			name: "with nanoseconds",
 			in:   "1979-05-27T00:32:00.999999",
 			out: toml.LocalDateTime{
-				Date: toml.LocalDate{
+				LocalDate: toml.LocalDate{
 					Year:  1979,
 					Month: 5,
 					Day:   27,
 				},
-				Time: toml.LocalTime{
+				LocalTime: toml.LocalTime{
 					Hour:       0,
 					Minute:     32,
 					Second:     0,
 					Nanosecond: 999999000,
+					Precision:  6,
 				},
 			},
 		},
@@ -1526,7 +1550,6 @@ func TestUnmarshalLocalDateTime(t *testing.T) {
 			var obj dateStruct
 
 			err := toml.Unmarshal([]byte(doc), &obj)
-
 			if err != nil {
 				t.Fatal(err)
 			}
@@ -1544,31 +1567,30 @@ func TestUnmarshalLocalDateTime(t *testing.T) {
 			var obj dateStruct
 
 			err := toml.Unmarshal([]byte(doc), &obj)
-
 			if err != nil {
 				t.Fatal(err)
 			}
 
-			if obj.Date.Year() != example.out.Date.Year {
+			if obj.Date.Year() != example.out.Year {
-				t.Errorf("expected year %d, got %d", example.out.Date.Year, obj.Date.Year())
+				t.Errorf("expected year %d, got %d", example.out.Year, obj.Date.Year())
 			}
-			if obj.Date.Month() != example.out.Date.Month {
+			if obj.Date.Month() != time.Month(example.out.Month) {
-				t.Errorf("expected month %d, got %d", example.out.Date.Month, obj.Date.Month())
+				t.Errorf("expected month %d, got %d", example.out.Month, obj.Date.Month())
 			}
-			if obj.Date.Day() != example.out.Date.Day {
+			if obj.Date.Day() != example.out.Day {
-				t.Errorf("expected day %d, got %d", example.out.Date.Day, obj.Date.Day())
+				t.Errorf("expected day %d, got %d", example.out.Day, obj.Date.Day())
 			}
-			if obj.Date.Hour() != example.out.Time.Hour {
+			if obj.Date.Hour() != example.out.Hour {
-				t.Errorf("expected hour %d, got %d", example.out.Time.Hour, obj.Date.Hour())
+				t.Errorf("expected hour %d, got %d", example.out.Hour, obj.Date.Hour())
 			}
-			if obj.Date.Minute() != example.out.Time.Minute {
+			if obj.Date.Minute() != example.out.Minute {
-				t.Errorf("expected minute %d, got %d", example.out.Time.Minute, obj.Date.Minute())
+				t.Errorf("expected minute %d, got %d", example.out.Minute, obj.Date.Minute())
 			}
-			if obj.Date.Second() != example.out.Time.Second {
+			if obj.Date.Second() != example.out.Second {
-				t.Errorf("expected second %d, got %d", example.out.Time.Second, obj.Date.Second())
+				t.Errorf("expected second %d, got %d", example.out.Second, obj.Date.Second())
 			}
-			if obj.Date.Nanosecond() != example.out.Time.Nanosecond {
+			if obj.Date.Nanosecond() != example.out.Nanosecond {
-				t.Errorf("expected nanoseconds %d, got %d", example.out.Time.Nanosecond, obj.Date.Nanosecond())
+				t.Errorf("expected nanoseconds %d, got %d", example.out.Nanosecond, obj.Date.Nanosecond())
 			}
 		})
 	}
@@ -1598,6 +1620,7 @@ func TestUnmarshalLocalTime(t *testing.T) {
 				Minute:     32,
 				Second:     0,
 				Nanosecond: 999999000,
+				Precision:  6,
 			},
 		},
 	}
@@ -1613,7 +1636,6 @@ func TestUnmarshalLocalTime(t *testing.T) {
 			var obj dateStruct
 
 			err := toml.Unmarshal([]byte(doc), &obj)
-
 			if err != nil {
 				t.Fatal(err)
 			}
@@ -1951,7 +1973,7 @@ func decoder(doc string) *toml.Decoder {
 
 func strictDecoder(doc string) *toml.Decoder {
 	d := decoder(doc)
-	d.SetStrict(true)
+	d.DisallowUnknownFields()
 	return d
 }
 
@@ -2283,8 +2305,7 @@ func (d *durationString) UnmarshalTOML(v interface{}) error {
 	return nil
 }
 
-type config437Error struct {
+type config437Error struct{}
-}
 
 func (e *config437Error) UnmarshalTOML(v interface{}) error {
 	return errors.New("expected")

internal/testsuite/add.go

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

internal/testsuite/json.go

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

internal/testsuite/parser.go

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

internal/testsuite/rm.go

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

internal/testsuite/testsuite.go

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

internal/tracker/key.go

@@ -11,19 +11,19 @@ type KeyTracker struct {
 }
 
 // UpdateTable sets the state of the tracker with the AST table node.
-func (t *KeyTracker) UpdateTable(node ast.Node) {
+func (t *KeyTracker) UpdateTable(node *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) {
+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) {
+func (t *KeyTracker) Push(node *ast.Node) {
 	it := node.Key()
 	for it.Next() {
 		t.k = append(t.k, string(it.Node().Data))
@@ -31,7 +31,7 @@ func (t *KeyTracker) Push(node ast.Node) {
 }
 
 // Pop key from stack.
-func (t *KeyTracker) Pop(node ast.Node) {
+func (t *KeyTracker) Pop(node *ast.Node) {
 	it := node.Key()
 	for it.Next() {
 		t.k = t.k[:len(t.k)-1]

internal/tracker/seen.go

@@ -1,7 +1,9 @@
 package tracker
 
 import (
+	"bytes"
 	"fmt"
+	"sync"
 
 	"github.com/pelletier/go-toml/v2/internal/ast"
 )
@@ -29,67 +31,132 @@ func (k keyKind) String() string {
 	panic("missing keyKind string mapping")
 }
 
-// SeenTracker tracks which keys have been seen with which TOML type to flag duplicates
+// SeenTracker tracks which keys have been seen with which TOML type to flag
-// and mismatches according to the spec.
+// 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 {
-	root    *info
+	entries    []entry
-	current *info
+	currentIdx int
 }
 
-type info struct {
+var pool sync.Pool
-	parent   *info
+
+func (s *SeenTracker) reset() {
+	// Always contains a root element at index 0.
+	s.currentIdx = 0
+	if len(s.entries) == 0 {
+		s.entries = make([]entry, 1, 2)
+	} else {
+		s.entries = s.entries[:1]
+	}
+	s.entries[0].child = -1
+	s.entries[0].next = -1
+}
+
+type entry struct {
+	// Use -1 to indicate no child or no sibling.
+	child int
+	next  int
+
+	name     []byte
 	kind     keyKind
-	children map[string]*info
 	explicit bool
+	kv       bool
 }
 
-func (i *info) Clear() {
+// Find the index of the child of parentIdx with key k. Returns -1 if
-	i.children = nil
+// it does not exist.
+func (s *SeenTracker) find(parentIdx int, k []byte) int {
+	for i := s.entries[parentIdx].child; i >= 0; i = s.entries[i].next {
+		if bytes.Equal(s.entries[i].name, k) {
+			return i
+		}
+	}
+	return -1
 }
 
-func (i *info) Has(k string) (*info, bool) {
+// Remove all descendants of node at position idx.
-	c, ok := i.children[k]
+func (s *SeenTracker) clear(idx int) {
-	return c, ok
+	if idx >= len(s.entries) {
-}
+		return
-
-func (i *info) SetKind(kind keyKind) {
-	i.kind = kind
-}
-
-func (i *info) CreateTable(k string, explicit bool) *info {
-	return i.createChild(k, tableKind, explicit)
-}
-
-func (i *info) CreateArrayTable(k string, explicit bool) *info {
-	return i.createChild(k, arrayTableKind, explicit)
-}
-
-func (i *info) createChild(k string, kind keyKind, explicit bool) *info {
-	if i.children == nil {
-		i.children = make(map[string]*info, 1)
 	}
 
-	x := &info{
+	for i := s.entries[idx].child; i >= 0; {
-		parent:   i,
+		next := s.entries[i].next
+		n := s.entries[0].next
+		s.entries[0].next = i
+		s.entries[i].next = n
+		s.entries[i].name = nil
+		s.clear(i)
+		i = next
+	}
+
+	s.entries[idx].child = -1
+}
+
+func (s *SeenTracker) create(parentIdx int, name []byte, kind keyKind, explicit bool, kv bool) int {
+	e := entry{
+		child: -1,
+		next:  s.entries[parentIdx].child,
+
+		name:     name,
 		kind:     kind,
 		explicit: explicit,
+		kv:       kv,
 	}
-	i.children[k] = x
+	var idx int
-	return x
+	if s.entries[0].next >= 0 {
+		idx = s.entries[0].next
+		s.entries[0].next = s.entries[idx].next
+		s.entries[idx] = e
+	} else {
+		idx = len(s.entries)
+		s.entries = append(s.entries, e)
+	}
+
+	s.entries[parentIdx].child = idx
+
+	return idx
 }
 
-// CheckExpression takes a top-level node and checks that it does not contain keys
+func (s *SeenTracker) setExplicitFlag(parentIdx int) {
-// that have been seen in previous calls, and validates that types are consistent.
+	for i := s.entries[parentIdx].child; i >= 0; i = s.entries[i].next {
-func (s *SeenTracker) CheckExpression(node ast.Node) error {
+		if s.entries[i].kv {
-	if s.root == nil {
+			s.entries[i].explicit = true
-		s.root = &info{
+			s.entries[i].kv = false
-			kind: tableKind,
 		}
-		s.current = s.root
+		s.setExplicitFlag(i)
+	}
+}
+
+// CheckExpression takes a top-level node and checks that it does not contain
+// keys that have been seen in previous calls, and validates that types are
+// consistent.
+func (s *SeenTracker) CheckExpression(node *ast.Node) error {
+	if s.entries == nil {
+		s.reset()
 	}
 	switch node.Kind {
 	case ast.KeyValue:
-		return s.checkKeyValue(s.current, node)
+		return s.checkKeyValue(node)
 	case ast.Table:
 		return s.checkTable(node)
 	case ast.ArrayTable:
@@ -97,104 +164,193 @@ func (s *SeenTracker) CheckExpression(node ast.Node) error {
 	default:
 		panic(fmt.Errorf("this should not be a top level node type: %s", node.Kind))
 	}
-
 }
-func (s *SeenTracker) checkTable(node ast.Node) error {
+
-	s.current = s.root
+func (s *SeenTracker) checkTable(node *ast.Node) error {
+	if s.currentIdx >= 0 {
+		s.setExplicitFlag(s.currentIdx)
+	}
 
 	it := node.Key()
-	// handle the first parts of the key, excluding the last one
+
+	parentIdx := 0
+
+	// This code is duplicated in checkArrayTable. This is because factoring
+	// it in a function requires to copy the iterator, or allocate it to the
+	// heap, which is not cheap.
 	for it.Next() {
-		if !it.Node().Next().Valid() {
+		if it.IsLast() {
 			break
 		}
 
-		k := string(it.Node().Data)
+		k := it.Node().Data
-		child, found := s.current.Has(k)
-		if !found {
-			child = s.current.CreateTable(k, false)
-		}
-		s.current = child
-	}
 
-	// handle the last part of the key
+		idx := s.find(parentIdx, k)
-	k := string(it.Node().Data)
 
-	i, found := s.current.Has(k)
+		if idx < 0 {
-	if found {
+			idx = s.create(parentIdx, k, tableKind, false, false)
-		if i.kind != tableKind {
-			return fmt.Errorf("toml: key %s should be a table, not a %s", k, i.kind)
-		}
-		if i.explicit {
-			return fmt.Errorf("toml: table %s already exists", k)
-		}
-		i.explicit = true
-		s.current = i
-	} else {
-		s.current = s.current.CreateTable(k, true)
-	}
-
-	return nil
-}
-
-func (s *SeenTracker) checkArrayTable(node ast.Node) error {
-	s.current = s.root
-
-	it := node.Key()
-
-	// handle the first parts of the key, excluding the last one
-	for it.Next() {
-		if !it.Node().Next().Valid() {
-			break
-		}
-
-		k := string(it.Node().Data)
-		child, found := s.current.Has(k)
-		if !found {
-			child = s.current.CreateTable(k, false)
-		}
-		s.current = child
-	}
-
-	// handle the last part of the key
-	k := string(it.Node().Data)
-
-	info, found := s.current.Has(k)
-	if found {
-		if info.kind != arrayTableKind {
-			return fmt.Errorf("toml: key %s already exists as a %s,  but should be an array table", info.kind, k)
-		}
-		info.Clear()
-	} else {
-		info = s.current.CreateArrayTable(k, true)
-	}
-
-	s.current = info
-	return nil
-}
-
-func (s *SeenTracker) checkKeyValue(context *info, node ast.Node) error {
-	it := node.Key()
-
-	// handle the first parts of the key, excluding the last one
-	for it.Next() {
-		k := string(it.Node().Data)
-		child, found := context.Has(k)
-		if found {
-			if child.kind != tableKind {
-				return fmt.Errorf("toml: expected %s to be a table, not a %s", k, child.kind)
-			}
 		} else {
-			child = context.CreateTable(k, false)
+			entry := s.entries[idx]
+			if entry.kind == valueKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			}
 		}
-		context = child
+		parentIdx = idx
 	}
 
-	if node.Value().Kind == ast.InlineTable {
+	k := it.Node().Data
-		context.SetKind(tableKind)
+	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 {
-		context.SetKind(valueKind)
+		idx = s.create(parentIdx, k, tableKind, true, false)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkArrayTable(node *ast.Node) error {
+	if s.currentIdx >= 0 {
+		s.setExplicitFlag(s.currentIdx)
+	}
+
+	it := node.Key()
+
+	parentIdx := 0
+
+	for it.Next() {
+		if it.IsLast() {
+			break
+		}
+
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false, false)
+		} else {
+			entry := s.entries[idx]
+			if entry.kind == valueKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			}
+		}
+
+		parentIdx = idx
+	}
+
+	k := it.Node().Data
+	idx := s.find(parentIdx, k)
+
+	if idx >= 0 {
+		kind := s.entries[idx].kind
+		if kind != arrayTableKind {
+			return fmt.Errorf("toml: key %s already exists as a %s,  but should be an array table", kind, string(k))
+		}
+		s.clear(idx)
+	} else {
+		idx = s.create(parentIdx, k, arrayTableKind, true, false)
+	}
+
+	s.currentIdx = idx
+
+	return nil
+}
+
+func (s *SeenTracker) checkKeyValue(node *ast.Node) error {
+	parentIdx := s.currentIdx
+	it := node.Key()
+
+	for it.Next() {
+		k := it.Node().Data
+
+		idx := s.find(parentIdx, k)
+
+		if idx < 0 {
+			idx = s.create(parentIdx, k, tableKind, false, true)
+		} else {
+			entry := s.entries[idx]
+			if it.IsLast() {
+				return fmt.Errorf("toml: key %s is already defined", string(k))
+			} else if entry.kind != tableKind {
+				return fmt.Errorf("toml: expected %s to be a table, not a %s", string(k), entry.kind)
+			} else if entry.explicit {
+				return fmt.Errorf("toml: cannot redefine table %s that has already been explicitly defined", string(k))
+			}
+		}
+
+		parentIdx = idx
+	}
+
+	s.entries[parentIdx].kind = valueKind
+
+	value := node.Value()
+
+	switch value.Kind {
+	case ast.InlineTable:
+		return s.checkInlineTable(value)
+	case ast.Array:
+		return s.checkArray(value)
 	}
 
 	return nil
 }
+
+func (s *SeenTracker) checkArray(node *ast.Node) error {
+	it := node.Children()
+	for it.Next() {
+		n := it.Node()
+		switch n.Kind {
+		case ast.InlineTable:
+			err := s.checkInlineTable(n)
+			if err != nil {
+				return err
+			}
+		case ast.Array:
+			err := s.checkArray(n)
+			if err != nil {
+				return err
+			}
+		}
+	}
+	return nil
+}
+
+func (s *SeenTracker) checkInlineTable(node *ast.Node) error {
+	if pool.New == nil {
+		pool.New = func() interface{} {
+			return &SeenTracker{}
+		}
+	}
+
+	s = pool.Get().(*SeenTracker)
+	s.reset()
+
+	it := node.Children()
+	for it.Next() {
+		n := it.Node()
+		err := s.checkKeyValue(n)
+		if err != nil {
+			return err
+		}
+	}
+
+	// As inline tables are self-contained, the tracker does not
+	// need to retain the details of what they contain. The
+	// keyValue element that creates the inline table is kept to
+	// mark the presence of the inline table and prevent
+	// redefinition of its keys: check* functions cannot walk into
+	// a value.
+	pool.Put(s)
+	return nil
+}

internal/tracker/seen_test.go

@@ -0,0 +1,16 @@
+package tracker
+
+import (
+	"testing"
+	"unsafe"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestEntrySize(t *testing.T) {
+	// Validate no regression on the size of entry{}. This is a critical bit for
+	// performance of unmarshaling documents. Should only be increased with care
+	// and a very good reason.
+	maxExpectedEntrySize := 48
+	require.LessOrEqual(t, int(unsafe.Sizeof(entry{})), maxExpectedEntrySize)
+}

localtime.go

@@ -1,300 +1,120 @@
-// Implementation of TOML's local date/time.
-// Copied over from https://github.com/googleapis/google-cloud-go/blob/master/civil/civil.go
-// to avoid pulling all the Google dependencies.
-//
-// Copyright 2016 Google LLC
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-// Package civil implements types for civil time, a time-zone-independent
-// representation of time that follows the rules of the proleptic
-// Gregorian calendar with exactly 24-hour days, 60-minute hours, and 60-second
-// minutes.
-//
-// Because they lack location information, these types do not represent unique
-// moments or intervals of time. Use time.Time for that purpose.
-
 package toml
 
 import (
 	"fmt"
+	"strings"
 	"time"
 )
 
-// A LocalDate represents a date (year, month, day).
+// LocalDate represents a calendar day in no specific timezone.
-//
-// This type does not include location information, and therefore does not
-// describe a unique 24-hour timespan.
 type LocalDate struct {
-	Year  int        // Year (e.g., 2014).
+	Year  int
-	Month time.Month // Month of the year (January = 1, ...).
+	Month int
-	Day   int        // Day of the month, starting at 1.
+	Day   int
 }
 
-// LocalDateOf returns the LocalDate in which a time occurs in that time's location.
+// AsTime converts d into a specific time instance at midnight in zone.
-func LocalDateOf(t time.Time) LocalDate {
+func (d LocalDate) AsTime(zone *time.Location) time.Time {
-	var d LocalDate
+	return time.Date(d.Year, time.Month(d.Month), d.Day, 0, 0, 0, 0, zone)
-	d.Year, d.Month, d.Day = t.Date()
-
-	return d
 }
 
-// ParseLocalDate parses a string in RFC3339 full-date format and returns the date value it represents.
+// String returns RFC 3339 representation of d.
-func ParseLocalDate(s string) (LocalDate, error) {
-	t, err := time.Parse("2006-01-02", s)
-	if err != nil {
-		return LocalDate{}, err
-	}
-
-	return LocalDateOf(t), nil
-}
-
-// String returns the date in RFC3339 full-date format.
 func (d LocalDate) String() string {
 	return fmt.Sprintf("%04d-%02d-%02d", d.Year, d.Month, d.Day)
 }
 
-// IsValid reports whether the date is valid.
+// MarshalText returns RFC 3339 representation of d.
-func (d LocalDate) IsValid() bool {
-	return LocalDateOf(d.In(time.UTC)) == d
-}
-
-// In returns the time corresponding to time 00:00:00 of the date in the location.
-//
-// In is always consistent with time.LocalDate, even when time.LocalDate returns a time
-// on a different day. For example, if loc is America/Indiana/Vincennes, then both
-//     time.LocalDate(1955, time.May, 1, 0, 0, 0, 0, loc)
-// and
-//     civil.LocalDate{Year: 1955, Month: time.May, Day: 1}.In(loc)
-// return 23:00:00 on April 30, 1955.
-//
-// In panics if loc is nil.
-func (d LocalDate) In(loc *time.Location) time.Time {
-	return time.Date(d.Year, d.Month, d.Day, 0, 0, 0, 0, loc)
-}
-
-// AddDays returns the date that is n days in the future.
-// n can also be negative to go into the past.
-func (d LocalDate) AddDays(n int) LocalDate {
-	return LocalDateOf(d.In(time.UTC).AddDate(0, 0, n))
-}
-
-// DaysSince returns the signed number of days between the date and s, not including the end day.
-// This is the inverse operation to AddDays.
-func (d LocalDate) DaysSince(s LocalDate) (days int) {
-	// We convert to Unix time so we do not have to worry about leap seconds:
-	// Unix time increases by exactly 86400 seconds per day.
-	deltaUnix := d.In(time.UTC).Unix() - s.In(time.UTC).Unix()
-
-	const secondsInADay = 86400
-
-	return int(deltaUnix / secondsInADay)
-}
-
-// Before reports whether d1 occurs before future date.
-func (d LocalDate) Before(future LocalDate) bool {
-	if d.Year != future.Year {
-		return d.Year < future.Year
-	}
-
-	if d.Month != future.Month {
-		return d.Month < future.Month
-	}
-
-	return d.Day < future.Day
-}
-
-// After reports whether d1 occurs after past date.
-func (d LocalDate) After(past LocalDate) bool {
-	return past.Before(d)
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of d.String().
 func (d LocalDate) MarshalText() ([]byte, error) {
 	return []byte(d.String()), nil
 }
 
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
+// UnmarshalText parses b using RFC 3339 to fill d.
-// The date is expected to be a string in a format accepted by ParseLocalDate.
+func (d *LocalDate) UnmarshalText(b []byte) error {
-func (d *LocalDate) UnmarshalText(data []byte) error {
+	res, err := parseLocalDate(b)
-	var err error
+	if err != nil {
-	*d, err = ParseLocalDate(string(data))
+		return err
-
+	}
-	return err
+	*d = res
+	return nil
 }
 
-// A LocalTime represents a time with nanosecond precision.
+// LocalTime represents a time of day of no specific day in no specific
-//
+// timezone.
-// This type does not include location information, and therefore does not
-// describe a unique moment in time.
-//
-// This type exists to represent the TIME type in storage-based APIs like BigQuery.
-// Most operations on Times are unlikely to be meaningful. Prefer the LocalDateTime type.
 type LocalTime struct {
-	Hour       int // The hour of the day in 24-hour format; range [0-23]
+	Hour       int // Hour of the day: [0; 24[
-	Minute     int // The minute of the hour; range [0-59]
+	Minute     int // Minute of the hour: [0; 60[
-	Second     int // The second of the minute; range [0-59]
+	Second     int // Second of the minute: [0; 60[
-	Nanosecond int // The nanosecond of the second; range [0-999999999]
+	Nanosecond int // Nanoseconds within the second:  [0, 1000000000[
+	Precision  int // Number of digits to display for Nanosecond.
 }
 
-// LocalTimeOf returns the LocalTime representing the time of day in which a time occurs
+// String returns RFC 3339 representation of d.
-// in that time's location. It ignores the date.
+// If d.Nanosecond and d.Precision are zero, the time won't have a nanosecond
-func LocalTimeOf(t time.Time) LocalTime {
+// component. If d.Nanosecond > 0 but d.Precision = 0, then the minimum number
-	var tm LocalTime
+// of digits for nanoseconds is provided.
-	tm.Hour, tm.Minute, tm.Second = t.Clock()
+func (d LocalTime) String() string {
-	tm.Nanosecond = t.Nanosecond()
+	s := fmt.Sprintf("%02d:%02d:%02d", d.Hour, d.Minute, d.Second)
 
-	return tm
+	if d.Precision > 0 {
+		s += fmt.Sprintf(".%09d", d.Nanosecond)[:d.Precision+1]
+	} else if d.Nanosecond > 0 {
+		// Nanoseconds are specified, but precision is not provided. Use the
+		// minimum.
+		s += strings.Trim(fmt.Sprintf(".%09d", d.Nanosecond), "0")
+	}
+
+	return s
 }
 
-// ParseLocalTime parses a string and returns the time value it represents.
+// MarshalText returns RFC 3339 representation of d.
-// ParseLocalTime accepts an extended form of the RFC3339 partial-time format. After
+func (d LocalTime) MarshalText() ([]byte, error) {
-// the HH:MM:SS part of the string, an optional fractional part may appear,
+	return []byte(d.String()), nil
-// consisting of a decimal point followed by one to nine decimal digits.
+}
-// (RFC3339 admits only one digit after the decimal point).
+
-func ParseLocalTime(s string) (LocalTime, error) {
+// UnmarshalText parses b using RFC 3339 to fill d.
-	t, err := time.Parse("15:04:05.999999999", s)
+func (d *LocalTime) UnmarshalText(b []byte) error {
+	res, left, err := parseLocalTime(b)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
+	}
 	if err != nil {
-		return LocalTime{}, err
+		return err
 	}
-
+	*d = res
-	return LocalTimeOf(t), nil
+	return nil
 }
 
-// String returns the date in the format described in ParseLocalTime. If Nanoseconds
+// LocalDateTime represents a time of a specific day in no specific timezone.
-// is zero, no fractional part will be generated. Otherwise, the result will
-// end with a fractional part consisting of a decimal point and nine digits.
-func (t LocalTime) String() string {
-	s := fmt.Sprintf("%02d:%02d:%02d", t.Hour, t.Minute, t.Second)
-	if t.Nanosecond == 0 {
-		return s
-	}
-
-	return s + fmt.Sprintf(".%09d", t.Nanosecond)
-}
-
-// IsValid reports whether the time is valid.
-func (t LocalTime) IsValid() bool {
-	// Construct a non-zero time.
-	tm := time.Date(2, 2, 2, t.Hour, t.Minute, t.Second, t.Nanosecond, time.UTC)
-
-	return LocalTimeOf(tm) == t
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of t.String().
-func (t LocalTime) MarshalText() ([]byte, error) {
-	return []byte(t.String()), nil
-}
-
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
-// The time is expected to be a string in a format accepted by ParseLocalTime.
-func (t *LocalTime) UnmarshalText(data []byte) error {
-	var err error
-	*t, err = ParseLocalTime(string(data))
-
-	return err
-}
-
-// A LocalDateTime represents a date and time.
-//
-// This type does not include location information, and therefore does not
-// describe a unique moment in time.
 type LocalDateTime struct {
-	Date LocalDate
+	LocalDate
-	Time LocalTime
+	LocalTime
 }
 
-// Note: We deliberately do not embed LocalDate into LocalDateTime, to avoid promoting AddDays and Sub.
+// AsTime converts d into a specific time instance in zone.
+func (d LocalDateTime) AsTime(zone *time.Location) time.Time {
+	return time.Date(d.Year, time.Month(d.Month), d.Day, d.Hour, d.Minute, d.Second, d.Nanosecond, zone)
+}
 
-// LocalDateTimeOf returns the LocalDateTime in which a time occurs in that time's location.
+// String returns RFC 3339 representation of d.
-func LocalDateTimeOf(t time.Time) LocalDateTime {
+func (d LocalDateTime) String() string {
-	return LocalDateTime{
+	return d.LocalDate.String() + "T" + d.LocalTime.String()
-		Date: LocalDateOf(t),
+}
-		Time: LocalTimeOf(t),
+
+// MarshalText returns RFC 3339 representation of d.
+func (d LocalDateTime) MarshalText() ([]byte, error) {
+	return []byte(d.String()), nil
+}
+
+// UnmarshalText parses b using RFC 3339 to fill d.
+func (d *LocalDateTime) UnmarshalText(data []byte) error {
+	res, left, err := parseLocalDateTime(data)
+	if err == nil && len(left) != 0 {
+		err = newDecodeError(left, "extra characters")
 	}
-}
-
-// ParseLocalDateTime parses a string and returns the LocalDateTime it represents.
-// ParseLocalDateTime accepts a variant of the RFC3339 date-time format that omits
-// the time offset but includes an optional fractional time, as described in
-// ParseLocalTime. Informally, the accepted format is
-//     YYYY-MM-DDTHH:MM:SS[.FFFFFFFFF]
-// where the 'T' may be a lower-case 't'.
-func ParseLocalDateTime(s string) (LocalDateTime, error) {
-	t, err := time.Parse("2006-01-02T15:04:05.999999999", s)
 	if err != nil {
-		t, err = time.Parse("2006-01-02t15:04:05.999999999", s)
+		return err
-		if err != nil {
-			return LocalDateTime{}, err
-		}
 	}
 
-	return LocalDateTimeOf(t), nil
+	*d = res
-}
+	return nil
-
-// String returns the date in the format described in ParseLocalDate.
-func (dt LocalDateTime) String() string {
-	return dt.Date.String() + "T" + dt.Time.String()
-}
-
-// IsValid reports whether the datetime is valid.
-func (dt LocalDateTime) IsValid() bool {
-	return dt.Date.IsValid() && dt.Time.IsValid()
-}
-
-// In returns the time corresponding to the LocalDateTime in the given location.
-//
-// If the time is missing or ambigous at the location, In returns the same
-// result as time.LocalDate. For example, if loc is America/Indiana/Vincennes, then
-// both
-//     time.LocalDate(1955, time.May, 1, 0, 30, 0, 0, loc)
-// and
-//     civil.LocalDateTime{
-//         civil.LocalDate{Year: 1955, Month: time.May, Day: 1}},
-//         civil.LocalTime{Minute: 30}}.In(loc)
-// return 23:30:00 on April 30, 1955.
-//
-// In panics if loc is nil.
-func (dt LocalDateTime) In(loc *time.Location) time.Time {
-	return time.Date(
-		dt.Date.Year, dt.Date.Month, dt.Date.Day,
-		dt.Time.Hour, dt.Time.Minute, dt.Time.Second, dt.Time.Nanosecond, loc,
-	)
-}
-
-// Before reports whether dt occurs before future.
-func (dt LocalDateTime) Before(future LocalDateTime) bool {
-	return dt.In(time.UTC).Before(future.In(time.UTC))
-}
-
-// After reports whether dt occurs after past.
-func (dt LocalDateTime) After(past LocalDateTime) bool {
-	return past.Before(dt)
-}
-
-// MarshalText implements the encoding.TextMarshaler interface.
-// The output is the result of dt.String().
-func (dt LocalDateTime) MarshalText() ([]byte, error) {
-	return []byte(dt.String()), nil
-}
-
-// UnmarshalText implements the encoding.TextUnmarshaler interface.
-// The datetime is expected to be a string in a format accepted by ParseLocalDateTime.
-func (dt *LocalDateTime) UnmarshalText(data []byte) error {
-	var err error
-	*dt, err = ParseLocalDateTime(string(data))
-
-	return err
 }

localtime_test.go

@@ -1,507 +1,118 @@
-// Copyright 2016 Google LLC
+package toml_test
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//      http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-package toml
 
 import (
-	"encoding/json"
-	"reflect"
 	"testing"
 	"time"
+
+	"github.com/pelletier/go-toml/v2"
+	"github.com/stretchr/testify/require"
 )
 
-func cmpEqual(x, y interface{}) bool {
+func TestLocalDate_AsTime(t *testing.T) {
-	return reflect.DeepEqual(x, y)
+	d := toml.LocalDate{2021, 6, 8}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 0, 0, 0, 0, time.UTC), cast)
 }
 
-func TestDates(t *testing.T) {
+func TestLocalDate_String(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDate{2021, 6, 8}
-
+	require.Equal(t, "2021-06-08", d.String())
-	for _, test := range []struct {
-		date     LocalDate
-		loc      *time.Location
-		wantStr  string
-		wantTime time.Time
-	}{
-		{
-			date:     LocalDate{2014, 7, 29},
-			loc:      time.Local,
-			wantStr:  "2014-07-29",
-			wantTime: time.Date(2014, time.July, 29, 0, 0, 0, 0, time.Local),
-		},
-		{
-			date:     LocalDateOf(time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local)),
-			loc:      time.UTC,
-			wantStr:  "2014-08-20",
-			wantTime: time.Date(2014, 8, 20, 0, 0, 0, 0, time.UTC),
-		},
-		{
-			date:     LocalDateOf(time.Date(999, time.January, 26, 0, 0, 0, 0, time.Local)),
-			loc:      time.UTC,
-			wantStr:  "0999-01-26",
-			wantTime: time.Date(999, 1, 26, 0, 0, 0, 0, time.UTC),
-		},
-	} {
-		if got := test.date.String(); got != test.wantStr {
-			t.Errorf("%#v.String() = %q, want %q", test.date, got, test.wantStr)
-		}
-
-		if got := test.date.In(test.loc); !got.Equal(test.wantTime) {
-			t.Errorf("%#v.In(%v) = %v, want %v", test.date, test.loc, got, test.wantTime)
-		}
-	}
 }
 
-func TestDateIsValid(t *testing.T) {
+func TestLocalDate_MarshalText(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDate{2021, 6, 8}
-
+	b, err := d.MarshalText()
-	for _, test := range []struct {
+	require.NoError(t, err)
-		date LocalDate
+	require.Equal(t, []byte("2021-06-08"), b)
-		want bool
-	}{
-		{LocalDate{2014, 7, 29}, true},
-		{LocalDate{2000, 2, 29}, true},
-		{LocalDate{10000, 12, 31}, true},
-		{LocalDate{1, 1, 1}, true},
-		{LocalDate{0, 1, 1}, true},  // year zero is OK
-		{LocalDate{-1, 1, 1}, true}, // negative year is OK
-		{LocalDate{1, 0, 1}, false},
-		{LocalDate{1, 1, 0}, false},
-		{LocalDate{2016, 1, 32}, false},
-		{LocalDate{2016, 13, 1}, false},
-		{LocalDate{1, -1, 1}, false},
-		{LocalDate{1, 1, -1}, false},
-	} {
-		got := test.date.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.date, got, test.want)
-		}
-	}
 }
 
-func TestParseDate(t *testing.T) {
+func TestLocalDate_UnmarshalMarshalText(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDate{}
+	err := d.UnmarshalText([]byte("2021-06-08"))
+	require.NoError(t, err)
+	require.Equal(t, toml.LocalDate{2021, 6, 8}, d)
 
-	var emptyDate LocalDate
+	err = d.UnmarshalText([]byte("what"))
-
+	require.Error(t, err)
-	for _, test := range []struct {
-		str  string
-		want LocalDate // if empty, expect an error
-	}{
-		{"2016-01-02", LocalDate{2016, 1, 2}},
-		{"2016-12-31", LocalDate{2016, 12, 31}},
-		{"0003-02-04", LocalDate{3, 2, 4}},
-		{"999-01-26", emptyDate},
-		{"", emptyDate},
-		{"2016-01-02x", emptyDate},
-	} {
-		got, err := ParseLocalDate(test.str)
-		if got != test.want {
-			t.Errorf("ParseLocalDate(%q) = %+v, want %+v", test.str, got, test.want)
-		}
-
-		if err != nil && test.want != (emptyDate) {
-			t.Errorf("Unexpected error %v from ParseLocalDate(%q)", err, test.str)
-		}
-	}
 }
 
-func TestDateArithmetic(t *testing.T) {
+func TestLocalTime_String(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalTime{20, 12, 1, 2, 9}
-
+	require.Equal(t, "20:12:01.000000002", d.String())
-	for _, test := range []struct {
+	d = toml.LocalTime{20, 12, 1, 0, 0}
-		desc  string
+	require.Equal(t, "20:12:01", d.String())
-		start LocalDate
+	d = toml.LocalTime{20, 12, 1, 0, 9}
-		end   LocalDate
+	require.Equal(t, "20:12:01.000000000", d.String())
-		days  int
+	d = toml.LocalTime{20, 12, 1, 100, 0}
-	}{
+	require.Equal(t, "20:12:01.0000001", d.String())
-		{
-			desc:  "zero days noop",
-			start: LocalDate{2014, 5, 9},
-			end:   LocalDate{2014, 5, 9},
-			days:  0,
-		},
-		{
-			desc:  "crossing a year boundary",
-			start: LocalDate{2014, 12, 31},
-			end:   LocalDate{2015, 1, 1},
-			days:  1,
-		},
-		{
-			desc:  "negative number of days",
-			start: LocalDate{2015, 1, 1},
-			end:   LocalDate{2014, 12, 31},
-			days:  -1,
-		},
-		{
-			desc:  "full leap year",
-			start: LocalDate{2004, 1, 1},
-			end:   LocalDate{2005, 1, 1},
-			days:  366,
-		},
-		{
-			desc:  "full non-leap year",
-			start: LocalDate{2001, 1, 1},
-			end:   LocalDate{2002, 1, 1},
-			days:  365,
-		},
-		{
-			desc:  "crossing a leap second",
-			start: LocalDate{1972, 6, 30},
-			end:   LocalDate{1972, 7, 1},
-			days:  1,
-		},
-		{
-			desc:  "dates before the unix epoch",
-			start: LocalDate{101, 1, 1},
-			end:   LocalDate{102, 1, 1},
-			days:  365,
-		},
-	} {
-		if got := test.start.AddDays(test.days); got != test.end {
-			t.Errorf("[%s] %#v.AddDays(%v) = %#v, want %#v", test.desc, test.start, test.days, got, test.end)
-		}
-
-		if got := test.end.DaysSince(test.start); got != test.days {
-			t.Errorf("[%s] %#v.Sub(%#v) = %v, want %v", test.desc, test.end, test.start, got, test.days)
-		}
-	}
 }
 
-func TestDateBefore(t *testing.T) {
+func TestLocalTime_MarshalText(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalTime{20, 12, 1, 2, 9}
-
+	b, err := d.MarshalText()
-	for _, test := range []struct {
+	require.NoError(t, err)
-		d1, d2 LocalDate
+	require.Equal(t, []byte("20:12:01.000000002"), b)
-		want   bool
-	}{
-		{LocalDate{2016, 12, 31}, LocalDate{2017, 1, 1}, true},
-		{LocalDate{2016, 1, 1}, LocalDate{2016, 1, 1}, false},
-		{LocalDate{2016, 12, 30}, LocalDate{2016, 12, 31}, true},
-		{LocalDate{2016, 1, 30}, LocalDate{2016, 12, 31}, true},
-	} {
-		if got := test.d1.Before(test.d2); got != test.want {
-			t.Errorf("%v.Before(%v): got %t, want %t", test.d1, test.d2, got, test.want)
-		}
-	}
 }
 
-func TestDateAfter(t *testing.T) {
+func TestLocalTime_UnmarshalMarshalText(t *testing.T) {
-	t.Parallel()
+	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)
 
-	for _, test := range []struct {
+	err = d.UnmarshalText([]byte("what"))
-		d1, d2 LocalDate
+	require.Error(t, err)
-		want   bool
+
-	}{
+	err = d.UnmarshalText([]byte("20:12:01.000000002 bad"))
-		{LocalDate{2016, 12, 31}, LocalDate{2017, 1, 1}, false},
+	require.Error(t, err)
-		{LocalDate{2016, 1, 1}, LocalDate{2016, 1, 1}, false},
-		{LocalDate{2016, 12, 30}, LocalDate{2016, 12, 31}, false},
-	} {
-		if got := test.d1.After(test.d2); got != test.want {
-			t.Errorf("%v.After(%v): got %t, want %t", test.d1, test.d2, got, test.want)
-		}
-	}
 }
 
-func TestTimeToString(t *testing.T) {
+func TestLocalTime_RoundTrip(t *testing.T) {
-	t.Parallel()
+	var d struct{ A toml.LocalTime }
-
+	err := toml.Unmarshal([]byte("a=20:12:01.500"), &d)
-	for _, test := range []struct {
+	require.NoError(t, err)
-		str       string
+	require.Equal(t, "20:12:01.500", d.A.String())
-		time      LocalTime
-		roundTrip bool // ParseLocalTime(str).String() == str?
-	}{
-		{"13:26:33", LocalTime{13, 26, 33, 0}, true},
-		{"01:02:03.000023456", LocalTime{1, 2, 3, 23456}, true},
-		{"00:00:00.000000001", LocalTime{0, 0, 0, 1}, true},
-		{"13:26:03.1", LocalTime{13, 26, 3, 100000000}, false},
-		{"13:26:33.0000003", LocalTime{13, 26, 33, 300}, false},
-	} {
-		gotTime, err := ParseLocalTime(test.str)
-		if err != nil {
-			t.Errorf("ParseLocalTime(%q): got error: %v", test.str, err)
-
-			continue
-		}
-
-		if gotTime != test.time {
-			t.Errorf("ParseLocalTime(%q) = %+v, want %+v", test.str, gotTime, test.time)
-		}
-
-		if test.roundTrip {
-			gotStr := test.time.String()
-			if gotStr != test.str {
-				t.Errorf("%#v.String() = %q, want %q", test.time, gotStr, test.str)
-			}
-		}
-	}
 }
 
-func TestTimeOf(t *testing.T) {
+func TestLocalDateTime_AsTime(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDateTime{
-
+		toml.LocalDate{2021, 6, 8},
-	for _, test := range []struct {
+		toml.LocalTime{20, 12, 1, 2, 9},
-		time time.Time
-		want LocalTime
-	}{
-		{time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local), LocalTime{15, 8, 43, 1}},
-		{time.Date(1, 1, 1, 0, 0, 0, 0, time.UTC), LocalTime{0, 0, 0, 0}},
-	} {
-		if got := LocalTimeOf(test.time); got != test.want {
-			t.Errorf("LocalTimeOf(%v) = %+v, want %+v", test.time, got, test.want)
-		}
 	}
+	cast := d.AsTime(time.UTC)
+	require.Equal(t, time.Date(2021, time.June, 8, 20, 12, 1, 2, time.UTC), cast)
 }
 
-func TestTimeIsValid(t *testing.T) {
+func TestLocalDateTime_String(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDateTime{
-
+		toml.LocalDate{2021, 6, 8},
-	for _, test := range []struct {
+		toml.LocalTime{20, 12, 1, 2, 9},
-		time LocalTime
-		want bool
-	}{
-		{LocalTime{0, 0, 0, 0}, true},
-		{LocalTime{23, 0, 0, 0}, true},
-		{LocalTime{23, 59, 59, 999999999}, true},
-		{LocalTime{24, 59, 59, 999999999}, false},
-		{LocalTime{23, 60, 59, 999999999}, false},
-		{LocalTime{23, 59, 60, 999999999}, false},
-		{LocalTime{23, 59, 59, 1000000000}, false},
-		{LocalTime{-1, 0, 0, 0}, false},
-		{LocalTime{0, -1, 0, 0}, false},
-		{LocalTime{0, 0, -1, 0}, false},
-		{LocalTime{0, 0, 0, -1}, false},
-	} {
-		got := test.time.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.time, got, test.want)
-		}
 	}
+	require.Equal(t, "2021-06-08T20:12:01.000000002", d.String())
 }
 
-func TestDateTimeToString(t *testing.T) {
+func TestLocalDateTime_MarshalText(t *testing.T) {
-	t.Parallel()
+	d := toml.LocalDateTime{
-
+		toml.LocalDate{2021, 6, 8},
-	for _, test := range []struct {
+		toml.LocalTime{20, 12, 1, 2, 9},
-		str       string
-		dateTime  LocalDateTime
-		roundTrip bool // ParseLocalDateTime(str).String() == str?
-	}{
-		{"2016-03-22T13:26:33", LocalDateTime{LocalDate{2016, 3, 22}, LocalTime{13, 26, 33, 0}}, true},
-		{"2016-03-22T13:26:33.000000600", LocalDateTime{LocalDate{2016, 3, 22}, LocalTime{13, 26, 33, 600}}, true},
-		{"2016-03-22t13:26:33", LocalDateTime{LocalDate{2016, 3, 22}, LocalTime{13, 26, 33, 0}}, false},
-	} {
-		gotDateTime, err := ParseLocalDateTime(test.str)
-		if err != nil {
-			t.Errorf("ParseLocalDateTime(%q): got error: %v", test.str, err)
-
-			continue
-		}
-
-		if gotDateTime != test.dateTime {
-			t.Errorf("ParseLocalDateTime(%q) = %+v, want %+v", test.str, gotDateTime, test.dateTime)
-		}
-
-		if test.roundTrip {
-			gotStr := test.dateTime.String()
-			if gotStr != test.str {
-				t.Errorf("%#v.String() = %q, want %q", test.dateTime, gotStr, test.str)
-			}
-		}
 	}
+	b, err := d.MarshalText()
+	require.NoError(t, err)
+	require.Equal(t, []byte("2021-06-08T20:12:01.000000002"), b)
 }
 
-func TestParseDateTimeErrors(t *testing.T) {
+func TestLocalDateTime_UnmarshalMarshalText(t *testing.T) {
-	t.Parallel()
+	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)
 
-	for _, str := range []string{
+	err = d.UnmarshalText([]byte("what"))
-		"",
+	require.Error(t, err)
-		"2016-03-22",           // just a date
+
-		"13:26:33",             // just a time
+	err = d.UnmarshalText([]byte("2021-06-08 20:12:01.000000002 bad"))
-		"2016-03-22 13:26:33",  // wrong separating character
+	require.Error(t, err)
-		"2016-03-22T13:26:33x", // extra at end
-	} {
-		if _, err := ParseLocalDateTime(str); err == nil {
-			t.Errorf("ParseLocalDateTime(%q) succeeded, want error", str)
-		}
-	}
-}
-
-func TestDateTimeOf(t *testing.T) {
-	t.Parallel()
-
-	for _, test := range []struct {
-		time time.Time
-		want LocalDateTime
-	}{
-		{
-			time.Date(2014, 8, 20, 15, 8, 43, 1, time.Local),
-			LocalDateTime{LocalDate{2014, 8, 20}, LocalTime{15, 8, 43, 1}},
-		},
-		{
-			time.Date(1, 1, 1, 0, 0, 0, 0, time.UTC),
-			LocalDateTime{LocalDate{1, 1, 1}, LocalTime{0, 0, 0, 0}},
-		},
-	} {
-		if got := LocalDateTimeOf(test.time); got != test.want {
-			t.Errorf("LocalDateTimeOf(%v) = %+v, want %+v", test.time, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeIsValid(t *testing.T) {
-	t.Parallel()
-
-	// No need to be exhaustive here; it's just LocalDate.IsValid && LocalTime.IsValid.
-	for _, test := range []struct {
-		dt   LocalDateTime
-		want bool
-	}{
-		{LocalDateTime{LocalDate{2016, 3, 20}, LocalTime{0, 0, 0, 0}}, true},
-		{LocalDateTime{LocalDate{2016, -3, 20}, LocalTime{0, 0, 0, 0}}, false},
-		{LocalDateTime{LocalDate{2016, 3, 20}, LocalTime{24, 0, 0, 0}}, false},
-	} {
-		got := test.dt.IsValid()
-		if got != test.want {
-			t.Errorf("%#v: got %t, want %t", test.dt, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeIn(t *testing.T) {
-	t.Parallel()
-
-	dt := LocalDateTime{LocalDate{2016, 1, 2}, LocalTime{3, 4, 5, 6}}
-
-	want := time.Date(2016, 1, 2, 3, 4, 5, 6, time.UTC)
-	if got := dt.In(time.UTC); !got.Equal(want) {
-		t.Errorf("got %v, want %v", got, want)
-	}
-}
-
-func TestDateTimeBefore(t *testing.T) {
-	t.Parallel()
-
-	d1 := LocalDate{2016, 12, 31}
-	d2 := LocalDate{2017, 1, 1}
-	t1 := LocalTime{5, 6, 7, 8}
-	t2 := LocalTime{5, 6, 7, 9}
-
-	for _, test := range []struct {
-		dt1, dt2 LocalDateTime
-		want     bool
-	}{
-		{LocalDateTime{d1, t1}, LocalDateTime{d2, t1}, true},
-		{LocalDateTime{d1, t1}, LocalDateTime{d1, t2}, true},
-		{LocalDateTime{d2, t1}, LocalDateTime{d1, t1}, false},
-		{LocalDateTime{d2, t1}, LocalDateTime{d2, t1}, false},
-	} {
-		if got := test.dt1.Before(test.dt2); got != test.want {
-			t.Errorf("%v.Before(%v): got %t, want %t", test.dt1, test.dt2, got, test.want)
-		}
-	}
-}
-
-func TestDateTimeAfter(t *testing.T) {
-	t.Parallel()
-
-	d1 := LocalDate{2016, 12, 31}
-	d2 := LocalDate{2017, 1, 1}
-	t1 := LocalTime{5, 6, 7, 8}
-	t2 := LocalTime{5, 6, 7, 9}
-
-	for _, test := range []struct {
-		dt1, dt2 LocalDateTime
-		want     bool
-	}{
-		{LocalDateTime{d1, t1}, LocalDateTime{d2, t1}, false},
-		{LocalDateTime{d1, t1}, LocalDateTime{d1, t2}, false},
-		{LocalDateTime{d2, t1}, LocalDateTime{d1, t1}, true},
-		{LocalDateTime{d2, t1}, LocalDateTime{d2, t1}, false},
-	} {
-		if got := test.dt1.After(test.dt2); got != test.want {
-			t.Errorf("%v.After(%v): got %t, want %t", test.dt1, test.dt2, got, test.want)
-		}
-	}
-}
-
-func TestMarshalJSON(t *testing.T) {
-	t.Parallel()
-
-	for _, test := range []struct {
-		value interface{}
-		want  string
-	}{
-		{LocalDate{1987, 4, 15}, `"1987-04-15"`},
-		{LocalTime{18, 54, 2, 0}, `"18:54:02"`},
-		{LocalDateTime{LocalDate{1987, 4, 15}, LocalTime{18, 54, 2, 0}}, `"1987-04-15T18:54:02"`},
-	} {
-		bgot, err := json.Marshal(test.value)
-		if err != nil {
-			t.Fatal(err)
-		}
-
-		if got := string(bgot); got != test.want {
-			t.Errorf("%#v: got %s, want %s", test.value, got, test.want)
-		}
-	}
-}
-
-func TestUnmarshalJSON(t *testing.T) {
-	t.Parallel()
-
-	var (
-		d  LocalDate
-		tm LocalTime
-		dt LocalDateTime
-	)
-
-	for _, test := range []struct {
-		data string
-		ptr  interface{}
-		want interface{}
-	}{
-		{`"1987-04-15"`, &d, &LocalDate{1987, 4, 15}},
-		{`"1987-04-\u0031\u0035"`, &d, &LocalDate{1987, 4, 15}},
-		{`"18:54:02"`, &tm, &LocalTime{18, 54, 2, 0}},
-		{`"1987-04-15T18:54:02"`, &dt, &LocalDateTime{LocalDate{1987, 4, 15}, LocalTime{18, 54, 2, 0}}},
-	} {
-		if err := json.Unmarshal([]byte(test.data), test.ptr); err != nil {
-			t.Fatalf("%s: %v", test.data, err)
-		}
-
-		if !cmpEqual(test.ptr, test.want) {
-			t.Errorf("%s: got %#v, want %#v", test.data, test.ptr, test.want)
-		}
-	}
-
-	for _, bad := range []string{
-		"", `""`, `"bad"`, `"1987-04-15x"`,
-		`19870415`,     // a JSON number
-		`11987-04-15x`, // not a JSON string
-
-	} {
-		if json.Unmarshal([]byte(bad), &d) == nil {
-			t.Errorf("%q, LocalDate: got nil, want error", bad)
-		}
-
-		if json.Unmarshal([]byte(bad), &tm) == nil {
-			t.Errorf("%q, LocalTime: got nil, want error", bad)
-		}
-
-		if json.Unmarshal([]byte(bad), &dt) == nil {
-			t.Errorf("%q, LocalDateTime: got nil, want error", bad)
-		}
-	}
 }

marshaler.go

@@ -5,11 +5,13 @@ import (
 	"encoding"
 	"fmt"
 	"io"
+	"math"
 	"reflect"
 	"sort"
 	"strconv"
 	"strings"
 	"time"
+	"unicode"
 )
 
 // Marshal serializes a Go value as a TOML document.
@@ -52,9 +54,10 @@ func NewEncoder(w io.Writer) *Encoder {
 // This behavior can be controlled on an individual struct field basis with the
 // inline tag:
 //
-//   MyField `inline:"true"`
+//	MyField `toml:",inline"`
-func (enc *Encoder) SetTablesInline(inline bool) {
+func (enc *Encoder) SetTablesInline(inline bool) *Encoder {
 	enc.tablesInline = inline
+	return enc
 }
 
 // SetArraysMultiline forces the encoder to emit all arrays with one element per
@@ -62,28 +65,31 @@ func (enc *Encoder) SetTablesInline(inline bool) {
 //
 // This behavior can be controlled on an individual struct field basis with the multiline tag:
 //
-//   MyField `multiline:"true"`
+//	MyField `multiline:"true"`
-func (enc *Encoder) SetArraysMultiline(multiline bool) {
+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) {
+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) {
+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
+// # Encoding rules
 //
 // A top level slice containing only maps or structs is encoded as [[table
 // array]].
@@ -98,27 +104,52 @@ func (enc *Encoder) SetIndentTables(indent bool) {
 // 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
+// a newline character or a single quote. In that case they are emitted as
-// strings.
+// quoted strings.
+//
+// Unsigned integers larger than math.MaxInt64 cannot be encoded. Doing so
+// results in an error. This rule exists because the TOML specification only
+// requires parsers to support at least the 64 bits integer range. Allowing
+// larger numbers would create non-standard TOML documents, which may not be
+// readable (at best) by other implementations. To encode such numbers, a
+// solution is a custom type that implements encoding.TextMarshaler.
 //
 // When encoding structs, fields are encoded in order of definition, with their
 // exact name.
 //
-// Struct tags
+// Tables and array tables are separated by empty lines. However, consecutive
+// subtables definitions are not. For example:
 //
-// The following struct tags are available to tweak encoding on a per-field
+//	[top1]
-// basis:
 //
-//   toml:"foo"
+//	[top2]
-//     Changes the name of the key to use for the field to foo.
+//	[top2.child1]
 //
-//   multiline:"true"
+//	[[array]]
-//     When the field contains a string, it will be emitted as a quoted
-//     multi-line TOML string.
 //
-//   inline:"true"
+//	[[array]]
-//     When the field would normally be encoded as a table, it is instead
+//	[array.child2]
-//     encoded as an inline table.
+//
+// # Struct tags
+//
+// The encoding of each public struct field can be customized by the format
+// string in the "toml" key of the struct field's tag. This follows
+// encoding/json's convention. The format string starts with the name of the
+// field, optionally followed by a comma-separated list of options. The name may
+// be empty in order to provide options without overriding the default name.
+//
+// The "multiline" option emits strings as quoted multi-line TOML strings. It
+// has no effect on fields that would not be encoded as strings.
+//
+// The "inline" option turns fields that would be emitted as tables into inline
+// tables instead. It has no effect on other fields.
+//
+// The "omitempty" option prevents empty values or groups from being emitted.
+//
+// In addition to the "toml" tag struct tag, a "comment" tag can be used to emit
+// a TOML comment before the value being annotated. Comments are ignored inside
+// inline tables. For array tables, the comment is only present before the first
+// element of the array.
 func (enc *Encoder) Encode(v interface{}) error {
 	var (
 		b   []byte
@@ -146,6 +177,8 @@ func (enc *Encoder) Encode(v interface{}) error {
 
 type valueOptions struct {
 	multiline bool
+	omitempty bool
+	comment   string
 }
 
 type encoderCtx struct {
@@ -195,16 +228,29 @@ 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 := v.Interface()
-		i, ok := v.Interface().(time.Time)
+
-		if ok {
+	switch x := i.(type) {
-			return i.AppendFormat(b, time.RFC3339), nil
+	case time.Time:
+		if x.Nanosecond() > 0 {
+			return x.AppendFormat(b, time.RFC3339Nano), nil
 		}
+		return x.AppendFormat(b, time.RFC3339), nil
+	case LocalTime:
+		return append(b, x.String()...), nil
+	case LocalDate:
+		return append(b, x.String()...), nil
+	case LocalDateTime:
+		return append(b, x.String()...), nil
 	}
 
-	if v.Type().Implements(textMarshalerType) {
+	hasTextMarshaler := v.Type().Implements(textMarshalerType)
+	if hasTextMarshaler || (v.CanAddr() && reflect.PtrTo(v.Type()).Implements(textMarshalerType)) {
+		if !hasTextMarshaler {
+			v = v.Addr()
+		}
+
 		if ctx.isRoot() {
 			return nil, fmt.Errorf("toml: type %s implementing the TextMarshaler interface cannot be a root element", v.Type())
 		}
@@ -244,9 +290,32 @@ func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, e
 	case reflect.String:
 		b = enc.encodeString(b, v.String(), ctx.options)
 	case reflect.Float32:
-		b = strconv.AppendFloat(b, v.Float(), 'f', -1, 32)
+		f := v.Float()
+
+		if math.IsNaN(f) {
+			b = append(b, "nan"...)
+		} else if f > math.MaxFloat32 {
+			b = append(b, "inf"...)
+		} else if f < -math.MaxFloat32 {
+			b = append(b, "-inf"...)
+		} else if math.Trunc(f) == f {
+			b = strconv.AppendFloat(b, f, 'f', 1, 32)
+		} else {
+			b = strconv.AppendFloat(b, f, 'f', -1, 32)
+		}
 	case reflect.Float64:
-		b = strconv.AppendFloat(b, v.Float(), 'f', -1, 64)
+		f := v.Float()
+		if math.IsNaN(f) {
+			b = append(b, "nan"...)
+		} else if f > math.MaxFloat64 {
+			b = append(b, "inf"...)
+		} else if f < -math.MaxFloat64 {
+			b = append(b, "-inf"...)
+		} else if math.Trunc(f) == f {
+			b = strconv.AppendFloat(b, f, 'f', 1, 64)
+		} else {
+			b = strconv.AppendFloat(b, f, 'f', -1, 64)
+		}
 	case reflect.Bool:
 		if v.Bool() {
 			b = append(b, "true"...)
@@ -254,7 +323,11 @@ func (enc *Encoder) encode(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, e
 			b = append(b, "false"...)
 		}
 	case reflect.Uint64, reflect.Uint32, reflect.Uint16, reflect.Uint8, reflect.Uint:
-		b = strconv.AppendUint(b, v.Uint(), 10)
+		x := v.Uint()
+		if x > uint64(math.MaxInt64) {
+			return nil, fmt.Errorf("toml: not encoding uint (%d) greater than max int64 (%d)", x, int64(math.MaxInt64))
+		}
+		b = strconv.AppendUint(b, x, 10)
 	case reflect.Int64, reflect.Int32, reflect.Int16, reflect.Int8, reflect.Int:
 		b = strconv.AppendInt(b, v.Int(), 10)
 	default:
@@ -273,19 +346,19 @@ func isNil(v reflect.Value) bool {
 	}
 }
 
+func shouldOmitEmpty(options valueOptions, v reflect.Value) bool {
+	return options.omitempty && isEmptyValue(v)
+}
+
 func (enc *Encoder) encodeKv(b []byte, ctx encoderCtx, options valueOptions, v reflect.Value) ([]byte, error) {
 	var err error
 
-	if !ctx.hasKey {
+	if !ctx.inline {
-		panic("caller of encodeKv should have set the key in the context")
+		b = enc.encodeComment(ctx.indent, options.comment, b)
 	}
+
 	b = enc.indent(ctx.indent, b)
-
+	b = enc.encodeKey(b, ctx.key)
-	b, err = enc.encodeKey(b, ctx.key)
-	if err != nil {
-		return nil, err
-	}
-
 	b = append(b, " = "...)
 
 	// create a copy of the context because the value of a KV shouldn't
@@ -303,6 +376,54 @@ func (enc *Encoder) encodeKv(b []byte, ctx encoderCtx, options valueOptions, v r
 	return b, nil
 }
 
+func isEmptyValue(v reflect.Value) bool {
+	switch v.Kind() {
+	case reflect.Struct:
+		return isEmptyStruct(v)
+	case reflect.Array, reflect.Map, reflect.Slice, reflect.String:
+		return v.Len() == 0
+	case reflect.Bool:
+		return !v.Bool()
+	case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
+		return v.Int() == 0
+	case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr:
+		return v.Uint() == 0
+	case reflect.Float32, reflect.Float64:
+		return v.Float() == 0
+	case reflect.Interface, reflect.Ptr:
+		return v.IsNil()
+	}
+	return false
+}
+
+func isEmptyStruct(v reflect.Value) bool {
+	// TODO: merge with walkStruct and cache.
+	typ := v.Type()
+	for i := 0; i < typ.NumField(); i++ {
+		fieldType := typ.Field(i)
+
+		// only consider exported fields
+		if fieldType.PkgPath != "" {
+			continue
+		}
+
+		tag := fieldType.Tag.Get("toml")
+
+		// special field name to skip field
+		if tag == "-" {
+			continue
+		}
+
+		f := v.Field(i)
+
+		if !isEmptyValue(f) {
+			return false
+		}
+	}
+
+	return true
+}
+
 const literalQuote = '\''
 
 func (enc *Encoder) encodeString(b []byte, v string, options valueOptions) []byte {
@@ -314,7 +435,13 @@ func (enc *Encoder) encodeString(b []byte, v string, options valueOptions) []byt
 }
 
 func needsQuoting(v string) bool {
-	return strings.ContainsAny(v, "'\b\f\n\r\t")
+	// TODO: vectorize
+	for _, b := range []byte(v) {
+		if b == '\'' || b == '\r' || b == '\n' || invalidAscii(b) {
+			return true
+		}
+	}
+	return false
 }
 
 // caller should have checked that the string does not contain new lines or ' .
@@ -326,7 +453,6 @@ func (enc *Encoder) encodeLiteralString(b []byte, v string) []byte {
 	return b
 }
 
-//nolint:cyclop
 func (enc *Encoder) encodeQuotedString(multiline bool, b []byte, v string) []byte {
 	stringQuote := `"`
 
@@ -386,7 +512,7 @@ func (enc *Encoder) encodeQuotedString(multiline bool, b []byte, v string) []byt
 	return b
 }
 
-// called should have checked that the string is in A-Z / a-z / 0-9 / - / _ .
+// caller should have checked that the string is in A-Z / a-z / 0-9 / - / _ .
 func (enc *Encoder) encodeUnquotedKey(b []byte, v string) []byte {
 	return append(b, v...)
 }
@@ -396,24 +522,17 @@ func (enc *Encoder) encodeTableHeader(ctx encoderCtx, b []byte) ([]byte, error)
 		return b, nil
 	}
 
+	b = enc.encodeComment(ctx.indent, ctx.options.comment, b)
+
 	b = enc.indent(ctx.indent, b)
 
 	b = append(b, '[')
 
-	var err error
+	b = enc.encodeKey(b, ctx.parentKey[0])
-
-	b, err = enc.encodeKey(b, ctx.parentKey[0])
-	if err != nil {
-		return nil, err
-	}
 
 	for _, k := range ctx.parentKey[1:] {
 		b = append(b, '.')
-
+		b = enc.encodeKey(b, k)
-		b, err = enc.encodeKey(b, k)
-		if err != nil {
-			return nil, err
-		}
 	}
 
 	b = append(b, "]\n"...)
@@ -422,19 +541,19 @@ func (enc *Encoder) encodeTableHeader(ctx encoderCtx, b []byte) ([]byte, error)
 }
 
 //nolint:cyclop
-func (enc *Encoder) encodeKey(b []byte, k string) ([]byte, error) {
+func (enc *Encoder) encodeKey(b []byte, k string) []byte {
 	needsQuotation := false
 	cannotUseLiteral := false
 
+	if len(k) == 0 {
+		return append(b, "''"...)
+	}
+
 	for _, c := range k {
 		if (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') || c == '-' || c == '_' {
 			continue
 		}
 
-		if c == '\n' {
-			return nil, fmt.Errorf("toml: new line characters in keys are not supported")
-		}
-
 		if c == literalQuote {
 			cannotUseLiteral = true
 		}
@@ -442,13 +561,17 @@ func (enc *Encoder) encodeKey(b []byte, k string) ([]byte, error) {
 		needsQuotation = true
 	}
 
+	if needsQuotation && needsQuoting(k) {
+		cannotUseLiteral = true
+	}
+
 	switch {
 	case cannotUseLiteral:
-		return enc.encodeQuotedString(false, b, k), nil
+		return enc.encodeQuotedString(false, b, k)
 	case needsQuotation:
-		return enc.encodeLiteralString(b, k), nil
+		return enc.encodeLiteralString(b, k)
 	default:
-		return enc.encodeUnquotedKey(b, k), nil
+		return enc.encodeUnquotedKey(b, k)
 	}
 }
 
@@ -502,18 +625,26 @@ type table struct {
 }
 
 func (t *table) pushKV(k string, v reflect.Value, options valueOptions) {
+	for _, e := range t.kvs {
+		if e.Key == k {
+			return
+		}
+	}
+
 	t.kvs = append(t.kvs, entry{Key: k, Value: v, Options: options})
 }
 
 func (t *table) pushTable(k string, v reflect.Value, options valueOptions) {
+	for _, e := range t.tables {
+		if e.Key == k {
+			return
+		}
+	}
 	t.tables = append(t.tables, entry{Key: k, Value: v, Options: options})
 }
 
-func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+func walkStruct(ctx encoderCtx, t *table, v reflect.Value) {
-	var t table
+	// TODO: cache this
-
-	//nolint:godox
-	// TODO: cache this?
 	typ := v.Type()
 	for i := 0; i < typ.NumField(); i++ {
 		fieldType := typ.Field(i)
@@ -523,45 +654,130 @@ func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]b
 			continue
 		}
 
-		k, ok := fieldType.Tag.Lookup("toml")
+		tag := fieldType.Tag.Get("toml")
-		if !ok {
-			k = fieldType.Name
-		}
 
 		// special field name to skip field
-		if k == "-" {
+		if tag == "-" {
 			continue
 		}
 
+		k, opts := parseTag(tag)
+		if !isValidName(k) {
+			k = ""
+		}
+
 		f := v.Field(i)
 
+		if k == "" {
+			if fieldType.Anonymous {
+				if fieldType.Type.Kind() == reflect.Struct {
+					walkStruct(ctx, t, f)
+				}
+				continue
+			} else {
+				k = fieldType.Name
+			}
+		}
+
 		if isNil(f) {
 			continue
 		}
 
 		options := valueOptions{
-			multiline: fieldBoolTag(fieldType, "multiline"),
+			multiline: opts.multiline,
+			omitempty: opts.omitempty,
+			comment:   fieldType.Tag.Get("comment"),
 		}
 
-		inline := fieldBoolTag(fieldType, "inline")
+		if opts.inline || !willConvertToTableOrArrayTable(ctx, f) {
-
-		if inline || !willConvertToTableOrArrayTable(ctx, f) {
 			t.pushKV(k, f, options)
 		} else {
 			t.pushTable(k, f, options)
 		}
 	}
+}
+
+func (enc *Encoder) encodeStruct(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
+	var t table
+
+	walkStruct(ctx, &t, v)
 
 	return enc.encodeTable(b, ctx, t)
 }
 
-func fieldBoolTag(field reflect.StructField, tag string) bool {
+func (enc *Encoder) encodeComment(indent int, comment string, b []byte) []byte {
-	x, ok := field.Tag.Lookup(tag)
+	for len(comment) > 0 {
-
+		var line string
-	return ok && x == "true"
+		idx := strings.IndexByte(comment, '\n')
+		if idx >= 0 {
+			line = comment[:idx]
+			comment = comment[idx+1:]
+		} else {
+			line = comment
+			comment = ""
+		}
+		b = enc.indent(indent, b)
+		b = append(b, "# "...)
+		b = append(b, line...)
+		b = append(b, '\n')
+	}
+	return b
+}
+
+func isValidName(s string) bool {
+	if s == "" {
+		return false
+	}
+	for _, c := range s {
+		switch {
+		case strings.ContainsRune("!#$%&()*+-./:;<=>?@[]^_{|}~ ", c):
+			// Backslash and quote chars are reserved, but
+			// otherwise any punctuation chars are allowed
+			// in a tag name.
+		case !unicode.IsLetter(c) && !unicode.IsDigit(c):
+			return false
+		}
+	}
+	return true
+}
+
+type tagOptions struct {
+	multiline bool
+	inline    bool
+	omitempty bool
+}
+
+func parseTag(tag string) (string, tagOptions) {
+	opts := tagOptions{}
+
+	idx := strings.Index(tag, ",")
+	if idx == -1 {
+		return tag, opts
+	}
+
+	raw := tag[idx+1:]
+	tag = string(tag[:idx])
+	for raw != "" {
+		var o string
+		i := strings.Index(raw, ",")
+		if i >= 0 {
+			o, raw = raw[:i], raw[i+1:]
+		} else {
+			o, raw = raw, ""
+		}
+		switch o {
+		case "multiline":
+			opts.multiline = true
+		case "inline":
+			opts.inline = true
+		case "omitempty":
+			opts.omitempty = true
+		}
+	}
+
+	return tag, opts
 }
 
-//nolint:cyclop
 func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, error) {
 	var err error
 
@@ -583,7 +799,13 @@ func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, erro
 	}
 	ctx.skipTableHeader = false
 
+	hasNonEmptyKV := false
 	for _, kv := range t.kvs {
+		if shouldOmitEmpty(kv.Options, kv.Value) {
+			continue
+		}
+		hasNonEmptyKV = true
+
 		ctx.setKey(kv.Key)
 
 		b, err = enc.encodeKv(b, ctx, kv.Options, kv.Value)
@@ -594,7 +816,20 @@ func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, erro
 		b = append(b, '\n')
 	}
 
+	first := true
 	for _, table := range t.tables {
+		if shouldOmitEmpty(table.Options, table.Value) {
+			continue
+		}
+		if first {
+			first = false
+			if hasNonEmptyKV {
+				b = append(b, '\n')
+			}
+		} else {
+			b = append(b, "\n"...)
+		}
+
 		ctx.setKey(table.Key)
 
 		ctx.options = table.Options
@@ -603,8 +838,6 @@ func (enc *Encoder) encodeTable(b []byte, ctx encoderCtx, t table) ([]byte, erro
 		if err != nil {
 			return nil, err
 		}
-
-		b = append(b, '\n')
 	}
 
 	return b, nil
@@ -617,6 +850,10 @@ func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte
 
 	first := true
 	for _, kv := range t.kvs {
+		if shouldOmitEmpty(kv.Options, kv.Value) {
+			continue
+		}
+
 		if first {
 			first = false
 		} else {
@@ -632,7 +869,7 @@ func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte
 	}
 
 	if len(t.tables) > 0 {
-		panic("inline table cannot contain nested tables, online key-values")
+		panic("inline table cannot contain nested tables, only key-values")
 	}
 
 	b = append(b, "}"...)
@@ -640,10 +877,11 @@ func (enc *Encoder) encodeTableInline(b []byte, ctx encoderCtx, t table) ([]byte
 	return b, nil
 }
 
-var textMarshalerType = reflect.TypeOf(new(encoding.TextMarshaler)).Elem()
-
 func willConvertToTable(ctx encoderCtx, v reflect.Value) bool {
-	if v.Type() == timeType || v.Type().Implements(textMarshalerType) {
+	if !v.IsValid() {
+		return false
+	}
+	if v.Type() == timeType || v.Type().Implements(textMarshalerType) || (v.Kind() != reflect.Ptr && v.CanAddr() && reflect.PtrTo(v.Type()).Implements(textMarshalerType)) {
 		return false
 	}
 
@@ -665,6 +903,9 @@ func willConvertToTable(ctx encoderCtx, v reflect.Value) bool {
 }
 
 func willConvertToTableOrArrayTable(ctx encoderCtx, v reflect.Value) bool {
+	if ctx.insideKv {
+		return false
+	}
 	t := v.Type()
 
 	if t.Kind() == reflect.Interface {
@@ -710,7 +951,6 @@ func (enc *Encoder) encodeSlice(b []byte, ctx encoderCtx, v reflect.Value) ([]by
 func (enc *Encoder) encodeSliceAsArrayTable(b []byte, ctx encoderCtx, v reflect.Value) ([]byte, error) {
 	ctx.shiftKey()
 
-	var err error
 	scratch := make([]byte, 0, 64)
 	scratch = append(scratch, "[["...)
 
@@ -719,18 +959,22 @@ func (enc *Encoder) encodeSliceAsArrayTable(b []byte, ctx encoderCtx, v reflect.
 			scratch = append(scratch, '.')
 		}
 
-		scratch, err = enc.encodeKey(scratch, k)
+		scratch = enc.encodeKey(scratch, k)
-		if err != nil {
-			return nil, err
-		}
 	}
 
 	scratch = append(scratch, "]]\n"...)
 	ctx.skipTableHeader = true
 
+	b = enc.encodeComment(ctx.indent, ctx.options.comment, b)
+
 	for i := 0; i < v.Len(); i++ {
+		if i != 0 {
+			b = append(b, "\n"...)
+		}
+
 		b = append(b, scratch...)
 
+		var err error
 		b, err = enc.encode(b, ctx, v.Index(i))
 		if err != nil {
 			return nil, err

parser.go

@@ -2,9 +2,10 @@ package toml
 
 import (
 	"bytes"
-	"strconv"
+	"unicode"
 
 	"github.com/pelletier/go-toml/v2/internal/ast"
+	"github.com/pelletier/go-toml/v2/internal/danger"
 )
 
 type parser struct {
@@ -16,9 +17,25 @@ type parser struct {
 	first   bool
 }
 
+func (p *parser) Range(b []byte) ast.Range {
+	return ast.Range{
+		Offset: uint32(danger.SubsliceOffset(p.data, b)),
+		Length: uint32(len(b)),
+	}
+}
+
+func (p *parser) Raw(raw ast.Range) []byte {
+	return p.data[raw.Offset : raw.Offset+raw.Length]
+}
+
+func (p *parser) SetRaw(ref ast.Reference, from []byte, to []byte) {
+	b := danger.BytesRange(from, to)
+	p.builder.NodeAt(ref).Raw = p.Range(b)
+}
+
 func (p *parser) Reset(b []byte) {
 	p.builder.Reset()
-	p.ref = ast.Reference{}
+	p.ref = ast.InvalidReference
 	p.data = b
 	p.left = b
 	p.err = nil
@@ -32,7 +49,7 @@ func (p *parser) NextExpression() bool {
 	}
 
 	p.builder.Reset()
-	p.ref = ast.Reference{}
+	p.ref = ast.InvalidReference
 
 	for {
 		if len(p.left) == 0 || p.err != nil {
@@ -61,7 +78,7 @@ func (p *parser) NextExpression() bool {
 	}
 }
 
-func (p *parser) Expression() ast.Node {
+func (p *parser) Expression() *ast.Node {
 	return p.builder.NodeAt(p.ref)
 }
 
@@ -86,7 +103,7 @@ func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
 	// expression =  ws [ comment ]
 	// expression =/ ws keyval ws [ comment ]
 	// expression =/ ws table ws [ comment ]
-	var ref ast.Reference
+	ref := ast.InvalidReference
 
 	b = p.parseWhitespace(b)
 
@@ -95,9 +112,8 @@ func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
 	}
 
 	if b[0] == '#' {
-		_, rest := scanComment(b)
+		_, rest, err := scanComment(b)
-
+		return ref, rest, err
-		return ref, rest, nil
 	}
 
 	if b[0] == '\n' || b[0] == '\r' {
@@ -118,9 +134,8 @@ func (p *parser) parseExpression(b []byte) (ast.Reference, []byte, error) {
 	b = p.parseWhitespace(b)
 
 	if len(b) > 0 && b[0] == '#' {
-		_, rest := scanComment(b)
+		_, rest, err := scanComment(b)
-
+		return ref, rest, err
-		return ref, rest, nil
 	}
 
 	return ref, b, nil
@@ -142,12 +157,14 @@ func (p *parser) parseArrayTable(b []byte) (ast.Reference, []byte, error) {
 	ref := p.builder.Push(ast.Node{
 		Kind: ast.ArrayTable,
 	})
+	start := b
 
 	b = b[2:]
 	b = p.parseWhitespace(b)
 
 	k, b, err := p.parseKey(b)
 	if err != nil {
+		p.SetRaw(ref, start, b)
 		return ref, nil, err
 	}
 
@@ -156,11 +173,12 @@ func (p *parser) parseArrayTable(b []byte) (ast.Reference, []byte, error) {
 
 	b, err = expect(']', b)
 	if err != nil {
+		p.SetRaw(ref, start, b)
 		return ref, nil, err
 	}
 
 	b, err = expect(']', b)
-
+	p.SetRaw(ref, start, b)
 	return ref, b, err
 }
 
@@ -171,12 +189,14 @@ func (p *parser) parseStdTable(b []byte) (ast.Reference, []byte, error) {
 	ref := p.builder.Push(ast.Node{
 		Kind: ast.Table,
 	})
+	start := b
 
 	b = b[1:]
 	b = p.parseWhitespace(b)
 
 	key, b, err := p.parseKey(b)
 	if err != nil {
+		p.SetRaw(ref, start, b)
 		return ref, nil, err
 	}
 
@@ -186,6 +206,7 @@ func (p *parser) parseStdTable(b []byte) (ast.Reference, []byte, error) {
 
 	b, err = expect(']', b)
 
+	p.SetRaw(ref, start, b)
 	return ref, b, err
 }
 
@@ -194,10 +215,12 @@ func (p *parser) parseKeyval(b []byte) (ast.Reference, []byte, error) {
 	ref := p.builder.Push(ast.Node{
 		Kind: ast.KeyValue,
 	})
+	start := b
 
 	key, b, err := p.parseKey(b)
 	if err != nil {
-		return ast.Reference{}, nil, err
+		p.SetRaw(ref, start, b)
+		return ast.InvalidReference, b, err
 	}
 
 	// keyval-sep = ws %x3D ws ; =
@@ -205,34 +228,38 @@ func (p *parser) parseKeyval(b []byte) (ast.Reference, []byte, error) {
 	b = p.parseWhitespace(b)
 
 	if len(b) == 0 {
-		return ast.Reference{}, nil, newDecodeError(b, "expected = after a key, but the document ends there")
+		p.SetRaw(ref, start, b)
+		return ast.InvalidReference, b, newDecodeError(b, "expected = after a key, but the document ends there")
 	}
 
 	b, err = expect('=', b)
 	if err != nil {
-		return ast.Reference{}, nil, err
+		p.SetRaw(ref, start, b)
+		return ast.InvalidReference, b, err
 	}
 
 	b = p.parseWhitespace(b)
 
 	valRef, b, err := p.parseVal(b)
 	if err != nil {
+		p.SetRaw(ref, start, b)
 		return ref, b, err
 	}
 
 	p.builder.Chain(valRef, key)
 	p.builder.AttachChild(ref, valRef)
 
+	p.SetRaw(ref, start, b)
 	return ref, b, err
 }
 
 //nolint:cyclop,funlen
 func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 	// val = string / boolean / array / inline-table / date-time / float / integer
-	var ref ast.Reference
+	ref := ast.InvalidReference
 
 	if len(b) == 0 {
-		return ref, nil, newDecodeError(b, "expected value, not eof")
+		return ref, b, newDecodeError(b, "expected value, not eof")
 	}
 
 	var err error
@@ -240,32 +267,36 @@ func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 
 	switch c {
 	case '"':
+		var raw []byte
 		var v []byte
 		if scanFollowsMultilineBasicStringDelimiter(b) {
-			v, b, err = p.parseMultilineBasicString(b)
+			raw, v, b, err = p.parseMultilineBasicString(b)
 		} else {
-			v, b, err = p.parseBasicString(b)
+			raw, v, b, err = p.parseBasicString(b)
 		}
 
 		if err == nil {
 			ref = p.builder.Push(ast.Node{
 				Kind: ast.String,
+				Raw:  p.Range(raw),
 				Data: v,
 			})
 		}
 
 		return ref, b, err
 	case '\'':
+		var raw []byte
 		var v []byte
 		if scanFollowsMultilineLiteralStringDelimiter(b) {
-			v, b, err = p.parseMultilineLiteralString(b)
+			raw, v, b, err = p.parseMultilineLiteralString(b)
 		} else {
-			v, b, err = p.parseLiteralString(b)
+			raw, v, b, err = p.parseLiteralString(b)
 		}
 
 		if err == nil {
 			ref = p.builder.Push(ast.Node{
 				Kind: ast.String,
+				Raw:  p.Range(raw),
 				Data: v,
 			})
 		}
@@ -273,23 +304,25 @@ func (p *parser) parseVal(b []byte) (ast.Reference, []byte, error) {
 		return ref, b, err
 	case 't':
 		if !scanFollowsTrue(b) {
-			return ref, nil, newDecodeError(atmost(b, 4), "expected 'true'")
+			return ref, b, newDecodeError(atmost(b, 4), "expected 'true'")
 		}
 
 		ref = p.builder.Push(ast.Node{
 			Kind: ast.Bool,
 			Data: b[:4],
+			Raw:  p.Range(b[:4]),
 		})
 
 		return ref, b[4:], nil
 	case 'f':
 		if !scanFollowsFalse(b) {
-			return ref, nil, newDecodeError(atmost(b, 5), "expected 'false'")
+			return ref, b, newDecodeError(atmost(b, 5), "expected 'false'")
 		}
 
 		ref = p.builder.Push(ast.Node{
 			Kind: ast.Bool,
 			Data: b[:5],
+			Raw:  p.Range(b[:5]),
 		})
 
 		return ref, b[5:], nil
@@ -310,13 +343,13 @@ func atmost(b []byte, n int) []byte {
 	return b[:n]
 }
 
-func (p *parser) parseLiteralString(b []byte) ([]byte, []byte, error) {
+func (p *parser) parseLiteralString(b []byte) ([]byte, []byte, []byte, error) {
 	v, rest, err := scanLiteralString(b)
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, rest, err
 	}
 
-	return v[1 : len(v)-1], rest, nil
+	return v, v[1 : len(v)-1], rest, nil
 }
 
 func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
@@ -328,6 +361,7 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 	parent := p.builder.Push(ast.Node{
 		Kind: ast.InlineTable,
 	})
+	start := b
 
 	first := true
 
@@ -338,7 +372,14 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 	var err error
 
 	for len(b) > 0 {
+		previousB := b
 		b = p.parseWhitespace(b)
+
+		if len(b) == 0 {
+			p.SetRaw(parent, start, b)
+			return parent, b, newDecodeError(previousB[:1], "inline table is incomplete")
+		}
+
 		if b[0] == '}' {
 			break
 		}
@@ -346,7 +387,8 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 		if !first {
 			b, err = expect(',', b)
 			if err != nil {
-				return parent, nil, err
+				p.SetRaw(parent, start, b)
+				return parent, b, err
 			}
 			b = p.parseWhitespace(b)
 		}
@@ -355,7 +397,8 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 
 		kv, b, err = p.parseKeyval(b)
 		if err != nil {
-			return parent, nil, err
+			p.SetRaw(parent, start, b)
+			return parent, b, err
 		}
 
 		if first {
@@ -370,6 +413,7 @@ func (p *parser) parseInlineTable(b []byte) (ast.Reference, []byte, error) {
 
 	rest, err := expect('}', b)
 
+	p.SetRaw(parent, start, b)
 	return parent, rest, err
 }
 
@@ -382,6 +426,8 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 	// array-values =/ ws-comment-newline val ws-comment-newline [ array-sep ]
 	// array-sep = %x2C  ; , Comma
 	// ws-comment-newline = *( wschar / [ comment ] newline )
+	arrayStart := b
+	start := b
 	b = b[1:]
 
 	parent := p.builder.Push(ast.Node{
@@ -396,11 +442,13 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 	for len(b) > 0 {
 		b, err = p.parseOptionalWhitespaceCommentNewline(b)
 		if err != nil {
-			return parent, nil, err
+			p.SetRaw(parent, start, b)
+			return parent, b, err
 		}
 
 		if len(b) == 0 {
-			return parent, nil, newDecodeError(b, "array is incomplete")
+			p.SetRaw(parent, start, b)
+			return parent, b, newDecodeError(arrayStart[:1], "array is incomplete")
 		}
 
 		if b[0] == ']' {
@@ -409,14 +457,19 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 
 		if b[0] == ',' {
 			if first {
-				return parent, nil, newDecodeError(b[0:1], "array cannot start with comma")
+				p.SetRaw(parent, start, b)
+				return parent, b, newDecodeError(b[0:1], "array cannot start with comma")
 			}
 			b = b[1:]
 
 			b, err = p.parseOptionalWhitespaceCommentNewline(b)
 			if err != nil {
-				return parent, nil, err
+				p.SetRaw(parent, start, b)
+				return parent, b, err
 			}
+		} else if !first {
+			p.SetRaw(parent, start, b)
+			return parent, b, newDecodeError(b[0:1], "array elements must be separated by commas")
 		}
 
 		// TOML allows trailing commas in arrays.
@@ -425,10 +478,10 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 		}
 
 		var valueRef ast.Reference
-
 		valueRef, b, err = p.parseVal(b)
 		if err != nil {
-			return parent, nil, err
+			p.SetRaw(parent, start, b)
+			return parent, b, err
 		}
 
 		if first {
@@ -440,13 +493,16 @@ func (p *parser) parseValArray(b []byte) (ast.Reference, []byte, error) {
 
 		b, err = p.parseOptionalWhitespaceCommentNewline(b)
 		if err != nil {
-			return parent, nil, err
+			p.SetRaw(parent, start, b)
+			return parent, b, err
 		}
 		first = false
 	}
 
 	rest, err := expect(']', b)
 
+	p.SetRaw(parent, start, rest)
+
 	return parent, rest, err
 }
 
@@ -456,7 +512,10 @@ func (p *parser) parseOptionalWhitespaceCommentNewline(b []byte) ([]byte, error)
 		b = p.parseWhitespace(b)
 
 		if len(b) > 0 && b[0] == '#' {
-			_, b = scanComment(b)
+			_, b, err = scanComment(b)
+			if err != nil {
+				return b, err
+			}
 		}
 
 		if len(b) == 0 {
@@ -466,7 +525,7 @@ func (p *parser) parseOptionalWhitespaceCommentNewline(b []byte) ([]byte, error)
 		if b[0] == '\n' || b[0] == '\r' {
 			b, err = p.parseNewline(b)
 			if err != nil {
-				return nil, err
+				return b, err
 			}
 		} else {
 			break
@@ -476,10 +535,10 @@ func (p *parser) parseOptionalWhitespaceCommentNewline(b []byte) ([]byte, error)
 	return b, nil
 }
 
-func (p *parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, error) {
+func (p *parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, []byte, error) {
 	token, rest, err := scanMultilineLiteralString(b)
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, rest, err
 	}
 
 	i := 3
@@ -491,11 +550,11 @@ func (p *parser) parseMultilineLiteralString(b []byte) ([]byte, []byte, error) {
 		i += 2
 	}
 
-	return token[i : len(token)-3], rest, err
+	return token, token[i : len(token)-3], rest, err
 }
 
 //nolint:funlen,gocognit,cyclop
-func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, error) {
+func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, []byte, error) {
 	// ml-basic-string = ml-basic-string-delim [ newline ] ml-basic-body
 	// ml-basic-string-delim
 	// ml-basic-string-delim = 3quotation-mark
@@ -506,13 +565,11 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, error) {
 	// mlb-quotes = 1*2quotation-mark
 	// mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
 	// mlb-escaped-nl = escape ws newline *( wschar / newline )
-	token, rest, err := scanMultilineBasicString(b)
+	token, escaped, rest, err := scanMultilineBasicString(b)
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, rest, err
 	}
 
-	var builder bytes.Buffer
-
 	i := 3
 
 	// skip the immediate new line
@@ -522,9 +579,24 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, error) {
 		i += 2
 	}
 
+	// fast path
+	startIdx := i
+	endIdx := len(token) - len(`"""`)
+
+	if !escaped {
+		str := token[startIdx:endIdx]
+		verr := utf8TomlValidAlreadyEscaped(str)
+		if verr.Zero() {
+			return token, str, rest, nil
+		}
+		return nil, nil, rest, newDecodeError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
+	}
+
+	var builder bytes.Buffer
+
 	// The scanner ensures that the token starts and ends with quotes and that
 	// escapes are balanced.
-	for ; i < len(token)-3; i++ {
+	for i < len(token)-3 {
 		c := token[i]
 
 		//nolint:nestif
@@ -532,17 +604,33 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, error) {
 			// When the last non-whitespace character on a line is an unescaped \,
 			// it will be trimmed along with all whitespace (including newlines) up
 			// to the next non-whitespace character or closing delimiter.
-			if token[i+1] == '\n' || (token[i+1] == '\r' && token[i+2] == '\n') {
+
-				i++ // skip the \
+			isLastNonWhitespaceOnLine := false
+			j := 1
+		findEOLLoop:
+			for ; j < len(token)-3-i; j++ {
+				switch token[i+j] {
+				case ' ', '\t':
+					continue
+				case '\r':
+					if token[i+j+1] == '\n' {
+						continue
+					}
+				case '\n':
+					isLastNonWhitespaceOnLine = true
+				}
+				break findEOLLoop
+			}
+			if isLastNonWhitespaceOnLine {
+				i += j
 				for ; i < len(token)-3; i++ {
 					c := token[i]
 					if !(c == '\n' || c == '\r' || c == ' ' || c == '\t') {
 						i--
-
 						break
 					}
 				}
-
+				i++
 				continue
 			}
 
@@ -563,31 +651,38 @@ func (p *parser) parseMultilineBasicString(b []byte) ([]byte, []byte, error) {
 				builder.WriteByte('\r')
 			case 't':
 				builder.WriteByte('\t')
+			case 'e':
+				builder.WriteByte(0x1B)
 			case 'u':
-				x, err := hexToString(atmost(token[i+1:], 4), 4)
+				x, err := hexToRune(atmost(token[i+1:], 4), 4)
 				if err != nil {
-					return nil, nil, err
+					return nil, nil, rest, err
 				}
-
+				builder.WriteRune(x)
-				builder.WriteString(x)
 				i += 4
 			case 'U':
-				x, err := hexToString(atmost(token[i+1:], 8), 8)
+				x, err := hexToRune(atmost(token[i+1:], 8), 8)
 				if err != nil {
-					return nil, nil, err
+					return nil, nil, rest, err
 				}
 
-				builder.WriteString(x)
+				builder.WriteRune(x)
 				i += 8
 			default:
-				return nil, nil, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
+				return nil, nil, rest, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
 			}
+			i++
 		} else {
-			builder.WriteByte(c)
+			size := utf8ValidNext(token[i:])
+			if size == 0 {
+				return nil, nil, rest, newDecodeError(token[i:i+1], "invalid character %#U", c)
+			}
+			builder.Write(token[i : i+size])
+			i += size
 		}
 	}
 
-	return builder.Bytes(), rest, nil
+	return token, builder.Bytes(), rest, nil
 }
 
 func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
@@ -599,13 +694,14 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 	// dotted-key = simple-key 1*( dot-sep simple-key )
 	//
 	// dot-sep   = ws %x2E ws  ; . Period
-	key, b, err := p.parseSimpleKey(b)
+	raw, key, b, err := p.parseSimpleKey(b)
 	if err != nil {
-		return ast.Reference{}, nil, err
+		return ast.InvalidReference, b, err
 	}
 
 	ref := p.builder.Push(ast.Node{
 		Kind: ast.Key,
+		Raw:  p.Range(raw),
 		Data: key,
 	})
 
@@ -614,13 +710,14 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 		if len(b) > 0 && b[0] == '.' {
 			b = p.parseWhitespace(b[1:])
 
-			key, b, err = p.parseSimpleKey(b)
+			raw, key, b, err = p.parseSimpleKey(b)
 			if err != nil {
-				return ref, nil, err
+				return ref, b, err
 			}
 
 			p.builder.PushAndChain(ast.Node{
 				Kind: ast.Key,
+				Raw:  p.Range(raw),
 				Data: key,
 			})
 		} else {
@@ -631,14 +728,14 @@ func (p *parser) parseKey(b []byte) (ast.Reference, []byte, error) {
 	return ref, b, nil
 }
 
-func (p *parser) parseSimpleKey(b []byte) (key, rest []byte, err error) {
+func (p *parser) parseSimpleKey(b []byte) (raw, key, rest []byte, err error) {
+	if len(b) == 0 {
+		return nil, nil, b, newDecodeError(b, "expected key but found none")
+	}
+
 	// 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
-	if len(b) == 0 {
-		return nil, nil, newDecodeError(b, "key is incomplete")
-	}
-
 	switch {
 	case b[0] == '\'':
 		return p.parseLiteralString(b)
@@ -646,14 +743,14 @@ func (p *parser) parseSimpleKey(b []byte) (key, rest []byte, err error) {
 		return p.parseBasicString(b)
 	case isUnquotedKeyChar(b[0]):
 		key, rest = scanUnquotedKey(b)
-		return key, rest, nil
+		return key, key, rest, nil
 	default:
-		return nil, nil, newDecodeError(b[0:1], "invalid character at start of key: %c", b[0])
+		return nil, nil, b[1:], newDecodeError(b[0:1], "invalid character at start of key: %c", b[0])
 	}
 }
 
 //nolint:funlen,cyclop
-func (p *parser) parseBasicString(b []byte) ([]byte, []byte, error) {
+func (p *parser) parseBasicString(b []byte) ([]byte, []byte, []byte, error) {
 	// basic-string = quotation-mark *basic-char quotation-mark
 	// quotation-mark = %x22            ; "
 	// basic-char = basic-unescaped / escaped
@@ -668,16 +765,33 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, error) {
 	// escape-seq-char =/ %x74         ; t    tab             U+0009
 	// escape-seq-char =/ %x75 4HEXDIG ; uXXXX                U+XXXX
 	// escape-seq-char =/ %x55 8HEXDIG ; UXXXXXXXX            U+XXXXXXXX
-	token, rest, err := scanBasicString(b)
+	token, escaped, rest, err := scanBasicString(b)
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, rest, err
 	}
 
+	startIdx := len(`"`)
+	endIdx := len(token) - len(`"`)
+
+	// Fast path. If there is no escape sequence, the string should just be
+	// an UTF-8 encoded string, which is the same as Go. In that case,
+	// validate the string and return a direct reference to the buffer.
+	if !escaped {
+		str := token[startIdx:endIdx]
+		verr := utf8TomlValidAlreadyEscaped(str)
+		if verr.Zero() {
+			return token, str, rest, nil
+		}
+		return nil, nil, rest, newDecodeError(str[verr.Index:verr.Index+verr.Size], "invalid UTF-8")
+	}
+
+	i := startIdx
+
 	var builder bytes.Buffer
 
 	// The scanner ensures that the token starts and ends with quotes and that
 	// escapes are balanced.
-	for i := 1; i < len(token)-1; i++ {
+	for i < len(token)-1 {
 		c := token[i]
 		if c == '\\' {
 			i++
@@ -696,47 +810,68 @@ func (p *parser) parseBasicString(b []byte) ([]byte, []byte, error) {
 				builder.WriteByte('\r')
 			case 't':
 				builder.WriteByte('\t')
+			case 'e':
+				builder.WriteByte(0x1B)
 			case 'u':
-				x, err := hexToString(token[i+1:len(token)-1], 4)
+				x, err := hexToRune(token[i+1:len(token)-1], 4)
 				if err != nil {
-					return nil, nil, err
+					return nil, nil, rest, err
 				}
 
-				builder.WriteString(x)
+				builder.WriteRune(x)
 				i += 4
 			case 'U':
-				x, err := hexToString(token[i+1:len(token)-1], 8)
+				x, err := hexToRune(token[i+1:len(token)-1], 8)
 				if err != nil {
-					return nil, nil, err
+					return nil, nil, rest, err
 				}
 
-				builder.WriteString(x)
+				builder.WriteRune(x)
 				i += 8
 			default:
-				return nil, nil, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
+				return nil, nil, rest, newDecodeError(token[i:i+1], "invalid escaped character %#U", c)
 			}
+			i++
 		} else {
-			builder.WriteByte(c)
+			size := utf8ValidNext(token[i:])
+			if size == 0 {
+				return nil, nil, rest, newDecodeError(token[i:i+1], "invalid character %#U", c)
+			}
+			builder.Write(token[i : i+size])
+			i += size
 		}
 	}
 
-	return builder.Bytes(), rest, nil
+	return token, builder.Bytes(), rest, nil
 }
 
-func hexToString(b []byte, length int) (string, error) {
+func hexToRune(b []byte, length int) (rune, error) {
 	if len(b) < length {
-		return "", newDecodeError(b, "unicode point needs %d character, not %d", length, len(b))
+		return -1, newDecodeError(b, "unicode point needs %d character, not %d", length, len(b))
 	}
 	b = b[:length]
 
-	//nolint:godox
+	var r uint32
-	// TODO: slow
+	for i, c := range b {
-	intcode, err := strconv.ParseInt(string(b), 16, 32)
+		d := uint32(0)
-	if err != nil {
+		switch {
-		return "", newDecodeError(b, "couldn't parse hexadecimal number: %w", err)
+		case '0' <= c && c <= '9':
+			d = uint32(c - '0')
+		case 'a' <= c && c <= 'f':
+			d = uint32(c - 'a' + 10)
+		case 'A' <= c && c <= 'F':
+			d = uint32(c - 'A' + 10)
+		default:
+			return -1, newDecodeError(b[i:i+1], "non-hex character")
+		}
+		r = r*16 + d
 	}
 
-	return string(rune(intcode)), nil
+	if r > unicode.MaxRune || 0xD800 <= r && r < 0xE000 {
+		return -1, newDecodeError(b, "escape sequence is invalid Unicode code point")
+	}
+
+	return rune(r), nil
 }
 
 func (p *parser) parseWhitespace(b []byte) []byte {
@@ -753,27 +888,28 @@ func (p *parser) parseIntOrFloatOrDateTime(b []byte) (ast.Reference, []byte, err
 	switch b[0] {
 	case 'i':
 		if !scanFollowsInf(b) {
-			return ast.Reference{}, nil, newDecodeError(atmost(b, 3), "expected 'inf'")
+			return ast.InvalidReference, b, newDecodeError(atmost(b, 3), "expected 'inf'")
 		}
 
 		return p.builder.Push(ast.Node{
 			Kind: ast.Float,
 			Data: b[:3],
+			Raw:  p.Range(b[:3]),
 		}), b[3:], nil
 	case 'n':
 		if !scanFollowsNan(b) {
-			return ast.Reference{}, nil, newDecodeError(atmost(b, 3), "expected 'nan'")
+			return ast.InvalidReference, b, newDecodeError(atmost(b, 3), "expected 'nan'")
 		}
 
 		return p.builder.Push(ast.Node{
 			Kind: ast.Float,
 			Data: b[:3],
+			Raw:  p.Range(b[:3]),
 		}), b[3:], nil
 	case '+', '-':
 		return p.scanIntOrFloat(b)
 	}
 
-	//nolint:gomnd
 	if len(b) < 3 {
 		return p.scanIntOrFloat(b)
 	}
@@ -791,26 +927,17 @@ func (p *parser) parseIntOrFloatOrDateTime(b []byte) (ast.Reference, []byte, err
 		if idx == 2 && c == ':' || (idx == 4 && c == '-') {
 			return p.scanDateTime(b)
 		}
+
+		break
 	}
 
 	return p.scanIntOrFloat(b)
 }
 
-func digitsToInt(b []byte) int {
-	x := 0
-
-	for _, d := range b {
-		x *= 10
-		x += int(d - '0')
-	}
-
-	return x
-}
-
-//nolint:gocognit,cyclop
 func (p *parser) scanDateTime(b []byte) (ast.Reference, []byte, error) {
 	// scans for contiguous characters in [0-9T:Z.+-], and up to one space if
 	// followed by a digit.
+	hasDate := false
 	hasTime := false
 	hasTz := false
 	seenSpace := false
@@ -823,17 +950,23 @@ byteLoop:
 		switch {
 		case isDigit(c):
 		case c == '-':
-			const offsetOfTz = 19
+			hasDate = true
-			if i == offsetOfTz {
+			const minOffsetOfTz = 8
+			if i >= minOffsetOfTz {
 				hasTz = true
 			}
-		case c == 'T' || c == ':' || c == '.':
+		case c == 'T' || c == 't' || c == ':' || c == '.':
 			hasTime = true
-		case c == '+' || c == '-' || c == 'Z':
+		case c == '+' || c == '-' || c == 'Z' || c == 'z':
 			hasTz = true
 		case c == ' ':
 			if !seenSpace && i+1 < len(b) && isDigit(b[i+1]) {
 				i += 2
+				// Avoid reaching past the end of the document in case the time
+				// is malformed. See TestIssue585.
+				if i >= len(b) {
+					i--
+				}
 				seenSpace = true
 				hasTime = true
 			} else {
@@ -847,10 +980,14 @@ byteLoop:
 	var kind ast.Kind
 
 	if hasTime {
-		if hasTz {
+		if hasDate {
-			kind = ast.DateTime
+			if hasTz {
+				kind = ast.DateTime
+			} else {
+				kind = ast.LocalDateTime
+			}
 		} else {
-			kind = ast.LocalDateTime
+			kind = ast.LocalTime
 		}
 	} else {
 		kind = ast.LocalDate
@@ -859,6 +996,7 @@ byteLoop:
 	return p.builder.Push(ast.Node{
 		Kind: kind,
 		Data: b[:i],
+		Raw:  p.Range(b[:i]),
 	}), b[i:], nil
 }
 
@@ -866,7 +1004,7 @@ byteLoop:
 func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 	i := 0
 
-	if len(b) > 2 && b[0] == '0' && b[1] != '.' {
+	if len(b) > 2 && b[0] == '0' && b[1] != '.' && b[1] != 'e' && b[1] != 'E' {
 		var isValidRune validRuneFn
 
 		switch b[1] {
@@ -892,6 +1030,7 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 		return p.builder.Push(ast.Node{
 			Kind: ast.Integer,
 			Data: b[:i],
+			Raw:  p.Range(b[:i]),
 		}), b[i:], nil
 	}
 
@@ -915,10 +1054,11 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 				return p.builder.Push(ast.Node{
 					Kind: ast.Float,
 					Data: b[:i+3],
+					Raw:  p.Range(b[:i+3]),
 				}), b[i+3:], nil
 			}
 
-			return ast.Reference{}, nil, newDecodeError(b[i:i+1], "unexpected character 'i' while scanning for a number")
+			return ast.InvalidReference, b[i:], newDecodeError(b[i:i+1], "unexpected character 'i' while scanning for a number")
 		}
 
 		if c == 'n' {
@@ -926,17 +1066,18 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 				return p.builder.Push(ast.Node{
 					Kind: ast.Float,
 					Data: b[:i+3],
+					Raw:  p.Range(b[:i+3]),
 				}), b[i+3:], nil
 			}
 
-			return ast.Reference{}, nil, newDecodeError(b[i:i+1], "unexpected character 'n' while scanning for a number")
+			return ast.InvalidReference, b[i:], newDecodeError(b[i:i+1], "unexpected character 'n' while scanning for a number")
 		}
 
 		break
 	}
 
 	if i == 0 {
-		return ast.Reference{}, b, newDecodeError(b, "incomplete number")
+		return ast.InvalidReference, b, newDecodeError(b, "incomplete number")
 	}
 
 	kind := ast.Integer
@@ -948,6 +1089,7 @@ func (p *parser) scanIntOrFloat(b []byte) (ast.Reference, []byte, error) {
 	return p.builder.Push(ast.Node{
 		Kind: kind,
 		Data: b[:i],
+		Raw:  p.Range(b[:i]),
 	}), b[i:], nil
 }
 
@@ -973,8 +1115,12 @@ func isValidBinaryRune(r byte) bool {
 }
 
 func expect(x byte, b []byte) ([]byte, error) {
+	if len(b) == 0 {
+		return b, newDecodeError(b, "expected character %c but the document ended here", x)
+	}
+
 	if b[0] != x {
-		return nil, newDecodeError(b[0:1], "expected character %U", x)
+		return b, newDecodeError(b[0:1], "expected character %c", x)
 	}
 
 	return b[1:], nil

parser_test.go

@@ -1,6 +1,8 @@
 package toml
 
 import (
+	"strconv"
+	"strings"
 	"testing"
 
 	"github.com/pelletier/go-toml/v2/internal/ast"
@@ -9,8 +11,6 @@ import (
 
 //nolint:funlen
 func TestParser_AST_Numbers(t *testing.T) {
-	t.Parallel()
-
 	examples := []struct {
 		desc  string
 		input string
@@ -137,7 +137,6 @@ func TestParser_AST_Numbers(t *testing.T) {
 	for _, e := range examples {
 		e := e
 		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
 			p := parser{}
 			p.Reset([]byte(`A = ` + e.input))
 			p.NextExpression()
@@ -168,7 +167,7 @@ type (
 	}
 )
 
-func compareNode(t *testing.T, e astNode, n ast.Node) {
+func compareNode(t *testing.T, e astNode, n *ast.Node) {
 	t.Helper()
 	require.Equal(t, e.Kind, n.Kind)
 	require.Equal(t, e.Data, n.Data)
@@ -200,8 +199,6 @@ func compareIterator(t *testing.T, expected []astNode, actual ast.Iterator) {
 
 //nolint:funlen
 func TestParser_AST(t *testing.T) {
-	t.Parallel()
-
 	examples := []struct {
 		desc  string
 		input string
@@ -340,7 +337,6 @@ func TestParser_AST(t *testing.T) {
 	for _, e := range examples {
 		e := e
 		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
 			p := parser{}
 			p.Reset([]byte(e.input))
 			p.NextExpression()
@@ -354,3 +350,101 @@ func TestParser_AST(t *testing.T) {
 		})
 	}
 }
+
+func BenchmarkParseBasicStringWithUnicode(b *testing.B) {
+	p := &parser{}
+	b.Run("4", func(b *testing.B) {
+		input := []byte(`"\u1234\u5678\u9ABC\u1234\u5678\u9ABC"`)
+		b.ReportAllocs()
+		b.SetBytes(int64(len(input)))
+
+		for i := 0; i < b.N; i++ {
+			p.parseBasicString(input)
+		}
+	})
+	b.Run("8", func(b *testing.B) {
+		input := []byte(`"\u12345678\u9ABCDEF0\u12345678\u9ABCDEF0"`)
+		b.ReportAllocs()
+		b.SetBytes(int64(len(input)))
+
+		for i := 0; i < b.N; i++ {
+			p.parseBasicString(input)
+		}
+	})
+}
+
+func BenchmarkParseBasicStringsEasy(b *testing.B) {
+	p := &parser{}
+
+	for _, size := range []int{1, 4, 8, 16, 21} {
+		b.Run(strconv.Itoa(size), func(b *testing.B) {
+			input := []byte(`"` + strings.Repeat("A", size) + `"`)
+
+			b.ReportAllocs()
+			b.SetBytes(int64(len(input)))
+
+			for i := 0; i < b.N; i++ {
+				p.parseBasicString(input)
+			}
+		})
+	}
+}
+
+func TestParser_AST_DateTimes(t *testing.T) {
+	examples := []struct {
+		desc  string
+		input string
+		kind  ast.Kind
+		err   bool
+	}{
+		{
+			desc:  "offset-date-time with delim 'T' and UTC offset",
+			input: `2021-07-21T12:08:05Z`,
+			kind:  ast.DateTime,
+		},
+		{
+			desc:  "offset-date-time with space delim and +8hours offset",
+			input: `2021-07-21 12:08:05+08:00`,
+			kind:  ast.DateTime,
+		},
+		{
+			desc:  "local-date-time with nano second",
+			input: `2021-07-21T12:08:05.666666666`,
+			kind:  ast.LocalDateTime,
+		},
+		{
+			desc:  "local-date-time",
+			input: `2021-07-21T12:08:05`,
+			kind:  ast.LocalDateTime,
+		},
+		{
+			desc:  "local-date",
+			input: `2021-07-21`,
+			kind:  ast.LocalDate,
+		},
+	}
+
+	for _, e := range examples {
+		e := e
+		t.Run(e.desc, func(t *testing.T) {
+			p := parser{}
+			p.Reset([]byte(`A = ` + e.input))
+			p.NextExpression()
+			err := p.Error()
+			if e.err {
+				require.Error(t, err)
+			} else {
+				require.NoError(t, err)
+
+				expected := astNode{
+					Kind: ast.KeyValue,
+					Children: []astNode{
+						{Kind: e.kind, Data: []byte(e.input)},
+						{Kind: ast.Key, Data: []byte(`A`)},
+					},
+				}
+				compareNode(t, expected, p.Expression())
+			}
+		})
+	}
+}

scanner.go

@@ -49,16 +49,21 @@ func scanLiteralString(b []byte) ([]byte, []byte, error) {
 	// literal-string = apostrophe *literal-char apostrophe
 	// apostrophe = %x27 ; ' apostrophe
 	// literal-char = %x09 / %x20-26 / %x28-7E / non-ascii
-	for i := 1; i < len(b); i++ {
+	for i := 1; i < len(b); {
 		switch b[i] {
 		case '\'':
 			return b[:i+1], b[i+1:], nil
-		case '\n':
+		case '\n', '\r':
-			return nil, nil, newDecodeError(b[i:i+1], "literal strings cannot have new lines")
+			return nil, b[i+1:], newDecodeError(b[i:i+1], "literal strings cannot have new lines")
 		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, b[i+1:], newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], "unterminated literal string")
+	return nil, b[len(b):], newDecodeError(b[len(b):], "unterminated literal string")
 }
 
 func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
@@ -70,23 +75,62 @@ func scanMultilineLiteralString(b []byte) ([]byte, []byte, error) {
 	// mll-content = mll-char / newline
 	// mll-char = %x09 / %x20-26 / %x28-7E / non-ascii
 	// mll-quotes = 1*2apostrophe
-	for i := 3; i < len(b); i++ {
+	for i := 3; i < len(b); {
-		if b[i] == '\'' && scanFollowsMultilineLiteralStringDelimiter(b[i:]) {
+		switch b[i] {
-			return b[:i+3], b[i+3:], nil
+		case '\'':
+			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, b[i:], newDecodeError(b[i-3:i+1], "''' not allowed in multiline literal string")
+				}
+
+				return b[:i], b[i:], nil
+			}
+		case '\r':
+			if len(b) < i+2 {
+				return nil, b[i:], newDecodeError(b[len(b):], `need a \n after \r`)
+			}
+			if b[i+1] != '\n' {
+				return nil, b[i+2:], newDecodeError(b[i:i+2], `need a \n after \r`)
+			}
+			i += 2 // skip the \n
+			continue
 		}
+		size := utf8ValidNext(b[i:])
+		if size == 0 {
+			return nil, b[i:], newDecodeError(b[i:i+1], "invalid character")
+		}
+		i += size
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], `multiline literal string not terminated by '''`)
+	return nil, b[len(b):], 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")
+		return nil, b, newDecodeError(b, "windows new line expected")
 	}
 
 	if b[1] != '\n' {
-		return nil, nil, newDecodeError(b, `windows new line should be \r\n`)
+		return nil, b[2:], newDecodeError(b, `windows new line should be \r\n`)
 	}
 
 	return b[:lenCRLF], b[lenCRLF:], nil
@@ -106,45 +150,62 @@ func scanWhitespace(b []byte) ([]byte, []byte) {
 }
 
 //nolint:unparam
-func scanComment(b []byte) ([]byte, []byte) {
+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); i++ {
+
+	for i := 1; i < len(b); {
 		if b[i] == '\n' {
-			return b[:i], b[i:]
+			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, b[i+1:], newDecodeError(b[i:i+1], "invalid character in comment")
+		}
+
+		i += size
 	}
 
-	return b, nil
+	return b, b[len(b):], nil
 }
 
-func scanBasicString(b []byte) ([]byte, []byte, error) {
+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
-	for i := 1; i < len(b); i++ {
+	escaped := false
+	i := 1
+
+	for ; i < len(b); i++ {
 		switch b[i] {
 		case '"':
-			return b[:i+1], b[i+1:], nil
+			return b[:i+1], escaped, b[i+1:], nil
-		case '\n':
+		case '\n', '\r':
-			return nil, nil, newDecodeError(b[i:i+1], "basic strings cannot have new lines")
+			return nil, escaped, b[i+1:], newDecodeError(b[i:i+1], "basic strings cannot have new lines")
 		case '\\':
 			if len(b) < i+2 {
-				return nil, nil, newDecodeError(b[i:i+1], "need a character after \\")
+				return nil, escaped, b[i+1:], newDecodeError(b[i:i+1], "need a character after \\")
 			}
+			escaped = true
 			i++ // skip the next character
 		}
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], `basic string not terminated by "`)
+	return nil, escaped, b[len(b):], newDecodeError(b[len(b):], `basic string not terminated by "`)
 }
 
-func scanMultilineBasicString(b []byte) ([]byte, []byte, error) {
+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
@@ -155,19 +216,54 @@ func scanMultilineBasicString(b []byte) ([]byte, []byte, error) {
 	// mlb-quotes = 1*2quotation-mark
 	// mlb-unescaped = wschar / %x21 / %x23-5B / %x5D-7E / non-ascii
 	// mlb-escaped-nl = escape ws newline *( wschar / newline )
-	for i := 3; i < len(b); i++ {
+
+	escaped := false
+	i := 3
+
+	for ; i < len(b); i++ {
 		switch b[i] {
 		case '"':
 			if scanFollowsMultilineBasicStringDelimiter(b[i:]) {
-				return b[:i+3], b[i+3:], nil
+				i += 3
+
+				// At that point we found 3 apostrophe, and i is the
+				// index of the byte after the third one. The scanner
+				// needs to be eager, because there can be an extra 2
+				// apostrophe that can be accepted at the end of the
+				// string.
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i >= len(b) || b[i] != '"' {
+					return b[:i], escaped, b[i:], nil
+				}
+				i++
+
+				if i < len(b) && b[i] == '"' {
+					return nil, escaped, b[i+1:], newDecodeError(b[i-3:i+1], `""" not allowed in multiline basic string`)
+				}
+
+				return b[:i], escaped, b[i:], nil
 			}
 		case '\\':
 			if len(b) < i+2 {
-				return nil, nil, newDecodeError(b[len(b):], "need a character after \\")
+				return nil, escaped, b[len(b):], newDecodeError(b[len(b):], "need a character after \\")
 			}
+			escaped = true
 			i++ // skip the next character
+		case '\r':
+			if len(b) < i+2 {
+				return nil, escaped, b[len(b):], newDecodeError(b[len(b):], `need a \n after \r`)
+			}
+			if b[i+1] != '\n' {
+				return nil, escaped, b[i+2:], newDecodeError(b[i:i+2], `need a \n after \r`)
+			}
+			i++ // skip the \n
 		}
 	}
 
-	return nil, nil, newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
+	return nil, escaped, b[len(b):], newDecodeError(b[len(b):], `multiline basic string not terminated by """`)
 }

strict.go

@@ -2,6 +2,7 @@ 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"
 )
 
@@ -14,7 +15,7 @@ type strict struct {
 	missing []decodeError
 }
 
-func (s *strict) EnterTable(node ast.Node) {
+func (s *strict) EnterTable(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -22,7 +23,7 @@ func (s *strict) EnterTable(node ast.Node) {
 	s.key.UpdateTable(node)
 }
 
-func (s *strict) EnterArrayTable(node ast.Node) {
+func (s *strict) EnterArrayTable(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -30,7 +31,7 @@ func (s *strict) EnterArrayTable(node ast.Node) {
 	s.key.UpdateArrayTable(node)
 }
 
-func (s *strict) EnterKeyValue(node ast.Node) {
+func (s *strict) EnterKeyValue(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -38,7 +39,7 @@ func (s *strict) EnterKeyValue(node ast.Node) {
 	s.key.Push(node)
 }
 
-func (s *strict) ExitKeyValue(node ast.Node) {
+func (s *strict) ExitKeyValue(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -46,7 +47,7 @@ func (s *strict) ExitKeyValue(node ast.Node) {
 	s.key.Pop(node)
 }
 
-func (s *strict) MissingTable(node ast.Node) {
+func (s *strict) MissingTable(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -58,7 +59,7 @@ func (s *strict) MissingTable(node ast.Node) {
 	})
 }
 
-func (s *strict) MissingField(node ast.Node) {
+func (s *strict) MissingField(node *ast.Node) {
 	if !s.Enabled {
 		return
 	}
@@ -86,3 +87,21 @@ func (s *strict) Error(doc []byte) error {
 
 	return err
 }
+
+func keyLocation(node *ast.Node) []byte {
+	k := node.Key()
+
+	hasOne := k.Next()
+	if !hasOne {
+		panic("should not be called with empty key")
+	}
+
+	start := k.Node().Data
+	end := k.Node().Data
+
+	for k.Next() {
+		end = k.Node().Data
+	}
+
+	return danger.BytesRange(start, end)
+}

targets.go

@@ -1,536 +0,0 @@
-package toml
-
-import (
-	"fmt"
-	"math"
-	"reflect"
-	"strings"
-	"sync"
-)
-
-type target interface {
-	// Dereferences the target.
-	get() reflect.Value
-
-	// Store a string at the target.
-	setString(v string)
-
-	// Store a boolean at the target
-	setBool(v bool)
-
-	// Store an int64 at the target
-	setInt64(v int64)
-
-	// Store a float64 at the target
-	setFloat64(v float64)
-
-	// Stores any value at the target
-	set(v reflect.Value)
-}
-
-// valueTarget just contains a reflect.Value that can be set.
-// It is used for struct fields.
-type valueTarget reflect.Value
-
-func (t valueTarget) get() reflect.Value {
-	return reflect.Value(t)
-}
-
-func (t valueTarget) set(v reflect.Value) {
-	reflect.Value(t).Set(v)
-}
-
-func (t valueTarget) setString(v string) {
-	t.get().SetString(v)
-}
-
-func (t valueTarget) setBool(v bool) {
-	t.get().SetBool(v)
-}
-
-func (t valueTarget) setInt64(v int64) {
-	t.get().SetInt(v)
-}
-
-func (t valueTarget) setFloat64(v float64) {
-	t.get().SetFloat(v)
-}
-
-// interfaceTarget wraps an other target to dereference on get.
-type interfaceTarget struct {
-	x target
-}
-
-func (t interfaceTarget) get() reflect.Value {
-	return t.x.get().Elem()
-}
-
-func (t interfaceTarget) set(v reflect.Value) {
-	t.x.set(v)
-}
-
-func (t interfaceTarget) setString(v string) {
-	panic("interface targets should always go through set")
-}
-
-func (t interfaceTarget) setBool(v bool) {
-	panic("interface targets should always go through set")
-}
-
-func (t interfaceTarget) setInt64(v int64) {
-	panic("interface targets should always go through set")
-}
-
-func (t interfaceTarget) setFloat64(v float64) {
-	panic("interface targets should always go through set")
-}
-
-// mapTarget targets a specific key of a map.
-type mapTarget struct {
-	v reflect.Value
-	k reflect.Value
-}
-
-func (t mapTarget) get() reflect.Value {
-	return t.v.MapIndex(t.k)
-}
-
-func (t mapTarget) set(v reflect.Value) {
-	t.v.SetMapIndex(t.k, v)
-}
-
-func (t mapTarget) setString(v string) {
-	t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setBool(v bool) {
-	t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setInt64(v int64) {
-	t.set(reflect.ValueOf(v))
-}
-
-func (t mapTarget) setFloat64(v float64) {
-	t.set(reflect.ValueOf(v))
-}
-
-// makes sure that the value pointed at by t is indexable (Slice, Array), or
-// dereferences to an indexable (Ptr, Interface).
-func ensureValueIndexable(t target) error {
-	f := t.get()
-
-	switch f.Type().Kind() {
-	case reflect.Slice:
-		if f.IsNil() {
-			t.set(reflect.MakeSlice(f.Type(), 0, 0))
-			return nil
-		}
-	case reflect.Interface:
-		if f.IsNil() || f.Elem().Type() != sliceInterfaceType {
-			t.set(reflect.MakeSlice(sliceInterfaceType, 0, 0))
-			return nil
-		}
-	case reflect.Ptr:
-		panic("pointer should have already been dereferenced")
-	case reflect.Array:
-		// arrays are always initialized.
-	default:
-		return fmt.Errorf("toml: cannot store array in a %s", f.Kind())
-	}
-
-	return nil
-}
-
-var (
-	sliceInterfaceType     = reflect.TypeOf([]interface{}{})
-	mapStringInterfaceType = reflect.TypeOf(map[string]interface{}{})
-)
-
-func ensureMapIfInterface(x target) {
-	v := x.get()
-
-	if v.Kind() == reflect.Interface && v.IsNil() {
-		newElement := reflect.MakeMap(mapStringInterfaceType)
-
-		x.set(newElement)
-	}
-}
-
-func setString(t target, v string) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.String:
-		t.setString(v)
-	case reflect.Interface:
-		t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("toml: cannot assign string to a %s", f.Kind())
-	}
-
-	return nil
-}
-
-func setBool(t target, v bool) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Bool:
-		t.setBool(v)
-	case reflect.Interface:
-		t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("toml: cannot assign boolean to a %s", f.Kind())
-	}
-
-	return nil
-}
-
-const (
-	maxInt = int64(^uint(0) >> 1)
-	minInt = -maxInt - 1
-)
-
-//nolint:funlen,gocognit,cyclop
-func setInt64(t target, v int64) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Int64:
-		t.setInt64(v)
-	case reflect.Int32:
-		if v < math.MinInt32 || v > math.MaxInt32 {
-			return fmt.Errorf("toml: number %d does not fit in an int32", v)
-		}
-
-		t.set(reflect.ValueOf(int32(v)))
-		return nil
-	case reflect.Int16:
-		if v < math.MinInt16 || v > math.MaxInt16 {
-			return fmt.Errorf("toml: number %d does not fit in an int16", v)
-		}
-
-		t.set(reflect.ValueOf(int16(v)))
-	case reflect.Int8:
-		if v < math.MinInt8 || v > math.MaxInt8 {
-			return fmt.Errorf("toml: number %d does not fit in an int8", v)
-		}
-
-		t.set(reflect.ValueOf(int8(v)))
-	case reflect.Int:
-		if v < minInt || v > maxInt {
-			return fmt.Errorf("toml: number %d does not fit in an int", v)
-		}
-
-		t.set(reflect.ValueOf(int(v)))
-	case reflect.Uint64:
-		if v < 0 {
-			return fmt.Errorf("toml: negative number %d does not fit in an uint64", v)
-		}
-
-		t.set(reflect.ValueOf(uint64(v)))
-	case reflect.Uint32:
-		if v < 0 || v > math.MaxUint32 {
-			return fmt.Errorf("toml: negative number %d does not fit in an uint32", v)
-		}
-
-		t.set(reflect.ValueOf(uint32(v)))
-	case reflect.Uint16:
-		if v < 0 || v > math.MaxUint16 {
-			return fmt.Errorf("toml: negative number %d does not fit in an uint16", v)
-		}
-
-		t.set(reflect.ValueOf(uint16(v)))
-	case reflect.Uint8:
-		if v < 0 || v > math.MaxUint8 {
-			return fmt.Errorf("toml: negative number %d does not fit in an uint8", v)
-		}
-
-		t.set(reflect.ValueOf(uint8(v)))
-	case reflect.Uint:
-		if v < 0 {
-			return fmt.Errorf("toml: negative number %d does not fit in an uint", v)
-		}
-
-		t.set(reflect.ValueOf(uint(v)))
-	case reflect.Interface:
-		t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("toml: integer cannot be assigned to %s", f.Kind())
-	}
-
-	return nil
-}
-
-func setFloat64(t target, v float64) error {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Float64:
-		t.setFloat64(v)
-	case reflect.Float32:
-		if v > math.MaxFloat32 {
-			return fmt.Errorf("toml: number %f does not fit in a float32", v)
-		}
-
-		t.set(reflect.ValueOf(float32(v)))
-	case reflect.Interface:
-		t.set(reflect.ValueOf(v))
-	default:
-		return fmt.Errorf("toml: float cannot be assigned to %s", f.Kind())
-	}
-
-	return nil
-}
-
-// Returns the element at idx of the value pointed at by target, or an error if
-// t does not point to an indexable.
-// If the target points to an Array and idx is out of bounds, it returns
-// (nil, nil) as this is not a fatal error (the unmarshaler will skip).
-func elementAt(t target, idx int) target {
-	f := t.get()
-
-	switch f.Kind() {
-	case reflect.Slice:
-		//nolint:godox
-		// TODO: use the idx function argument and avoid alloc if possible.
-		idx := f.Len()
-
-		t.set(reflect.Append(f, reflect.New(f.Type().Elem()).Elem()))
-
-		return valueTarget(t.get().Index(idx))
-	case reflect.Array:
-		if idx >= f.Len() {
-			return nil
-		}
-
-		return valueTarget(f.Index(idx))
-	case reflect.Interface:
-		// This function is called after ensureValueIndexable, so it's
-		// guaranteed that f contains an initialized slice.
-		ifaceElem := f.Elem()
-		idx := ifaceElem.Len()
-		newElem := reflect.New(ifaceElem.Type().Elem()).Elem()
-		newSlice := reflect.Append(ifaceElem, newElem)
-
-		t.set(newSlice)
-
-		return valueTarget(t.get().Elem().Index(idx))
-	default:
-		// Why ensureValueIndexable let it go through?
-		panic(fmt.Errorf("elementAt received unhandled value type: %s", f.Kind()))
-	}
-}
-
-func (d *decoder) scopeTableTarget(shouldAppend bool, t target, name string) (target, bool, error) {
-	x := t.get()
-
-	switch x.Kind() {
-	// Kinds that need to recurse
-	case reflect.Interface:
-		t := scopeInterface(shouldAppend, t)
-		return d.scopeTableTarget(shouldAppend, t, name)
-	case reflect.Ptr:
-		t := scopePtr(t)
-		return d.scopeTableTarget(shouldAppend, t, name)
-	case reflect.Slice:
-		t := scopeSlice(shouldAppend, t)
-		shouldAppend = false
-		return d.scopeTableTarget(shouldAppend, t, name)
-	case reflect.Array:
-		t, err := d.scopeArray(shouldAppend, t)
-		if err != nil {
-			return t, false, err
-		}
-		shouldAppend = false
-
-		return d.scopeTableTarget(shouldAppend, t, name)
-
-	// Terminal kinds
-	case reflect.Struct:
-		return scopeStruct(x, name)
-	case reflect.Map:
-		if x.IsNil() {
-			t.set(reflect.MakeMap(x.Type()))
-			x = t.get()
-		}
-
-		return scopeMap(x, name)
-	default:
-		panic(fmt.Sprintf("can't scope on a %s", x.Kind()))
-	}
-}
-
-func scopeInterface(shouldAppend bool, t target) target {
-	initInterface(shouldAppend, t)
-	return interfaceTarget{t}
-}
-
-func scopePtr(t target) target {
-	initPtr(t)
-	return valueTarget(t.get().Elem())
-}
-
-func initPtr(t target) {
-	x := t.get()
-	if !x.IsNil() {
-		return
-	}
-
-	t.set(reflect.New(x.Type().Elem()))
-}
-
-// initInterface makes sure that the interface pointed at by the target is not
-// nil.
-// Returns the target to the initialized value of the target.
-func initInterface(shouldAppend bool, t target) {
-	x := t.get()
-
-	if x.Kind() != reflect.Interface {
-		panic("this should only be called on interfaces")
-	}
-
-	if !x.IsNil() && (x.Elem().Type() == sliceInterfaceType || x.Elem().Type() == mapStringInterfaceType) {
-		return
-	}
-
-	var newElement reflect.Value
-	if shouldAppend {
-		newElement = reflect.MakeSlice(sliceInterfaceType, 0, 0)
-	} else {
-		newElement = reflect.MakeMap(mapStringInterfaceType)
-	}
-
-	t.set(newElement)
-}
-
-func scopeSlice(shouldAppend bool, t target) target {
-	v := t.get()
-
-	if shouldAppend {
-		newElem := reflect.New(v.Type().Elem())
-		newSlice := reflect.Append(v, newElem.Elem())
-
-		t.set(newSlice)
-
-		v = t.get()
-	}
-
-	return valueTarget(v.Index(v.Len() - 1))
-}
-
-func (d *decoder) scopeArray(shouldAppend bool, t target) (target, error) {
-	v := t.get()
-
-	idx := d.arrayIndex(shouldAppend, v)
-
-	if idx >= v.Len() {
-		return nil, fmt.Errorf("toml: impossible to insert element beyond array's size: %d", v.Len())
-	}
-
-	return valueTarget(v.Index(idx)), nil
-}
-
-func scopeMap(v reflect.Value, name string) (target, bool, error) {
-	k := reflect.ValueOf(name)
-
-	keyType := v.Type().Key()
-	if !k.Type().AssignableTo(keyType) {
-		if !k.Type().ConvertibleTo(keyType) {
-			return nil, false, fmt.Errorf("toml: cannot convert map key of type %s to expected type %s", k.Type(), keyType)
-		}
-
-		k = k.Convert(keyType)
-	}
-
-	if !v.MapIndex(k).IsValid() {
-		newElem := reflect.New(v.Type().Elem())
-		v.SetMapIndex(k, newElem.Elem())
-	}
-
-	return mapTarget{
-		v: v,
-		k: k,
-	}, true, nil
-}
-
-type fieldPathsMap = map[string][]int
-
-type fieldPathsCache struct {
-	m map[reflect.Type]fieldPathsMap
-	l sync.RWMutex
-}
-
-func (c *fieldPathsCache) get(t reflect.Type) (fieldPathsMap, bool) {
-	c.l.RLock()
-	paths, ok := c.m[t]
-	c.l.RUnlock()
-
-	return paths, ok
-}
-
-func (c *fieldPathsCache) set(t reflect.Type, m fieldPathsMap) {
-	c.l.Lock()
-	c.m[t] = m
-	c.l.Unlock()
-}
-
-var globalFieldPathsCache = fieldPathsCache{
-	m: map[reflect.Type]fieldPathsMap{},
-	l: sync.RWMutex{},
-}
-
-func scopeStruct(v reflect.Value, name string) (target, bool, error) {
-	//nolint:godox
-	// TODO: cache this, and reduce allocations
-	fieldPaths, ok := globalFieldPathsCache.get(v.Type())
-	if !ok {
-		fieldPaths = map[string][]int{}
-
-		path := make([]int, 0, 16)
-
-		var walk func(reflect.Value)
-		walk = func(v reflect.Value) {
-			t := v.Type()
-			for i := 0; i < t.NumField(); i++ {
-				l := len(path)
-				path = append(path, i)
-				f := t.Field(i)
-
-				if f.Anonymous {
-					walk(v.Field(i))
-				} else if f.PkgPath == "" {
-					// only consider exported fields
-					fieldName, ok := f.Tag.Lookup("toml")
-					if !ok {
-						fieldName = f.Name
-					}
-
-					pathCopy := make([]int, len(path))
-					copy(pathCopy, path)
-
-					fieldPaths[fieldName] = pathCopy
-					// extra copy for the case-insensitive match
-					fieldPaths[strings.ToLower(fieldName)] = pathCopy
-				}
-				path = path[:l]
-			}
-		}
-
-		walk(v)
-
-		globalFieldPathsCache.set(v.Type(), fieldPaths)
-	}
-
-	path, ok := fieldPaths[name]
-	if !ok {
-		path, ok = fieldPaths[strings.ToLower(name)]
-	}
-
-	if !ok {
-		return nil, false, nil
-	}
-
-	return valueTarget(v.FieldByIndex(path)), true, nil
-}

targets_test.go

@@ -1,207 +0,0 @@
-package toml
-
-import (
-	"reflect"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestStructTarget_Ensure(t *testing.T) {
-	t.Parallel()
-
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		test  func(v reflect.Value, err error)
-	}{
-		{
-			desc:  "handle a nil slice of string",
-			input: reflect.ValueOf(&struct{ A []string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				assert.False(t, v.IsNil())
-			},
-		},
-		{
-			desc:  "handle an existing slice of string",
-			input: reflect.ValueOf(&struct{ A []string }{A: []string{"foo"}}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				require.False(t, v.IsNil())
-
-				s, ok := v.Interface().([]string)
-				if !ok {
-					t.Errorf("interface %v should be castable into []string", s)
-					return
-				}
-
-				assert.Equal(t, []string{"foo"}, s)
-			},
-		},
-	}
-
-	for _, e := range examples {
-		e := e
-		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
-
-			d := decoder{}
-			target, _, err := d.scopeTableTarget(false, valueTarget(e.input), e.name)
-			require.NoError(t, err)
-			err = ensureValueIndexable(target)
-			v := target.get()
-			e.test(v, err)
-		})
-	}
-}
-
-func TestStructTarget_SetString(t *testing.T) {
-	t.Parallel()
-
-	str := "value"
-
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		test  func(v reflect.Value, err error)
-	}{
-		{
-			desc:  "sets a string",
-			input: reflect.ValueOf(&struct{ A string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.NoError(t, err)
-				assert.Equal(t, str, v.String())
-			},
-		},
-		{
-			desc:  "fails on a float",
-			input: reflect.ValueOf(&struct{ A float64 }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.Error(t, err)
-			},
-		},
-		{
-			desc:  "fails on a slice",
-			input: reflect.ValueOf(&struct{ A []string }{}).Elem(),
-			name:  "A",
-			test: func(v reflect.Value, err error) {
-				assert.Error(t, err)
-			},
-		},
-	}
-
-	for _, e := range examples {
-		e := e
-		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
-
-			d := decoder{}
-			target, _, err := d.scopeTableTarget(false, valueTarget(e.input), e.name)
-			require.NoError(t, err)
-			err = setString(target, str)
-			v := target.get()
-			e.test(v, err)
-		})
-	}
-}
-
-func TestPushNew(t *testing.T) {
-	t.Parallel()
-
-	t.Run("slice of strings", func(t *testing.T) {
-		t.Parallel()
-
-		type Doc struct {
-			A []string
-		}
-		d := Doc{}
-
-		dec := decoder{}
-		x, _, err := dec.scopeTableTarget(false, valueTarget(reflect.ValueOf(&d).Elem()), "A")
-		require.NoError(t, err)
-
-		n := elementAt(x, 0)
-		n.setString("hello")
-		require.Equal(t, []string{"hello"}, d.A)
-
-		n = elementAt(x, 1)
-		n.setString("world")
-		require.Equal(t, []string{"hello", "world"}, d.A)
-	})
-
-	t.Run("slice of interfaces", func(t *testing.T) {
-		t.Parallel()
-
-		type Doc struct {
-			A []interface{}
-		}
-		d := Doc{}
-
-		dec := decoder{}
-		x, _, err := dec.scopeTableTarget(false, valueTarget(reflect.ValueOf(&d).Elem()), "A")
-		require.NoError(t, err)
-
-		n := elementAt(x, 0)
-		require.NoError(t, setString(n, "hello"))
-		require.Equal(t, []interface{}{"hello"}, d.A)
-
-		n = elementAt(x, 1)
-		require.NoError(t, setString(n, "world"))
-		require.Equal(t, []interface{}{"hello", "world"}, d.A)
-	})
-}
-
-func TestScope_Struct(t *testing.T) {
-	t.Parallel()
-
-	examples := []struct {
-		desc  string
-		input reflect.Value
-		name  string
-		err   bool
-		found bool
-		idx   []int
-	}{
-		{
-			desc:  "simple field",
-			input: reflect.ValueOf(&struct{ A string }{}).Elem(),
-			name:  "A",
-			idx:   []int{0},
-			found: true,
-		},
-		{
-			desc:  "fails not-exported field",
-			input: reflect.ValueOf(&struct{ a string }{}).Elem(),
-			name:  "a",
-			err:   false,
-			found: false,
-		},
-	}
-
-	for _, e := range examples {
-		e := e
-		t.Run(e.desc, func(t *testing.T) {
-			t.Parallel()
-
-			dec := decoder{}
-			x, found, err := dec.scopeTableTarget(false, valueTarget(e.input), e.name)
-			assert.Equal(t, e.found, found)
-			if e.err {
-				assert.Error(t, err)
-			}
-			if found {
-				x2, ok := x.(valueTarget)
-				require.True(t, ok)
-				x2.get()
-			}
-		})
-	}
-}

testdata/fuzz/FuzzUnmarshal/0c08b924aa26af23d16b77029b4f5567a96af93402e0472fef7b837b37580d7c

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

testdata/fuzz/FuzzUnmarshal/249a017d8ffa0d88d41c594ada5399f764833f64050180cb39f106d12666853f

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

testdata/fuzz/FuzzUnmarshal/391ce67b866d58464d067c5981e99951fd499896453a2fe2779a6bfe7df11bf5

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

testdata/fuzz/FuzzUnmarshal/49ae83283c5cfbed874bcacbfb342de107b69c2a3d3c35c2d99f67345fe89946

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

testdata/fuzz/FuzzUnmarshal/69ea82e85bb4ad966ae86bb05eea3af69ed1a826c846f1bf2282f329b4d8fb36

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

testdata/fuzz/FuzzUnmarshal/746910fac6deb42e7e62dce60b333c81c39d844a122c9075894c023400e8fdbf

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

testdata/fuzz/FuzzUnmarshal/844d1171a3adc95e51627f0e9378bbaa193fe2d0284f4ba8ed19a5226a8cbc74

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

testdata/fuzz/FuzzUnmarshal/900c54b5071113ae2a5365b3156377ddd775b9d38add61629544903cb9c94b00

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

testdata/fuzz/FuzzUnmarshal/c87e8eb4e8fa40fd3c89d8ed16a049ce81592ded710965ee29a4e506eb5427b3

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

testdata/fuzz/FuzzUnmarshal/d0e5d6fc3b865cfae5480c01f301b87c932722ae0c1d692deea413349ea028be

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

testdata/fuzz/FuzzUnmarshal/d2cecea5b8ee5f148419671cef0644178cdba1ee91d3f295be38da576e89ef40

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

toml_testgen_support_test.go

@@ -1,14 +1,14 @@
+//go:generate go run ./cmd/tomltestgen/main.go -o toml_testgen_test.go
+
 // This is a support file for toml_testgen_test.go
 package toml_test
 
 import (
 	"encoding/json"
-	"fmt"
-	"strconv"
 	"testing"
-	"time"
 
 	"github.com/pelletier/go-toml/v2"
+	"github.com/pelletier/go-toml/v2/internal/testsuite"
 	"github.com/stretchr/testify/require"
 )
 
@@ -17,10 +17,14 @@ func testgenInvalid(t *testing.T, input string) {
 	t.Logf("Input TOML:\n%s", input)
 
 	doc := map[string]interface{}{}
-	err := toml.Unmarshal([]byte(input), &doc)
+	err := testsuite.Unmarshal([]byte(input), &doc)
 
 	if err == nil {
-		t.Log(json.Marshal(doc))
+		out, err := json.Marshal(doc)
+		if err != nil {
+			panic("could not marshal map to json")
+		}
+		t.Log("JSON output from unmarshal:", string(out))
 		t.Fatalf("test did not fail")
 	}
 }
@@ -29,124 +33,26 @@ func testgenValid(t *testing.T, input string, jsonRef string) {
 	t.Helper()
 	t.Logf("Input TOML:\n%s", input)
 
-	doc := map[string]interface{}{}
+	// TODO: change this to interface{}
+	var doc map[string]interface{}
 
-	err := toml.Unmarshal([]byte(input), &doc)
+	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)
-	refDoc := testgenBuildRefDoc(jsonRef)
-
-	require.Equal(t, refDoc, doc)
-
-	out, err := toml.Marshal(doc)
 	require.NoError(t, err)
 
-	doc2 := map[string]interface{}{}
+	var ref interface{}
-	err = toml.Unmarshal(out, &doc2)
+	err = json.Unmarshal([]byte(jsonRef), &ref)
 	require.NoError(t, err)
 
-	require.Equal(t, refDoc, doc2)
+	var actual interface{}
-}
+	err = json.Unmarshal([]byte(j), &actual)
-
+	require.NoError(t, err)
-func testgenBuildRefDoc(jsonRef string) map[string]interface{} {
+
-	descTree := map[string]interface{}{}
+	testsuite.CmpJSON(t, "", ref, actual)
-
-	err := json.Unmarshal([]byte(jsonRef), &descTree)
-	if err != nil {
-		panic(fmt.Sprintf("reference doc should be valid JSON: %s", err))
-	}
-
-	doc := testGenTranslateDesc(descTree)
-	if doc == nil {
-		return map[string]interface{}{}
-	}
-
-	return doc.(map[string]interface{})
-}
-
-//nolint:funlen,gocognit,cyclop
-func testGenTranslateDesc(input interface{}) interface{} {
-	a, ok := input.([]interface{})
-	if ok {
-		xs := make([]interface{}, len(a))
-		for i, v := range a {
-			xs[i] = testGenTranslateDesc(v)
-		}
-
-		return xs
-	}
-
-	d, ok := input.(map[string]interface{})
-	if !ok {
-		panic(fmt.Sprintf("input should be valid map[string]: %v", input))
-	}
-
-	var (
-		dtype  string
-		dvalue interface{}
-	)
-
-	//nolint:nestif
-	if len(d) == 2 {
-		dtypeiface, ok := d["type"]
-		if ok {
-			dvalue, ok = d["value"]
-			if ok {
-				dtype = dtypeiface.(string)
-
-				switch dtype {
-				case "string":
-					return dvalue.(string)
-				case "float":
-					v, err := strconv.ParseFloat(dvalue.(string), 64)
-					if err != nil {
-						panic(fmt.Sprintf("invalid float '%s': %s", dvalue, err))
-					}
-
-					return v
-				case "integer":
-					v, err := strconv.ParseInt(dvalue.(string), 10, 64)
-					if err != nil {
-						panic(fmt.Sprintf("invalid int '%s': %s", dvalue, err))
-					}
-
-					return v
-				case "bool":
-					return dvalue.(string) == "true"
-				case "datetime":
-					dt, err := time.Parse("2006-01-02T15:04:05Z", dvalue.(string))
-					if err != nil {
-						panic(fmt.Sprintf("invalid datetime '%s': %s", dvalue, err))
-					}
-
-					return dt
-				case "array":
-					if dvalue == nil {
-						return nil
-					}
-
-					a := dvalue.([]interface{})
-
-					xs := make([]interface{}, len(a))
-
-					for i, v := range a {
-						xs[i] = testGenTranslateDesc(v)
-					}
-
-					return xs
-				}
-
-				panic(fmt.Sprintf("unknown type: %s", dtype))
-			}
-		}
-	}
-
-	dest := map[string]interface{}{}
-	for k, v := range d {
-		dest[k] = testGenTranslateDesc(v)
-	}
-
-	return dest
 }

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
+)