|
| 1 | +--- |
| 2 | +id: 89 |
| 3 | +title: TOC generator directive and MDS035 auto-fix |
| 4 | +status: "🔲" |
| 5 | +summary: >- |
| 6 | + Add a `<?toc?>...<?/toc?>` generated-section |
| 7 | + directive that emits a nested list of the |
| 8 | + current document's headings linked to their |
| 9 | + anchors (MDS036). Upgrade MDS035 (plan 88) to |
| 10 | + auto-fix each detected renderer-specific TOC |
| 11 | + token by replacing it with a `<?toc?>` block, |
| 12 | + which the directive then regenerates on |
| 13 | + `mdsmith fix`. |
| 14 | +--- |
| 15 | +# TOC generator directive and MDS035 auto-fix |
| 16 | + |
| 17 | +## Goal |
| 18 | + |
| 19 | +Give mdsmith a native heading-level TOC |
| 20 | +generator so MDS035's detection can be followed |
| 21 | +by an actual fix. Today MDS035 only flags |
| 22 | +`[TOC]`, `[[_TOC_]]`, `[[toc]]`, `${toc}` and |
| 23 | +points at `<?catalog?>`, which is a file-index |
| 24 | +directive, not a heading-outline replacement. |
| 25 | + |
| 26 | +## Context |
| 27 | + |
| 28 | +Plan 88 landed MDS035 as detection-only because |
| 29 | +mdsmith had no equivalent for in-document |
| 30 | +heading TOCs — the common case of those tokens. |
| 31 | +`<?catalog?>` lists *other* files matching a |
| 32 | +glob, so it is the right replacement only on |
| 33 | +index pages, not on the per-document TOCs |
| 34 | +authors usually want. |
| 35 | + |
| 36 | +mdsmith already has the generated-section |
| 37 | +machinery needed here. |
| 38 | +[`<?catalog?>`][catalog] and |
| 39 | +[`<?include?>`][include] each read a directive |
| 40 | +body, compute content, and emit it between |
| 41 | +`<?name ...?>` and `<?/name?>` markers. A rule |
| 42 | +keeps the content current on `mdsmith fix`. |
| 43 | +See the [generated-section archetype][gensection] |
| 44 | +for the shared mechanics. |
| 45 | + |
| 46 | +[catalog]: ../internal/rules/MDS019-catalog/README.md |
| 47 | +[include]: ../internal/rules/MDS021-include/README.md |
| 48 | +[gensection]: ../docs/background/archetypes/generated-section/README.md |
| 49 | + |
| 50 | +### Why both pieces in one plan |
| 51 | + |
| 52 | +The `<?toc?>` directive is only useful to |
| 53 | +MDS035 if MDS035 knows to rewrite its detected |
| 54 | +tokens to it; conversely, MDS035's auto-fix is |
| 55 | +only possible once `<?toc?>` exists. Splitting |
| 56 | +the work across two plans would leave one half |
| 57 | +dead code until the other lands. |
| 58 | + |
| 59 | +## Design |
| 60 | + |
| 61 | +### Directive syntax |
| 62 | + |
| 63 | +```text |
| 64 | +<?toc |
| 65 | +min-level: 2 |
| 66 | +max-level: 3 |
| 67 | +?> |
| 68 | +- [First heading](#first-heading) |
| 69 | + - [Subheading](#subheading) |
| 70 | +- [Second heading](#second-heading) |
| 71 | +<?/toc?> |
| 72 | +``` |
| 73 | + |
| 74 | +Parameters (all optional): |
| 75 | + |
| 76 | +| Name | Type | Default | Description | |
| 77 | +|-------------|------|---------|-----------------------------------------------------| |
| 78 | +| `min-level` | int | `2` | Lowest heading level to include (1–6) | |
| 79 | +| `max-level` | int | `6` | Highest heading level to include (1–6, ≥ min-level) | |
| 80 | + |
| 81 | +`min-level: 2` excludes the document title by |
| 82 | +default, matching what Python-Markdown's `[TOC]` |
| 83 | +produces with default settings. |
| 84 | + |
| 85 | +### Generated content |
| 86 | + |
| 87 | +A nested unordered list, one item per heading |
| 88 | +in source order. Each item links to a |
| 89 | +GitHub-flavor slug of the heading text: |
| 90 | + |
| 91 | +- Lowercase the heading plain text. |
| 92 | +- Replace spaces with `-`. |
| 93 | +- Strip characters outside `[a-z0-9-_]`. |
| 94 | +- Disambiguate repeats by appending `-1`, `-2`, … |
| 95 | + in source order (matching goldmark-meta / |
| 96 | + GitHub's behavior). |
| 97 | + |
| 98 | +Indentation is `<?listindent?>`-aware: use the |
| 99 | +same per-level indent as MDS016 (two spaces by |
| 100 | +default). |
| 101 | + |
| 102 | +List items respect the heading structure, not |
| 103 | +the raw level. Given H2 → H4 → H2, the tree is: |
| 104 | + |
| 105 | +```text |
| 106 | +- [h2a](#h2a) |
| 107 | + - [h4](#h4) |
| 108 | +- [h2b](#h2b) |
| 109 | +``` |
| 110 | + |
| 111 | +…not a flat list keyed on absolute level. |
| 112 | + |
| 113 | +### Rule: MDS036 (toc) |
| 114 | + |
| 115 | +- ID: `MDS036` |
| 116 | +- Name: `toc` |
| 117 | +- Category: `meta` |
| 118 | +- Default: enabled (generated sections are |
| 119 | + enabled by default for the project; users opt |
| 120 | + out by removing the directive) |
| 121 | +- Fixable: yes |
| 122 | + |
| 123 | +`Check` diff-compares the body between the |
| 124 | +markers against the regenerated output. |
| 125 | +`Fix` rewrites the body. |
| 126 | + |
| 127 | +Use the shared `internal/archetype/gensection` |
| 128 | +engine that `catalog` already uses. New |
| 129 | +directive logic lives in |
| 130 | +`internal/rules/toc/`. |
| 131 | + |
| 132 | +### MDS035 auto-fix |
| 133 | + |
| 134 | +MDS035 becomes a `FixableRule`. On `Fix`, for |
| 135 | +each matched directive line inside a paragraph: |
| 136 | + |
| 137 | +- `[TOC]` (unresolved), `[[_TOC_]]`, `[[toc]]`, |
| 138 | + `${toc}` → replace the single line with: |
| 139 | + |
| 140 | + ```text |
| 141 | + <?toc |
| 142 | + ?> |
| 143 | + <?/toc?> |
| 144 | + ``` |
| 145 | + |
| 146 | + The empty YAML body means "use defaults". A |
| 147 | + trailing newline separator is inserted if the |
| 148 | + surrounding paragraph would otherwise fuse |
| 149 | + the directive into adjacent text. |
| 150 | + |
| 151 | +- `[TOC]` with a matching link reference |
| 152 | + definition → leave untouched (already |
| 153 | + suppressed by Check; must stay out of Fix). |
| 154 | + |
| 155 | +The Fix output is plain `<?toc?>...<?/toc?>` |
| 156 | +with an **empty** body. MDS036 runs in a |
| 157 | +subsequent pass of the same `mdsmith fix` |
| 158 | +invocation (mdsmith already supports |
| 159 | +multi-pass fix, used by MDS019/MDS021) and |
| 160 | +fills in the heading list. If the one-pass |
| 161 | +semantics become brittle, an alternative is for |
| 162 | +MDS035 to call into `gensection` directly and |
| 163 | +emit populated content; start with multi-pass |
| 164 | +and fall back only if needed. |
| 165 | + |
| 166 | +### Interaction with existing rules |
| 167 | + |
| 168 | +- MDS015 (blank line around fenced code / |
| 169 | + generated sections): the Fix output adds the |
| 170 | + required blank-line padding around the |
| 171 | + inserted block. |
| 172 | +- MDS020 (required-structure): no change; TOC |
| 173 | + blocks are not a required section. |
| 174 | +- MDS019 (catalog), MDS021 (include): orthogonal |
| 175 | + — different directive names. |
| 176 | + |
| 177 | +### Out of scope |
| 178 | + |
| 179 | +- Custom link anchor format overrides |
| 180 | + (e.g., non-GitHub slugger). Always GitHub- |
| 181 | + style for the first release; add a |
| 182 | + `slugger:` parameter later if needed. |
| 183 | +- Skipping specific headings via frontmatter |
| 184 | + or attribute syntax. Use `min-level` / |
| 185 | + `max-level` for now. |
| 186 | +- Rendering `[TOC]` into `<?toc?>` with |
| 187 | + preserved non-default parameters. The four |
| 188 | + renderer-specific tokens have no shared |
| 189 | + parameter surface; always emit the default |
| 190 | + `<?toc?>`. |
| 191 | + |
| 192 | +## Tasks |
| 193 | + |
| 194 | +1. Create the `<?toc?>` directive in a new |
| 195 | + `internal/rules/toc/` package using the |
| 196 | + shared `internal/archetype/gensection.Engine` |
| 197 | + (same engine as `catalog`). Add a slug |
| 198 | + helper in `internal/mdtext/` or the toc |
| 199 | + package. Register as MDS036 in category |
| 200 | + `meta`, enabled by default, `FixableRule`. |
| 201 | +2. Add MDS036 fixtures under |
| 202 | + `internal/rules/MDS036-toc/`: `good/` with a |
| 203 | + correct body, `bad/` with a stale body to |
| 204 | + verify Check, `fixed/` with the expected |
| 205 | + output after Fix. Cover default parameters, |
| 206 | + custom `min-level`/`max-level`, single-level |
| 207 | + docs, and deeply nested structures. |
| 208 | +3. Upgrade MDS035 to `FixableRule`: replace |
| 209 | + matched directive lines with |
| 210 | + `<?toc?>\n<?/toc?>` blocks preserving |
| 211 | + surrounding blank lines; leave `[TOC]` |
| 212 | + untouched when a link-ref definition |
| 213 | + suppresses Check; add `fixed/` fixtures for |
| 214 | + each of the four variants. |
| 215 | +4. Update MDS035's diagnostic message to point |
| 216 | + at `<?toc?>` (MDS036) instead of |
| 217 | + `<?catalog?>` (MDS019); wording: |
| 218 | + `unsupported TOC directive \`{token}\`; use \`<?toc?>\` (MDS036)` |
| 219 | +5. Update MDS035 README (message + examples), |
| 220 | + plan 88 status/deviation note, and the |
| 221 | + renderer-portability section in |
| 222 | + [docs/background/markdown-linters.md][lnt]. |
| 223 | +6. Update the generated-section archetype doc |
| 224 | + to list `toc` alongside `catalog` and |
| 225 | + `include`. |
| 226 | +7. Verify multi-pass `mdsmith fix` end-to-end: |
| 227 | + starting from a document containing `[TOC]`, |
| 228 | + a single `fix` run must yield a populated |
| 229 | + `<?toc?>...<?/toc?>` block. Add an |
| 230 | + integration test asserting this. |
| 231 | + |
| 232 | +[lnt]: ../docs/background/markdown-linters.md |
| 233 | + |
| 234 | +## Acceptance Criteria |
| 235 | + |
| 236 | +- [ ] `<?toc?>...<?/toc?>` in a document |
| 237 | + regenerates on `mdsmith fix` with a |
| 238 | + nested list of headings linked to |
| 239 | + GitHub-style slugs |
| 240 | +- [ ] `min-level` and `max-level` parameters |
| 241 | + gate which headings appear |
| 242 | +- [ ] `<?toc?>` with a stale body produces |
| 243 | + an MDS036 diagnostic on `check` |
| 244 | +- [ ] MDS035 `fix` rewrites `[TOC]`, |
| 245 | + `[[_TOC_]]`, `[[toc]]`, and `${toc}` |
| 246 | + on their own line to |
| 247 | + `<?toc?>\n<?/toc?>` blocks |
| 248 | +- [ ] MDS035 `fix` leaves `[TOC]` untouched |
| 249 | + when a matching link reference |
| 250 | + definition is present |
| 251 | +- [ ] A single `mdsmith fix` run converts a |
| 252 | + source containing `[TOC]` into a source |
| 253 | + containing a populated `<?toc?>` block |
| 254 | +- [ ] Merge driver regenerates `<?toc?>` |
| 255 | + bodies on conflict, same as `<?catalog?>` |
| 256 | +- [ ] MDS035 diagnostic message names |
| 257 | + `<?toc?>` (MDS036) as the replacement |
| 258 | +- [ ] All tests pass: `go test ./...` |
| 259 | +- [ ] `go tool golangci-lint run` reports |
| 260 | + no issues |
0 commit comments