| 1 | // markdownPipeline — the isomorphic Markdown→HAST parse pipeline (Phase E of |
| 2 | // the history refactor). It runs the EXACT processor chain react-markdown uses |
| 3 | // in MarkdownRenderer, factored into a side-effect-free function that runs both |
| 4 | // on the main thread (jsdom/tests, Worker-less fallback) and inside |
| 5 | // markdown.worker.ts: |
| 6 | // |
| 7 | // normalizeMath pre-pass |
| 8 | // → remark-parse + remarkGfm + remarkMath + remarkMathPolicy + remarkLocalPathLinks |
| 9 | // → remark-rehype (allowDangerousHtml, same as react-markdown's default) |
| 10 | // → rehypeReasonixKatex |
| 11 | // → react-markdown's post-transform (raw → text, urlTransform on URL attrs) |
| 12 | // |
| 13 | // The document is parsed WHOLE so definitions, footnotes, and reference links |
| 14 | // resolve globally; sliceHastBlocks then cuts the resulting HAST top-level |
| 15 | // children into renderable blocks (the remark-rehype footnote section stays its |
| 16 | // own trailing block). Concatenating the rendered blocks is byte-identical to |
| 17 | // rendering the unsliced root. |
| 18 | // |
| 19 | // This module must stay free of DOM/React imports so the inline worker bundle |
| 20 | // contains no component code. |
| 21 | |
| 22 | import { unified } from "unified"; |
| 23 | import { VFile } from "vfile"; |
| 24 | import remarkParse from "remark-parse"; |
| 25 | import remarkRehype from "remark-rehype"; |
| 26 | import { urlAttributes } from "html-url-attributes"; |
| 27 | import { visit } from "unist-util-visit"; |
| 28 | import type { Element as HastElement, Root as HastRoot, RootContent as HastRootContent } from "hast"; |
| 29 | import { normalizeMath } from "../components/mathNormalize"; |
| 30 | import { reasonixRemarkPlugins } from "../components/markdownRemarkPlugins"; |
| 31 | import { reasonixRehypePlugins } from "../components/rehypeReasonixKatex"; |
| 32 | import { |
| 33 | extractLargePlainMarkdownTables, |
| 34 | type VirtualMarkdownTableData, |
| 35 | } from "./largeMarkdownTable"; |
| 36 | import { isLocalFileHref } from "./localFileUrl"; |
| 37 | import { contentRevision } from "./contentRevision"; |
| 38 | import { markdownSelectionTextFromBlocks } from "./markdownSelectionProjection"; |
| 39 | export { estimateHastBytes } from "./markdownByteEstimate"; |
| 40 | |
| 41 | export type { HastRoot, HastRootContent }; |
| 42 | |
| 43 | export interface MarkdownBlock { |
| 44 | /** Stable within one parse: the block's top-level index. */ |
| 45 | key: string; |
| 46 | children: HastRootContent[]; |
| 47 | /** Lightweight representation for a large table with plain-text cells. */ |
| 48 | virtualTable?: VirtualMarkdownTableData; |
| 49 | /** |
| 50 | * Content identity of this block, computed by the parse that produced it. |
| 51 | * Two blocks from two parses are the same block when their keys and |
| 52 | * fingerprints match, which lets the render path keep the previous AST |
| 53 | * object (and therefore native selection and disclosure state) without |
| 54 | * serializing either tree. The value is opaque: never interpret it. |
| 55 | */ |
| 56 | fingerprint: number; |
| 57 | /** Parsed HAST element count used to bound progressive DOM publication. */ |
| 58 | elementCount?: number; |
| 59 | } |
| 60 | |
| 61 | export interface MarkdownParseResult { |
| 62 | blocks: MarkdownBlock[]; |
| 63 | selectionText: string; |
| 64 | selectionRevision: number; |
| 65 | } |
| 66 | |
| 67 | const SAFE_PROTOCOL_RE = /^(https?|ircs?|mailto|xmpp)$/i; |
| 68 | |
| 69 | // Mirror of react-markdown's exported defaultUrlTransform (which itself mirrors |
| 70 | // micromark-util-sanitize-uri without the encode pass). Copied so the worker |
| 71 | // bundle does not pull React in through react-markdown; the pipeline parity |
| 72 | // test pins this copy against the original over a URL corpus. |
| 73 | export function defaultMarkdownUrlTransform(value: string): string { |
| 74 | const colon = value.indexOf(":"); |
| 75 | const questionMark = value.indexOf("?"); |
| 76 | const numberSign = value.indexOf("#"); |
| 77 | const slash = value.indexOf("/"); |
| 78 | |
| 79 | if ( |
| 80 | // If there is no protocol, it’s relative. |
| 81 | colon === -1 |
| 82 | // If the first colon is after a `?`, `#`, or `/`, it’s not a protocol. |
| 83 | || (slash !== -1 && colon > slash) |
| 84 | || (questionMark !== -1 && colon > questionMark) |
| 85 | || (numberSign !== -1 && colon > numberSign) |
| 86 | // It is a protocol, it should be allowed. |
| 87 | || SAFE_PROTOCOL_RE.test(value.slice(0, colon)) |
| 88 | ) { |
| 89 | return value; |
| 90 | } |
| 91 | |
| 92 | return ""; |
| 93 | } |
| 94 | |
| 95 | // Local file hrefs come from local-path linkification (remarkLocalPathLinks) |
| 96 | // or explicit Markdown links and must survive URL sanitisation, which would |
| 97 | // otherwise blank them along with javascript: and friends. |
| 98 | export function markdownUrlTransform(value: string): string { |
| 99 | return isLocalFileHref(value) ? value : defaultMarkdownUrlTransform(value); |
| 100 | } |
| 101 | |
| 102 | // Images use a separate protocol policy because their bytes are resolved and |
| 103 | // validated by ResolveMarkdownImageForTab before reaching the WebView. Ordinary |
| 104 | // links intentionally continue to reject data: URLs. |
| 105 | export function markdownImageUrlTransform(value: string): string { |
| 106 | if (/^data:image\/(?:png|jpeg|gif|webp)(?:;base64)?,/i.test(value)) return value; |
| 107 | return isLocalFileHref(value) ? value : defaultMarkdownUrlTransform(value); |
| 108 | } |
| 109 | |
| 110 | // The same transform react-markdown applies between the rehype plugins and |
| 111 | // hast-util-to-jsx-runtime (its internal `post()` visitor): raw HTML nodes |
| 112 | // become text (skipHtml is false), and every URL attribute passes through |
| 113 | // urlTransform. applied here so worker output arrives render-ready. |
| 114 | function applyReactMarkdownTransforms(tree: HastRoot): void { |
| 115 | visit(tree, (node, index, parent) => { |
| 116 | if ((node.type as string) === "raw" && parent && typeof index === "number") { |
| 117 | parent.children[index] = { type: "text", value: (node as unknown as { value: string }).value }; |
| 118 | return index; |
| 119 | } |
| 120 | if (node.type === "element") { |
| 121 | const element = node as HastElement; |
| 122 | for (const key in urlAttributes) { |
| 123 | const hasAttr = Object.prototype.hasOwnProperty.call(urlAttributes, key) |
| 124 | && Object.prototype.hasOwnProperty.call(element.properties, key); |
| 125 | if (!hasAttr) continue; |
| 126 | const value = element.properties[key]; |
| 127 | const test = urlAttributes[key as keyof typeof urlAttributes]; |
| 128 | if (test === null || (test as readonly string[]).includes(element.tagName)) { |
| 129 | element.properties[key] = element.tagName === "img" && key === "src" |
| 130 | ? markdownImageUrlTransform(String(value || "")) |
| 131 | : markdownUrlTransform(String(value || "")); |
| 132 | } |
| 133 | } |
| 134 | } |
| 135 | return undefined; |
| 136 | }); |
| 137 | } |
| 138 | |
| 139 | /** |
| 140 | * Parse markdown text into a render-ready HAST root using the exact chain the |
| 141 | * main-thread renderer uses. Synchronous (unified runSync), DOM-free, safe to |
| 142 | * run inside a Web Worker. |
| 143 | */ |
| 144 | export function parseMarkdownToHast(text: string): HastRoot { |
| 145 | return parseNormalizedMarkdownToHast(normalizeMath(text)); |
| 146 | } |
| 147 | |
| 148 | function parseNormalizedMarkdownToHast(normalized: string): HastRoot { |
| 149 | const processor = unified() |
| 150 | .use(remarkParse) |
| 151 | .use(reasonixRemarkPlugins) |
| 152 | .use(remarkRehype, { allowDangerousHtml: true }) |
| 153 | .use(reasonixRehypePlugins); |
| 154 | // The same VFile must flow through parse and runSync: remarkMathPolicy |
| 155 | // slices original math sources out of file.value by node position. |
| 156 | const file = new VFile({ value: normalized }); |
| 157 | const tree = processor.runSync(processor.parse(file), file) as unknown as HastRoot; |
| 158 | applyReactMarkdownTransforms(tree); |
| 159 | return tree; |
| 160 | } |
| 161 | |
| 162 | const VIRTUAL_TABLE_TAG = "reasonix-virtual-table"; |
| 163 | |
| 164 | function injectVirtualTablePlaceholders( |
| 165 | root: HastRoot, |
| 166 | markerPrefix: string, |
| 167 | tableCount: number, |
| 168 | ): void { |
| 169 | const walk = (node: HastRoot | HastElement): void => { |
| 170 | for (let index = 0; index < node.children.length; index += 1) { |
| 171 | const child = node.children[index]; |
| 172 | if (child.type === "element") { |
| 173 | const paragraph = child as HastElement; |
| 174 | const only = paragraph.tagName === "p" && paragraph.children.length === 1 |
| 175 | ? paragraph.children[0] |
| 176 | : undefined; |
| 177 | const value = only?.type === "text" ? only.value : ""; |
| 178 | if (value.startsWith(markerPrefix)) { |
| 179 | const tableIndex = Number(value.slice(markerPrefix.length)); |
| 180 | if (Number.isInteger(tableIndex) && tableIndex >= 0 && tableIndex < tableCount) { |
| 181 | node.children[index] = { |
| 182 | type: "element", |
| 183 | tagName: VIRTUAL_TABLE_TAG, |
| 184 | properties: { dataReasonixTableIndex: tableIndex }, |
| 185 | children: [], |
| 186 | }; |
| 187 | continue; |
| 188 | } |
| 189 | } |
| 190 | walk(paragraph); |
| 191 | } |
| 192 | } |
| 193 | }; |
| 194 | walk(root); |
| 195 | } |
| 196 | |
| 197 | /** |
| 198 | * Slice a parsed HAST root into top-level blocks. Each element child starts a |
| 199 | * new block; interstitial whitespace text attaches to the FOLLOWING block so |
| 200 | * concatenating the rendered blocks reproduces the unsliced render byte for |
| 201 | * byte. Whole-document parsing has already resolved footnote/reference |
| 202 | * definitions, and remark-rehype's footnote section arrives as a trailing |
| 203 | * top-level element, so it becomes its own trailing block. |
| 204 | */ |
| 205 | /** A block before its parse has stamped a fingerprint. */ |
| 206 | export type UnfingerprintedBlock = Omit<MarkdownBlock, "fingerprint">; |
| 207 | |
| 208 | export function sliceHastBlocks(root: HastRoot): UnfingerprintedBlock[] { |
| 209 | const blocks: UnfingerprintedBlock[] = []; |
| 210 | let pending: HastRootContent[] = []; |
| 211 | for (const child of root.children) { |
| 212 | if (child.type === "element") { |
| 213 | blocks.push({ key: `b${blocks.length}`, children: [...pending, child] }); |
| 214 | pending = []; |
| 215 | } else if (blocks.length === 0) { |
| 216 | pending.push(child); |
| 217 | } else { |
| 218 | blocks[blocks.length - 1].children.push(child); |
| 219 | } |
| 220 | } |
| 221 | if (pending.length > 0) { |
| 222 | if (blocks.length === 0) blocks.push({ key: "b0", children: pending }); |
| 223 | else blocks[blocks.length - 1].children.push(...pending); |
| 224 | } |
| 225 | return blocks; |
| 226 | } |
| 227 | |
| 228 | /** Parse + slice in one call (the worker entry point). */ |
| 229 | export function parseMarkdownToBlocks(text: string): MarkdownBlock[] { |
| 230 | const normalized = normalizeMath(text); |
| 231 | const extracted = extractLargePlainMarkdownTables(normalized); |
| 232 | if (extracted.tables.length === 0) return fingerprintBlocks(sliceHastBlocks(parseNormalizedMarkdownToHast(normalized))); |
| 233 | |
| 234 | const processor = unified() |
| 235 | .use(remarkParse) |
| 236 | .use(reasonixRemarkPlugins) |
| 237 | .use(remarkRehype, { allowDangerousHtml: true }) |
| 238 | .use(reasonixRehypePlugins); |
| 239 | const file = new VFile({ value: extracted.text }); |
| 240 | const tree = processor.runSync(processor.parse(file), file) as unknown as HastRoot; |
| 241 | applyReactMarkdownTransforms(tree); |
| 242 | injectVirtualTablePlaceholders(tree, extracted.markerPrefix, extracted.tables.length); |
| 243 | |
| 244 | return fingerprintBlocks(sliceHastBlocks(tree).map((block) => { |
| 245 | const placeholder = block.children.find( |
| 246 | (child): child is HastElement => child.type === "element" && child.tagName === VIRTUAL_TABLE_TAG, |
| 247 | ); |
| 248 | if (!placeholder) return block; |
| 249 | const index = Number(placeholder.properties.dataReasonixTableIndex); |
| 250 | return { |
| 251 | ...block, |
| 252 | children: block.children.filter((child) => child !== placeholder), |
| 253 | virtualTable: extracted.tables[index], |
| 254 | }; |
| 255 | })); |
| 256 | } |
| 257 | |
| 258 | // 32-bit FNV-1a over a value's JSON-visible shape. Property order is sorted |
| 259 | // rather than preserved: both sides of every comparison come from this |
| 260 | // function, so what matters is that equal shapes hash equal and different |
| 261 | // shapes almost never collide — not that the walk matches JSON.stringify |
| 262 | // byte for byte. Values JSON.stringify would drop (`undefined`, functions) |
| 263 | // are skipped here for the same reason. |
| 264 | function hashValue(hash: number, value: unknown, depth: number): number { |
| 265 | if (depth > 64) return hash; |
| 266 | if (value === null) return hashText(hash, "null"); |
| 267 | switch (typeof value) { |
| 268 | case "string": return hashText(hash, value); |
| 269 | case "number": return hashText(hash, String(value)); |
| 270 | case "boolean": return hashText(hash, value ? "t" : "f"); |
| 271 | case "undefined": |
| 272 | case "function": |
| 273 | return hash; |
| 274 | case "object": { |
| 275 | if (Array.isArray(value)) { |
| 276 | let next = hashText(hash, "["); |
| 277 | for (const item of value) next = hashValue(next, item, depth + 1); |
| 278 | return hashText(next, "]"); |
| 279 | } |
| 280 | let next = hashText(hash, "{"); |
| 281 | const record = value as Record<string, unknown>; |
| 282 | for (const key of Object.keys(record).sort()) { |
| 283 | next = hashValue(hashText(next, key), record[key], depth + 1); |
| 284 | } |
| 285 | return hashText(next, "}"); |
| 286 | } |
| 287 | default: |
| 288 | return hash; |
| 289 | } |
| 290 | } |
| 291 | |
| 292 | function hashText(hash: number, text: string): number { |
| 293 | let next = hash; |
| 294 | for (let index = 0; index < text.length; index += 1) { |
| 295 | next ^= text.charCodeAt(index); |
| 296 | next = Math.imul(next, 0x01000193) >>> 0; |
| 297 | } |
| 298 | // A separator so `["ab"]` and `["a","b"]` cannot hash alike. |
| 299 | next ^= 0x1f; |
| 300 | return Math.imul(next, 0x01000193) >>> 0; |
| 301 | } |
| 302 | |
| 303 | /** Stamps every block with its content fingerprint. Call once, on final blocks. */ |
| 304 | export function fingerprintBlocks(blocks: UnfingerprintedBlock[]): MarkdownBlock[] { |
| 305 | for (const block of blocks) { |
| 306 | let hash = hashText(0x811c9dc5, block.key); |
| 307 | for (const child of block.children) hash = hashValue(hash, child, 0); |
| 308 | if (block.virtualTable) hash = hashValue(hash, block.virtualTable, 0); |
| 309 | (block as MarkdownBlock).fingerprint = hash; |
| 310 | (block as MarkdownBlock).elementCount = countHastElements(block.children); |
| 311 | } |
| 312 | return blocks as MarkdownBlock[]; |
| 313 | } |
| 314 | |
| 315 | function countHastElements(children: HastRootContent[]): number { |
| 316 | let count = 0; |
| 317 | const visitChildren = (nodes: HastRootContent[]): void => { |
| 318 | for (const node of nodes) { |
| 319 | if (node.type !== "element") continue; |
| 320 | count += 1; |
| 321 | visitChildren(node.children as HastRootContent[]); |
| 322 | } |
| 323 | }; |
| 324 | visitChildren(children); |
| 325 | return count; |
| 326 | } |
| 327 | |
| 328 | /** Parse once and derive both the render tree and copy projection. */ |
| 329 | export function parseMarkdown(text: string): MarkdownParseResult { |
| 330 | const blocks = parseMarkdownToBlocks(text); |
| 331 | const selectionText = markdownSelectionTextFromBlocks(blocks); |
| 332 | return { |
| 333 | blocks, |
| 334 | selectionText, |
| 335 | selectionRevision: contentRevision(selectionText), |
| 336 | }; |
| 337 | } |
| 338 | |
| 339 | /** |
| 340 | * Content-derived cache revision for the transcript markdown cache: an FNV-1a |
| 341 | * fingerprint of the source text. Cache entries also store the source itself, |
| 342 | * so a (practically impossible) hash collision is caught by comparison. |
| 343 | */ |
| 344 | export function markdownContentRevision(text: string): number { |
| 345 | return contentRevision(text); |
| 346 | } |
| 347 |