okf-catalog 0.1.0

package/dist/bundle/markdown.js

@@ -0,0 +1,210 @@
+import { fromMarkdown } from "mdast-util-from-markdown";
+import { gfmFootnoteFromMarkdown } from "mdast-util-gfm-footnote";
+import { gfmTableFromMarkdown } from "mdast-util-gfm-table";
+import { gfmFootnote } from "micromark-extension-gfm-footnote";
+import { gfmTable } from "micromark-extension-gfm-table";
+const SENTENCE_CAP = 200;
+export const ANALYSIS_BUDGET = 256 * 1024;
+const MAX_NESTING = 256;
+const MAX_EMPHASIS_RUNS = 2000;
+const MAX_DEFINITIONS = 1000;
+const SCRIPT_LIKE = /<\s*\/?\s*(script|style|iframe)\b/i;
+const DELIMITER_RUN = /[*_]+/g;
+const DEFINITION_LINE = /^ {0,3}\[[^\]]{1,999}\]:/;
+/** An emphasis delimiter run that could open or close: one not wedged between two word characters. */
+function isFlankingRun(text, start, end) {
+    const before = start === 0 ? " " : (text[start - 1] ?? " ");
+    const after = end >= text.length ? " " : (text[end] ?? " ");
+    const wordy = (c) => /[\p{L}\p{N}]/u.test(c);
+    return !(wordy(before) && wordy(after));
+}
+/**
+ * Why a body must not be parsed, or `undefined` when it may be. The parser's cost grows with the square of the
+ * number of emphasis delimiter runs and of link reference definitions, and its stack with nesting depth, so
+ * each is counted before any parse. The counts are deterministic and cheap.
+ */
+export function analysisBounds(text) {
+    if (nestingDepth(text) > MAX_NESTING)
+        return "nesting";
+    let runs = 0;
+    for (const match of text.matchAll(DELIMITER_RUN)) {
+        if (isFlankingRun(text, match.index, match.index + match[0].length))
+            runs += 1;
+        if (runs > MAX_EMPHASIS_RUNS)
+            return "emphasis";
+    }
+    let definitions = 0;
+    for (const line of text.split("\n")) {
+        if (DEFINITION_LINE.test(line))
+            definitions += 1;
+        if (definitions > MAX_DEFINITIONS)
+            return "definitions";
+    }
+    return undefined;
+}
+/** The deepest run of block-quote markers or unclosed brackets, measured before any parse so the bound is deterministic. */
+export function nestingDepth(body) {
+    let deepest = 0;
+    let brackets = 0;
+    for (const line of body.split("\n")) {
+        const quotes = (/^[ \t>]*/.exec(line)?.[0].match(/>/g) ?? []).length;
+        if (quotes > deepest)
+            deepest = quotes;
+        for (const ch of line) {
+            if (ch === "[")
+                brackets += 1;
+            else if (ch === "]" && brackets > 0)
+                brackets -= 1;
+            if (brackets > deepest)
+                deepest = brackets;
+        }
+    }
+    return deepest;
+}
+function isParent(node) {
+    return "children" in node && Array.isArray(node.children);
+}
+/** The prose of a node, leaving out HTML and footnote marks, whitespace collapsed. */
+function proseOf(node) {
+    const parts = [];
+    const stack = [node];
+    while (stack.length > 0) {
+        const current = stack.pop();
+        if (current.type === "html" || current.type === "footnoteReference")
+            continue;
+        if ("value" in current && typeof current.value === "string")
+            parts.push(current.value);
+        else if (isParent(current))
+            for (let i = current.children.length - 1; i >= 0; i--)
+                stack.push(current.children[i]);
+    }
+    return parts.join("").replace(/\s+/g, " ").trim();
+}
+const BLOCKS = new Set([
+    "paragraph",
+    "heading",
+    "blockquote",
+    "list",
+    "listItem",
+    "code",
+    "table",
+    "tableRow",
+    "tableCell",
+    "thematicBreak",
+    "footnoteDefinition",
+]);
+/** The whole tree's prose, one space between blocks, HTML and footnote marks left out, whitespace collapsed. */
+function proseWithBlocks(tree) {
+    const parts = [];
+    const stack = [tree];
+    while (stack.length > 0) {
+        const current = stack.pop();
+        if (current.type === "html" || current.type === "footnoteReference")
+            continue;
+        if (BLOCKS.has(current.type))
+            parts.push(" ");
+        if ("value" in current && typeof current.value === "string")
+            parts.push(current.value);
+        else if (isParent(current)) {
+            for (let i = current.children.length - 1; i >= 0; i--)
+                stack.push(current.children[i]);
+        }
+        if (BLOCKS.has(current.type))
+            parts.push(" ");
+    }
+    return parts.join("").replace(/\s+/g, " ").trim();
+}
+function firstSentenceOf(text) {
+    const match = /^(.*?[.!?])(?=\s|$)/.exec(text);
+    const sentence = match?.[1] ?? text;
+    return sentence.length > SENTENCE_CAP ? sentence.slice(0, SENTENCE_CAP).trimEnd() : sentence;
+}
+const EMPTY = {
+    firstHeading: undefined,
+    firstSentence: undefined,
+    links: [],
+    footnoteReferences: [],
+    htmlBlocks: 0,
+    inlineHtml: 0,
+    hasScriptLike: false,
+};
+/**
+ * Reads the facts the OKF layer needs from a Markdown body: CommonMark with footnotes and tables, never the
+ * autolink extension, whose cost grows without bound on nested brackets. A body nested past the limit is left
+ * unanalysed rather than parsed; a body past the budget is analysed up to it. The tree is walked iteratively.
+ */
+export function readBody(body) {
+    const truncated = body.length > ANALYSIS_BUDGET;
+    const text = truncated ? body.slice(0, ANALYSIS_BUDGET) : body;
+    if (analysisBounds(text) !== undefined)
+        return { ...EMPTY, links: [], footnoteReferences: [], unanalysed: true, truncated };
+    const tree = fromMarkdown(text, {
+        extensions: [gfmFootnote(), gfmTable()],
+        mdastExtensions: [gfmFootnoteFromMarkdown(), gfmTableFromMarkdown()],
+    });
+    const facts = {
+        ...EMPTY,
+        links: [],
+        footnoteReferences: [],
+        unanalysed: false,
+        truncated,
+    };
+    const definitions = new Map();
+    const references = [];
+    const stack = [
+        { node: tree, parent: undefined, skipped: false },
+    ];
+    while (stack.length > 0) {
+        const { node, parent, skipped } = stack.pop();
+        switch (node.type) {
+            case "definition":
+                definitions.set(node.identifier, node.url);
+                break;
+            case "heading":
+                if (facts.firstHeading === undefined)
+                    facts.firstHeading = proseOf(node);
+                break;
+            case "paragraph":
+                if (!skipped && facts.firstSentence === undefined) {
+                    const sentence = firstSentenceOf(proseOf(node));
+                    if (sentence.length > 0)
+                        facts.firstSentence = sentence;
+                }
+                break;
+            case "link":
+                facts.links.push({ url: node.url, text: proseOf(node) });
+                break;
+            case "linkReference":
+                references.push({ identifier: node.identifier, slot: facts.links.length });
+                facts.links.push({ url: "", text: proseOf(node) });
+                break;
+            case "footnoteReference":
+                facts.footnoteReferences.push(node.identifier);
+                break;
+            case "html":
+                if (parent !== undefined && parent.type === "paragraph")
+                    facts.inlineHtml += 1;
+                else
+                    facts.htmlBlocks += 1;
+                if (SCRIPT_LIKE.test(node.value))
+                    facts.hasScriptLike = true;
+                break;
+            default:
+                break;
+        }
+        if (isParent(node)) {
+            const skip = skipped || node.type === "footnoteDefinition" || node.type === "table";
+            for (let i = node.children.length - 1; i >= 0; i--)
+                stack.push({ node: node.children[i], parent: node, skipped: skip });
+        }
+    }
+    for (const ref of references) {
+        const slot = facts.links[ref.slot];
+        if (slot !== undefined)
+            slot.url = definitions.get(ref.identifier) ?? "";
+    }
+    facts.links = facts.links.filter((l) => l.url !== "");
+    facts.prose = proseWithBlocks(tree);
+    return facts;
+}
+//# sourceMappingURL=markdown.js.map

package/dist/bundle/markdown.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"markdown.js","sourceRoot":"","sources":["../../src/bundle/markdown.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,YAAY,EAAE,MAAM,0BAA0B,CAAC;AACxD,OAAO,EAAE,uBAAuB,EAAE,MAAM,yBAAyB,CAAC;AAClE,OAAO,EAAE,oBAAoB,EAAE,MAAM,sBAAsB,CAAC;AAC5D,OAAO,EAAE,WAAW,EAAE,MAAM,kCAAkC,CAAC;AAC/D,OAAO,EAAE,QAAQ,EAAE,MAAM,+BAA+B,CAAC;AAkBzD,MAAM,YAAY,GAAG,GAAG,CAAC;AACzB,MAAM,CAAC,MAAM,eAAe,GAAG,GAAG,GAAG,IAAI,CAAC;AAC1C,MAAM,WAAW,GAAG,GAAG,CAAC;AACxB,MAAM,iBAAiB,GAAG,IAAI,CAAC;AAC/B,MAAM,eAAe,GAAG,IAAI,CAAC;AAC7B,MAAM,WAAW,GAAG,oCAAoC,CAAC;AACzD,MAAM,aAAa,GAAG,QAAQ,CAAC;AAC/B,MAAM,eAAe,GAAG,0BAA0B,CAAC;AAEnD,sGAAsG;AACtG,SAAS,aAAa,CAAC,IAAY,EAAE,KAAa,EAAE,GAAW;IAC7D,MAAM,MAAM,GAAG,KAAK,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,GAAG,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC;IAC5D,MAAM,KAAK,GAAG,GAAG,IAAI,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,GAAG,CAAC,CAAC;IAC5D,MAAM,KAAK,GAAG,CAAC,CAAS,EAAW,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IAC9D,OAAO,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC;AAC1C,CAAC;AAED;;;;GAIG;AACH,MAAM,UAAU,cAAc,CAAC,IAAY;IACzC,IAAI,YAAY,CAAC,IAAI,CAAC,GAAG,WAAW;QAAE,OAAO,SAAS,CAAC;IACvD,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,QAAQ,CAAC,aAAa,CAAC,EAAE,CAAC;QACjD,IAAI,aAAa,CAAC,IAAI,EAAE,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC;YAAE,IAAI,IAAI,CAAC,CAAC;QAC/E,IAAI,IAAI,GAAG,iBAAiB;YAAE,OAAO,UAAU,CAAC;IAClD,CAAC;IACD,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACpC,IAAI,eAAe,CAAC,IAAI,CAAC,IAAI,CAAC;YAAE,WAAW,IAAI,CAAC,CAAC;QACjD,IAAI,WAAW,GAAG,eAAe;YAAE,OAAO,aAAa,CAAC;IAC1D,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED,4HAA4H;AAC5H,MAAM,UAAU,YAAY,CAAC,IAAY;IACvC,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,IAAI,QAAQ,GAAG,CAAC,CAAC;IACjB,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACpC,MAAM,MAAM,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC;QACrE,IAAI,MAAM,GAAG,OAAO;YAAE,OAAO,GAAG,MAAM,CAAC;QACvC,KAAK,MAAM,EAAE,IAAI,IAAI,EAAE,CAAC;YACtB,IAAI,EAAE,KAAK,GAAG;gBAAE,QAAQ,IAAI,CAAC,CAAC;iBACzB,IAAI,EAAE,KAAK,GAAG,IAAI,QAAQ,GAAG,CAAC;gBAAE,QAAQ,IAAI,CAAC,CAAC;YACnD,IAAI,QAAQ,GAAG,OAAO;gBAAE,OAAO,GAAG,QAAQ,CAAC;QAC7C,CAAC;IACH,CAAC;IACD,OAAO,OAAO,CAAC;AACjB,CAAC;AAED,SAAS,QAAQ,CAAC,IAAW;IAC3B,OAAO,UAAU,IAAI,IAAI,IAAI,KAAK,CAAC,OAAO,CAAE,IAAe,CAAC,QAAQ,CAAC,CAAC;AACxE,CAAC;AAED,sFAAsF;AACtF,SAAS,OAAO,CAAC,IAAW;IAC1B,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAY,CAAC,IAAI,CAAC,CAAC;IAC9B,OAAO,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACxB,MAAM,OAAO,GAAG,KAAK,CAAC,GAAG,EAAW,CAAC;QACrC,IAAI,OAAO,CAAC,IAAI,KAAK,MAAM,IAAI,OAAO,CAAC,IAAI,KAAK,mBAAmB;YAAE,SAAS;QAC9E,IAAI,OAAO,IAAI,OAAO,IAAI,OAAO,OAAO,CAAC,KAAK,KAAK,QAAQ;YAAE,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;aAClF,IAAI,QAAQ,CAAC,OAAO,CAAC;YACxB,KAAK,IAAI,CAAC,GAAG,OAAO,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE;gBACnD,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAU,CAAC,CAAC;IAC/C,CAAC;IACD,OAAO,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;AACpD,CAAC;AAED,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC;IACrB,WAAW;IACX,SAAS;IACT,YAAY;IACZ,MAAM;IACN,UAAU;IACV,MAAM;IACN,OAAO;IACP,UAAU;IACV,WAAW;IACX,eAAe;IACf,oBAAoB;CACrB,CAAC,CAAC;AAEH,gHAAgH;AAChH,SAAS,eAAe,CAAC,IAAW;IAClC,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAY,CAAC,IAAI,CAAC,CAAC;IAC9B,OAAO,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACxB,MAAM,OAAO,GAAG,KAAK,CAAC,GAAG,EAAW,CAAC;QACrC,IAAI,OAAO,CAAC,IAAI,KAAK,MAAM,IAAI,OAAO,CAAC,IAAI,KAAK,mBAAmB;YAAE,SAAS;QAC9E,IAAI,MAAM,CAAC,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC;YAAE,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;QAC9C,IAAI,OAAO,IAAI,OAAO,IAAI,OAAO,OAAO,CAAC,KAAK,KAAK,QAAQ;YAAE,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,CAAC;aAClF,IAAI,QAAQ,CAAC,OAAO,CAAC,EAAE,CAAC;YAC3B,KAAK,IAAI,CAAC,GAAG,OAAO,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE;gBACnD,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC,CAAU,CAAC,CAAC;QAC7C,CAAC;QACD,IAAI,MAAM,CAAC,GAAG,CAAC,OAAO,CAAC,IAAI,CAAC;YAAE,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;IAChD,CAAC;IACD,OAAO,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;AACpD,CAAC;AAED,SAAS,eAAe,CAAC,IAAY;IACnC,MAAM,KAAK,GAAG,qBAAqB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC/C,MAAM,QAAQ,GAAG,KAAK,EAAE,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC;IACpC,OAAO,QAAQ,CAAC,MAAM,GAAG,YAAY,CAAC,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,EAAE,YAAY,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC;AAC/F,CAAC;AAED,MAAM,KAAK,GAAgD;IACzD,YAAY,EAAE,SAAS;IACvB,aAAa,EAAE,SAAS;IACxB,KAAK,EAAE,EAAE;IACT,kBAAkB,EAAE,EAAE;IACtB,UAAU,EAAE,CAAC;IACb,UAAU,EAAE,CAAC;IACb,aAAa,EAAE,KAAK;CACrB,CAAC;AAEF;;;;GAIG;AACH,MAAM,UAAU,QAAQ,CAAC,IAAY;IACnC,MAAM,SAAS,GAAG,IAAI,CAAC,MAAM,GAAG,eAAe,CAAC;IAChD,MAAM,IAAI,GAAG,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,eAAe,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;IAC/D,IAAI,cAAc,CAAC,IAAI,CAAC,KAAK,SAAS;QACpC,OAAO,EAAE,GAAG,KAAK,EAAE,KAAK,EAAE,EAAE,EAAE,kBAAkB,EAAE,EAAE,EAAE,UAAU,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;IACtF,MAAM,IAAI,GAAG,YAAY,CAAC,IAAI,EAAE;QAC9B,UAAU,EAAE,CAAC,WAAW,EAAE,EAAE,QAAQ,EAAE,CAAC;QACvC,eAAe,EAAE,CAAC,uBAAuB,EAAE,EAAE,oBAAoB,EAAE,CAAC;KACrE,CAAC,CAAC;IACH,MAAM,KAAK,GAAc;QACvB,GAAG,KAAK;QACR,KAAK,EAAE,EAAE;QACT,kBAAkB,EAAE,EAAE;QACtB,UAAU,EAAE,KAAK;QACjB,SAAS;KACV,CAAC;IACF,MAAM,WAAW,GAAG,IAAI,GAAG,EAAkB,CAAC;IAC9C,MAAM,UAAU,GAAgD,EAAE,CAAC;IACnE,MAAM,KAAK,GAAwE;QACjF,EAAE,IAAI,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,OAAO,EAAE,KAAK,EAAE;KAClD,CAAC;IACF,OAAO,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACxB,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,GAAG,KAAK,CAAC,GAAG,EAI1C,CAAC;QACF,QAAQ,IAAI,CAAC,IAAI,EAAE,CAAC;YAClB,KAAK,YAAY;gBACf,WAAW,CAAC,GAAG,CAAC,IAAI,CAAC,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC;gBAC3C,MAAM;YACR,KAAK,SAAS;gBACZ,IAAI,KAAK,CAAC,YAAY,KAAK,SAAS;oBAAE,KAAK,CAAC,YAAY,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;gBACzE,MAAM;YACR,KAAK,WAAW;gBACd,IAAI,CAAC,OAAO,IAAI,KAAK,CAAC,aAAa,KAAK,SAAS,EAAE,CAAC;oBAClD,MAAM,QAAQ,GAAG,eAAe,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;oBAChD,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC;wBAAE,KAAK,CAAC,aAAa,GAAG,QAAQ,CAAC;gBAC1D,CAAC;gBACD,MAAM;YACR,KAAK,MAAM;gBACT,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,EAAE,IAAI,CAAC,GAAG,EAAE,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;gBACzD,MAAM;YACR,KAAK,eAAe;gBAClB,UAAU,CAAC,IAAI,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,EAAE,IAAI,EAAE,KAAK,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC;gBAC3E,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,EAAE,EAAE,EAAE,IAAI,EAAE,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;gBACnD,MAAM;YACR,KAAK,mBAAmB;gBACtB,KAAK,CAAC,kBAAkB,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;gBAC/C,MAAM;YACR,KAAK,MAAM;gBACT,IAAI,MAAM,KAAK,SAAS,IAAI,MAAM,CAAC,IAAI,KAAK,WAAW;oBAAE,KAAK,CAAC,UAAU,IAAI,CAAC,CAAC;;oBAC1E,KAAK,CAAC,UAAU,IAAI,CAAC,CAAC;gBAC3B,IAAI,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC;oBAAE,KAAK,CAAC,aAAa,GAAG,IAAI,CAAC;gBAC7D,MAAM;YACR;gBACE,MAAM;QACV,CAAC;QACD,IAAI,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC;YACnB,MAAM,IAAI,GAAG,OAAO,IAAI,IAAI,CAAC,IAAI,KAAK,oBAAoB,IAAI,IAAI,CAAC,IAAI,KAAK,OAAO,CAAC;YACpF,KAAK,IAAI,CAAC,GAAG,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE;gBAChD,KAAK,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAU,EAAE,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;QACjF,CAAC;IACH,CAAC;IACD,KAAK,MAAM,GAAG,IAAI,UAAU,EAAE,CAAC;QAC7B,MAAM,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;QACnC,IAAI,IAAI,KAAK,SAAS;YAAE,IAAI,CAAC,GAAG,GAAG,WAAW,CAAC,GAAG,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,EAAE,CAAC;IAC3E,CAAC;IACD,KAAK,CAAC,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,GAAG,KAAK,EAAE,CAAC,CAAC;IACtD,KAAK,CAAC,KAAK,GAAG,eAAe,CAAC,IAAI,CAAC,CAAC;IACpC,OAAO,KAAK,CAAC;AACf,CAAC","sourcesContent":["import type { Nodes, Parent } from \"mdast\";\nimport { fromMarkdown } from \"mdast-util-from-markdown\";\nimport { gfmFootnoteFromMarkdown } from \"mdast-util-gfm-footnote\";\nimport { gfmTableFromMarkdown } from \"mdast-util-gfm-table\";\nimport { gfmFootnote } from \"micromark-extension-gfm-footnote\";\nimport { gfmTable } from \"micromark-extension-gfm-table\";\n\nexport interface BodyFacts {\n  firstHeading: string | undefined;\n  firstSentence: string | undefined;\n  links: Array<{ url: string; text: string }>;\n  footnoteReferences: string[];\n  htmlBlocks: number;\n  inlineHtml: number;\n  hasScriptLike: boolean;\n  /** The body nests deeper than the analyser allows; no facts were taken from it. */\n  unanalysed: boolean;\n  /** Only the first part of the body was analysed. */\n  truncated: boolean;\n  /** The body's text with blocks separated by one space, HTML and footnote marks left out; absent when unanalysed. */\n  prose?: string;\n}\n\nconst SENTENCE_CAP = 200;\nexport const ANALYSIS_BUDGET = 256 * 1024;\nconst MAX_NESTING = 256;\nconst MAX_EMPHASIS_RUNS = 2000;\nconst MAX_DEFINITIONS = 1000;\nconst SCRIPT_LIKE = /<\\s*\\/?\\s*(script|style|iframe)\\b/i;\nconst DELIMITER_RUN = /[*_]+/g;\nconst DEFINITION_LINE = /^ {0,3}\\[[^\\]]{1,999}\\]:/;\n\n/** An emphasis delimiter run that could open or close: one not wedged between two word characters. */\nfunction isFlankingRun(text: string, start: number, end: number): boolean {\n  const before = start === 0 ? \" \" : (text[start - 1] ?? \" \");\n  const after = end >= text.length ? \" \" : (text[end] ?? \" \");\n  const wordy = (c: string): boolean => /[\\p{L}\\p{N}]/u.test(c);\n  return !(wordy(before) && wordy(after));\n}\n\n/**\n * Why a body must not be parsed, or `undefined` when it may be. The parser's cost grows with the square of the\n * number of emphasis delimiter runs and of link reference definitions, and its stack with nesting depth, so\n * each is counted before any parse. The counts are deterministic and cheap.\n */\nexport function analysisBounds(text: string): \"nesting\" | \"emphasis\" | \"definitions\" | undefined {\n  if (nestingDepth(text) > MAX_NESTING) return \"nesting\";\n  let runs = 0;\n  for (const match of text.matchAll(DELIMITER_RUN)) {\n    if (isFlankingRun(text, match.index, match.index + match[0].length)) runs += 1;\n    if (runs > MAX_EMPHASIS_RUNS) return \"emphasis\";\n  }\n  let definitions = 0;\n  for (const line of text.split(\"\\n\")) {\n    if (DEFINITION_LINE.test(line)) definitions += 1;\n    if (definitions > MAX_DEFINITIONS) return \"definitions\";\n  }\n  return undefined;\n}\n\n/** The deepest run of block-quote markers or unclosed brackets, measured before any parse so the bound is deterministic. */\nexport function nestingDepth(body: string): number {\n  let deepest = 0;\n  let brackets = 0;\n  for (const line of body.split(\"\\n\")) {\n    const quotes = (/^[ \\t>]*/.exec(line)?.[0].match(/>/g) ?? []).length;\n    if (quotes > deepest) deepest = quotes;\n    for (const ch of line) {\n      if (ch === \"[\") brackets += 1;\n      else if (ch === \"]\" && brackets > 0) brackets -= 1;\n      if (brackets > deepest) deepest = brackets;\n    }\n  }\n  return deepest;\n}\n\nfunction isParent(node: Nodes): node is Nodes & Parent {\n  return \"children\" in node && Array.isArray((node as Parent).children);\n}\n\n/** The prose of a node, leaving out HTML and footnote marks, whitespace collapsed. */\nfunction proseOf(node: Nodes): string {\n  const parts: string[] = [];\n  const stack: Nodes[] = [node];\n  while (stack.length > 0) {\n    const current = stack.pop() as Nodes;\n    if (current.type === \"html\" || current.type === \"footnoteReference\") continue;\n    if (\"value\" in current && typeof current.value === \"string\") parts.push(current.value);\n    else if (isParent(current))\n      for (let i = current.children.length - 1; i >= 0; i--)\n        stack.push(current.children[i] as Nodes);\n  }\n  return parts.join(\"\").replace(/\\s+/g, \" \").trim();\n}\n\nconst BLOCKS = new Set([\n  \"paragraph\",\n  \"heading\",\n  \"blockquote\",\n  \"list\",\n  \"listItem\",\n  \"code\",\n  \"table\",\n  \"tableRow\",\n  \"tableCell\",\n  \"thematicBreak\",\n  \"footnoteDefinition\",\n]);\n\n/** The whole tree's prose, one space between blocks, HTML and footnote marks left out, whitespace collapsed. */\nfunction proseWithBlocks(tree: Nodes): string {\n  const parts: string[] = [];\n  const stack: Nodes[] = [tree];\n  while (stack.length > 0) {\n    const current = stack.pop() as Nodes;\n    if (current.type === \"html\" || current.type === \"footnoteReference\") continue;\n    if (BLOCKS.has(current.type)) parts.push(\" \");\n    if (\"value\" in current && typeof current.value === \"string\") parts.push(current.value);\n    else if (isParent(current)) {\n      for (let i = current.children.length - 1; i >= 0; i--)\n        stack.push(current.children[i] as Nodes);\n    }\n    if (BLOCKS.has(current.type)) parts.push(\" \");\n  }\n  return parts.join(\"\").replace(/\\s+/g, \" \").trim();\n}\n\nfunction firstSentenceOf(text: string): string {\n  const match = /^(.*?[.!?])(?=\\s|$)/.exec(text);\n  const sentence = match?.[1] ?? text;\n  return sentence.length > SENTENCE_CAP ? sentence.slice(0, SENTENCE_CAP).trimEnd() : sentence;\n}\n\nconst EMPTY: Omit<BodyFacts, \"unanalysed\" | \"truncated\"> = {\n  firstHeading: undefined,\n  firstSentence: undefined,\n  links: [],\n  footnoteReferences: [],\n  htmlBlocks: 0,\n  inlineHtml: 0,\n  hasScriptLike: false,\n};\n\n/**\n * Reads the facts the OKF layer needs from a Markdown body: CommonMark with footnotes and tables, never the\n * autolink extension, whose cost grows without bound on nested brackets. A body nested past the limit is left\n * unanalysed rather than parsed; a body past the budget is analysed up to it. The tree is walked iteratively.\n */\nexport function readBody(body: string): BodyFacts {\n  const truncated = body.length > ANALYSIS_BUDGET;\n  const text = truncated ? body.slice(0, ANALYSIS_BUDGET) : body;\n  if (analysisBounds(text) !== undefined)\n    return { ...EMPTY, links: [], footnoteReferences: [], unanalysed: true, truncated };\n  const tree = fromMarkdown(text, {\n    extensions: [gfmFootnote(), gfmTable()],\n    mdastExtensions: [gfmFootnoteFromMarkdown(), gfmTableFromMarkdown()],\n  });\n  const facts: BodyFacts = {\n    ...EMPTY,\n    links: [],\n    footnoteReferences: [],\n    unanalysed: false,\n    truncated,\n  };\n  const definitions = new Map<string, string>();\n  const references: Array<{ identifier: string; slot: number }> = [];\n  const stack: Array<{ node: Nodes; parent: Nodes | undefined; skipped: boolean }> = [\n    { node: tree, parent: undefined, skipped: false },\n  ];\n  while (stack.length > 0) {\n    const { node, parent, skipped } = stack.pop() as {\n      node: Nodes;\n      parent: Nodes | undefined;\n      skipped: boolean;\n    };\n    switch (node.type) {\n      case \"definition\":\n        definitions.set(node.identifier, node.url);\n        break;\n      case \"heading\":\n        if (facts.firstHeading === undefined) facts.firstHeading = proseOf(node);\n        break;\n      case \"paragraph\":\n        if (!skipped && facts.firstSentence === undefined) {\n          const sentence = firstSentenceOf(proseOf(node));\n          if (sentence.length > 0) facts.firstSentence = sentence;\n        }\n        break;\n      case \"link\":\n        facts.links.push({ url: node.url, text: proseOf(node) });\n        break;\n      case \"linkReference\":\n        references.push({ identifier: node.identifier, slot: facts.links.length });\n        facts.links.push({ url: \"\", text: proseOf(node) });\n        break;\n      case \"footnoteReference\":\n        facts.footnoteReferences.push(node.identifier);\n        break;\n      case \"html\":\n        if (parent !== undefined && parent.type === \"paragraph\") facts.inlineHtml += 1;\n        else facts.htmlBlocks += 1;\n        if (SCRIPT_LIKE.test(node.value)) facts.hasScriptLike = true;\n        break;\n      default:\n        break;\n    }\n    if (isParent(node)) {\n      const skip = skipped || node.type === \"footnoteDefinition\" || node.type === \"table\";\n      for (let i = node.children.length - 1; i >= 0; i--)\n        stack.push({ node: node.children[i] as Nodes, parent: node, skipped: skip });\n    }\n  }\n  for (const ref of references) {\n    const slot = facts.links[ref.slot];\n    if (slot !== undefined) slot.url = definitions.get(ref.identifier) ?? \"\";\n  }\n  facts.links = facts.links.filter((l) => l.url !== \"\");\n  facts.prose = proseWithBlocks(tree);\n  return facts;\n}\n"]}

package/dist/bundle/model.js

@@ -0,0 +1,7 @@
+/** Core types. Pure data; nothing here touches the file system, the engine or the protocol. */
+export const DEFAULT_CAPS = {
+    fileBytes: 2 * 1024 * 1024,
+    files: 20_000,
+    treeBytes: 512 * 1024 * 1024,
+};
+//# sourceMappingURL=model.js.map

package/dist/bundle/model.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"model.js","sourceRoot":"","sources":["../../src/bundle/model.ts"],"names":[],"mappings":"AAAA,+FAA+F;AA+K/F,MAAM,CAAC,MAAM,YAAY,GAAS;IAChC,SAAS,EAAE,CAAC,GAAG,IAAI,GAAG,IAAI;IAC1B,KAAK,EAAE,MAAM;IACb,SAAS,EAAE,GAAG,GAAG,IAAI,GAAG,IAAI;CAC7B,CAAC","sourcesContent":["/** Core types. Pure data; nothing here touches the file system, the engine or the protocol. */\n\n/** A bundle-relative path with POSIX separators and no leading slash. */\nexport type PagePath = string;\n\nexport interface BundleFile {\n  path: string;\n  bytes: Uint8Array;\n}\n\nexport type Status = \"draft\" | \"stable\" | \"deprecated\";\nexport type Trust = \"unverified\" | \"machine-confirmed\" | \"human-reviewed\";\nexport type SpecText = \"2026-08-15\" | \"2026-08-21\";\n\n/** A timestamp as written, with its instant when it parsed. */\nexport interface Timestamp {\n  raw: string;\n  at?: Date;\n}\n\nexport interface StaleAfter {\n  raw: string;\n  /** date: overdue from the start of that UTC day; datetime: overdue from that instant; unparseable: never overdue. */\n  form: \"date\" | \"datetime\" | \"unparseable\";\n  at?: Date;\n}\n\nexport interface Source {\n  resource: string;\n  id?: string;\n  title?: string;\n  author?: string;\n  usageCount?: number;\n  lastModified?: string;\n  usageWindow?: { from: string; to: string };\n}\n\nexport interface Verification {\n  by: string;\n  at?: Timestamp;\n}\n\nexport type LinkKind =\n  | \"page\"\n  | \"folder\"\n  | \"reserved\"\n  | \"attachment\"\n  | \"anchor\"\n  | \"external\"\n  | \"broken\";\n\nexport interface Link {\n  raw: string;\n  kind: LinkKind;\n  /** The bundle path the link resolved to, for page, folder, reserved and attachment links. */\n  target?: string;\n}\n\nexport type DegradationCode =\n  | \"title-from-heading\"\n  | \"title-from-filename\"\n  | \"description-from-body\"\n  | \"description-missing\"\n  | \"scalar-coerced\"\n  | \"field-ignored\"\n  | \"tags-not-list\"\n  | \"status-unknown\"\n  | \"stale-after-unparseable\"\n  | \"stale-after-no-offset\"\n  | \"stale-after-unexpected-form\"\n  | \"timestamp-invalid\"\n  | \"generated-malformed\"\n  | \"verified-entry-malformed\"\n  | \"source-malformed\"\n  | \"footnote-without-source\"\n  | \"verification-without-at\"\n  | \"body-html\"\n  | \"body-unanalysed\"\n  | \"body-truncated\"\n  | \"index-lists-unserved\"\n  | \"replacement-missing\"\n  | \"replacement-broken\"\n  | \"replacement-external\"\n  | \"replacement-not-served\"\n  | \"replacement-self\"\n  | \"replacement-not-a-page\"\n  | \"frontmatter-warning\"\n  | \"reserved-frontmatter-unparseable\"\n  | \"reserved-unanalysed\"\n  | \"okf-version-unknown\";\n\nexport interface Degradation {\n  path: string;\n  code: DegradationCode;\n  field: string;\n  detail: string;\n}\n\nexport type RefusalRule =\n  | \"no-frontmatter\"\n  | \"frontmatter-unparseable\"\n  | \"body-unreadable\"\n  | \"no-type\"\n  | \"not-utf8\"\n  | \"symlink\"\n  | \"gitlink\"\n  | \"path-escape\"\n  | \"special-file\"\n  | \"unreadable\"\n  | \"engine-config\"\n  | \"hash-mismatch\"\n  | \"size-mismatch\"\n  | \"not-in-manifest\"\n  | \"manifest-missing\"\n  | \"manifest-invalid\"\n  | \"oversize\"\n  | \"too-many-files\"\n  | \"tree-too-large\"\n  | \"bundle-path-missing\";\n\nexport interface Refusal {\n  path: string;\n  rule: RefusalRule;\n  detail: string;\n}\n\nexport interface Page {\n  path: PagePath;\n  folder: string;\n  hash: string;\n  type: string;\n  title: string;\n  titleSource: \"frontmatter\" | \"heading\" | \"filename\";\n  description?: string;\n  descriptionSource: \"frontmatter\" | \"body\" | \"none\";\n  tags: string[];\n  status: Status;\n  statusSource: \"frontmatter\" | \"default\";\n  statusRaw?: string;\n  staleAfter?: StaleAfter;\n  generated?: { by: string; at?: Timestamp };\n  verified: Verification[];\n  /** The verification with the latest instant; among entries without one, the last listed. */\n  latestVerification?: Verification;\n  trust: Trust;\n  sources: Source[];\n  usageWindow?: { from: string; to: string };\n  resource?: string;\n  replacement?: PagePath;\n  links: Link[];\n  footnoteReferences: string[];\n  frontmatter: Record<string, unknown>;\n  body: string;\n  /** The body's prose for snippets, captured once at load; absent when the body could not be analysed. */\n  prose?: string;\n  degradations: Degradation[];\n}\n\nexport interface ReservedFile {\n  kind: \"index\" | \"log\";\n  path: string;\n  folder: string;\n  frontmatter?: Record<string, unknown>;\n  body: string;\n  text: string;\n  okfVersion?: string;\n  degradations: Degradation[];\n}\n\nexport interface Caps {\n  fileBytes: number;\n  files: number;\n  treeBytes: number;\n}\n\nexport const DEFAULT_CAPS: Caps = {\n  fileBytes: 2 * 1024 * 1024,\n  files: 20_000,\n  treeBytes: 512 * 1024 * 1024,\n};\n\nexport interface LoadOptions {\n  admit: Status[];\n  dev: boolean;\n  integrity: \"require-manifest\" | \"none\";\n  specText: SpecText;\n  caps: Caps;\n  types?: string[];\n  /** Refusals the file walker produced before the core saw the files (symbolic links, escapes, oversize). */\n  walkRefusals?: Refusal[];\n  /** Dot-leading paths the walker found and never read; counted, present for the manifest, refused when they are the engine's. */\n  hiddenPaths?: string[];\n  /** The hidden paths that are folders: a manifest entry beneath one is present even though no file arrived. */\n  hiddenFolders?: string[];\n  /** The walker's own bundle-level refusal, when the caps stopped it. */\n  walkFatal?: Refusal;\n}\n\nexport interface Report {\n  loadedAt: Date;\n  commit?: string;\n  /** A bundle-level refusal; when set, nothing should be served. */\n  fatal?: Refusal;\n  /** Whether the manifest was verified against the files, or integrity was not required. */\n  integrity: \"checked\" | \"skipped\";\n  admitted: number;\n  excludedByStatus: number;\n  attachments: number;\n  hidden: number;\n  refusals: Refusal[];\n  degradations: Degradation[];\n  unknownTypes: string[];\n  unknownStatuses: Array<{ path: string; value: string }>;\n  brokenLinks: Array<{ from: PagePath; raw: string }>;\n  linksToUnserved: Array<{ from: PagePath; raw: string; target: string }>;\n  missingOnDisk: string[];\n  foldersWithoutIndex: string[];\n  encodedFolders: string[];\n}\n"]}

package/dist/bundle/page.js

@@ -0,0 +1,374 @@
+import { decodeUtf8, parseFrontmatter, splitFrontmatter } from "./frontmatter.js";
+import { resolveLink } from "./links.js";
+import { sha256Hex } from "./manifest.js";
+import { readBody } from "./markdown.js";
+import { folderOf } from "./paths.js";
+import { parseTimestamp } from "./timestamp.js";
+const STATUSES = new Set(["draft", "stable", "deprecated"]);
+function isRecord(value) {
+    return typeof value === "object" && value !== null && !Array.isArray(value);
+}
+/** What a value is, for a sentence: "a list", "a mapping", "a number", "empty". */
+function kindOf(value) {
+    if (value === null || value === undefined)
+        return "empty";
+    if (Array.isArray(value))
+        return "a list";
+    if (typeof value === "object")
+        return "a mapping";
+    return `a ${typeof value}`;
+}
+function stemOf(path) {
+    return path.slice(path.lastIndexOf("/") + 1).replace(/\.md$/, "");
+}
+/** The verification with the latest instant; among entries without one, the last listed. */
+export function latestVerification(verified) {
+    let best;
+    let bestAt = Number.NEGATIVE_INFINITY;
+    for (const v of verified) {
+        const at = v.at?.at?.getTime();
+        if (at === undefined) {
+            if (best === undefined)
+                best = v;
+            continue;
+        }
+        if (best === undefined || best.at?.at === undefined || at >= bestAt) {
+            best = v;
+            bestAt = at;
+        }
+    }
+    if (best !== undefined && best.at?.at === undefined) {
+        for (const v of verified)
+            if (v.at?.at === undefined)
+                best = v;
+    }
+    return best;
+}
+/** Turns one file into a page, deriving every field with its source and a degradation wherever a fallback was used. */
+export function parsePage(file, ctx) {
+    const path = file.path;
+    const degradations = [];
+    const degrade = (code, field, detail) => {
+        degradations.push({ path, code, field, detail });
+    };
+    const refuse = (rule, detail) => ({
+        ok: false,
+        refusal: { path, rule, detail },
+    });
+    const decoded = decodeUtf8(file.bytes);
+    if (!decoded.ok)
+        return refuse("not-utf8", "the file is not valid UTF-8");
+    const split = splitFrontmatter(decoded.text);
+    if (split.block === undefined) {
+        return refuse("no-frontmatter", split.unclosed === true
+            ? "the frontmatter block opened on line 1 is never closed"
+            : "no frontmatter block opens the file");
+    }
+    const parsed = parseFrontmatter(split.block, 1);
+    if (!parsed.ok)
+        return refuse("frontmatter-unparseable", parsed.error);
+    const { data, sources: scalarText } = parsed;
+    for (const warning of parsed.warnings)
+        degrade("frontmatter-warning", "frontmatter", warning);
+    /** A top-level value read as text: strings as they are, numbers and booleans by their source text, reported. */
+    const text = (key) => {
+        const value = data[key];
+        if (typeof value === "string")
+            return value;
+        if (typeof value === "number" || typeof value === "boolean") {
+            const asText = scalarText[key] ?? String(value);
+            degrade("scalar-coerced", key, `${key} is a ${typeof value}, read as "${asText}"`);
+            return asText;
+        }
+        return undefined;
+    };
+    if (!Object.hasOwn(data, "type"))
+        return refuse("no-type", "the frontmatter has no type");
+    const type = text("type")?.trim();
+    if (type === undefined)
+        return refuse("no-type", `type is ${kindOf(data.type)}, not text`);
+    if (type.length === 0)
+        return refuse("no-type", "type is empty");
+    const facts = readBody(split.body);
+    if (facts.unanalysed) {
+        degrade("body-unanalysed", "body", "the body nests deeper than the analyser allows; no title, description or links were taken from it");
+    }
+    if (facts.truncated)
+        degrade("body-truncated", "body", "only the first 256 KiB of the body were analysed");
+    let title = text("title")?.trim();
+    let titleSource = "frontmatter";
+    if (Object.hasOwn(data, "title") && title === undefined) {
+        degrade("field-ignored", "title", `title is ${kindOf(data.title)}, not text; ignored`);
+    }
+    if (title === undefined || title.length === 0) {
+        if (facts.firstHeading !== undefined && facts.firstHeading.length > 0) {
+            title = facts.firstHeading;
+            titleSource = "heading";
+            degrade("title-from-heading", "title", `no title field; the first heading "${title}" is used`);
+        }
+        else {
+            title = stemOf(path);
+            titleSource = "filename";
+            degrade("title-from-filename", "title", `no title field and no heading; the file name "${title}" is used`);
+        }
+    }
+    let description = text("description")?.trim();
+    let descriptionSource = "frontmatter";
+    if (Object.hasOwn(data, "description") && description === undefined) {
+        degrade("field-ignored", "description", `description is ${kindOf(data.description)}, not text; ignored`);
+    }
+    if (description === undefined || description.length === 0) {
+        if (facts.firstSentence !== undefined) {
+            description = facts.firstSentence;
+            descriptionSource = "body";
+            degrade("description-from-body", "description", "no description field; the body's first sentence is used");
+        }
+        else {
+            description = undefined;
+            descriptionSource = "none";
+            degrade("description-missing", "description", "no description field and no body text to take one from");
+        }
+    }
+    const tags = [];
+    const rawTags = data.tags;
+    if (Array.isArray(rawTags)) {
+        rawTags.forEach((tag, i) => {
+            if (typeof tag === "string")
+                tags.push(tag);
+            else if (typeof tag === "number" || typeof tag === "boolean")
+                tags.push(String(tag));
+            else
+                degrade("field-ignored", "tags", `tags[${i}] is ${kindOf(tag)}, not text; ignored`);
+        });
+    }
+    else if (typeof rawTags === "string") {
+        tags.push(rawTags);
+        degrade("tags-not-list", "tags", "tags is a single string, read as one tag");
+    }
+    else if (rawTags !== undefined && rawTags !== null) {
+        degrade("tags-not-list", "tags", `tags is ${kindOf(rawTags)}, not a list; ignored`);
+    }
+    const statusResult = normaliseStatus(data.status, Object.hasOwn(data, "status"), degrade);
+    let staleAfter;
+    if (Object.hasOwn(data, "stale_after") &&
+        data.stale_after !== null &&
+        data.stale_after !== undefined) {
+        const raw = text("stale_after") ?? JSON.stringify(data.stale_after);
+        const ts = parseTimestamp(raw);
+        const expected = ctx.specText === "2026-08-15" ? "date" : "datetime";
+        if (ts.kind === "date")
+            staleAfter = { raw, form: "date", at: ts.at };
+        else if (ts.kind === "datetime") {
+            staleAfter = { raw, form: "datetime", at: ts.at };
+            if (!ts.offset)
+                degrade("stale-after-no-offset", "stale_after", `${raw} has no offset and is read as UTC`);
+        }
+        else {
+            staleAfter = { raw, form: "unparseable" };
+            degrade("stale-after-unparseable", "stale_after", `${raw} is not a date or a datetime; the page is never overdue`);
+        }
+        if (staleAfter.form !== "unparseable" && staleAfter.form !== expected) {
+            degrade("stale-after-unexpected-form", "stale_after", `the ${ctx.specText} text expects a ${expected}; ${raw} is a ${staleAfter.form}`);
+        }
+    }
+    const timestamp = (value, field) => {
+        const raw = typeof value === "string" ? value : JSON.stringify(value);
+        const ts = parseTimestamp(raw);
+        if (ts.kind === "invalid") {
+            degrade("timestamp-invalid", field, `${raw} is not a date or a datetime`);
+            return { raw };
+        }
+        return { raw, at: ts.at };
+    };
+    let generated;
+    if (Object.hasOwn(data, "generated")) {
+        const g = data.generated;
+        if (isRecord(g) && typeof g.by === "string" && g.by.trim().length > 0) {
+            generated = { by: g.by.trim() };
+            if (Object.hasOwn(g, "at"))
+                generated.at = timestamp(g.at, "generated.at");
+        }
+        else {
+            degrade("generated-malformed", "generated", `generated is ${kindOf(g)} without a by actor; ignored`);
+        }
+    }
+    const verified = [];
+    if (Object.hasOwn(data, "verified")) {
+        const v = data.verified;
+        const entries = Array.isArray(v) ? v : isRecord(v) ? [v] : [];
+        if (!Array.isArray(v) && !isRecord(v)) {
+            degrade("verified-entry-malformed", "verified", `verified is ${kindOf(v)}, neither a list nor a mapping; ignored`);
+        }
+        entries.forEach((entry, i) => {
+            if (isRecord(entry) && typeof entry.by === "string" && entry.by.trim().length > 0) {
+                const verification = { by: entry.by.trim() };
+                if (Object.hasOwn(entry, "at") && entry.at !== null)
+                    verification.at = timestamp(entry.at, `verified[${i}].at`);
+                else
+                    degrade("verification-without-at", "verified", `verified[${i}] by ${verification.by} has no at`);
+                verified.push(verification);
+            }
+            else {
+                degrade("verified-entry-malformed", "verified", `verified[${i}] has no by actor; ignored`);
+            }
+        });
+    }
+    const trust = verified.length === 0
+        ? "unverified"
+        : verified.some((v) => v.by.startsWith("human:"))
+            ? "human-reviewed"
+            : "machine-confirmed";
+    const sources = [];
+    if (Object.hasOwn(data, "sources")) {
+        const s = data.sources;
+        if (Array.isArray(s)) {
+            s.forEach((entry, i) => {
+                if (isRecord(entry) && typeof entry.resource === "string" && entry.resource.length > 0) {
+                    const source = { resource: entry.resource };
+                    if (typeof entry.id === "string")
+                        source.id = entry.id;
+                    else if (typeof entry.id === "number" || typeof entry.id === "boolean")
+                        source.id = String(entry.id);
+                    if (typeof entry.title === "string")
+                        source.title = entry.title;
+                    if (typeof entry.author === "string")
+                        source.author = entry.author;
+                    if (typeof entry.usage_count === "number") {
+                        if (Number.isFinite(entry.usage_count))
+                            source.usageCount = entry.usage_count;
+                        else
+                            degrade("source-malformed", "sources", `sources[${i}].usage_count is not a finite number; ignored`);
+                    }
+                    if (typeof entry.last_modified === "string")
+                        source.lastModified = entry.last_modified;
+                    const w = entry.usage_window;
+                    if (isRecord(w) && typeof w.from === "string" && typeof w.to === "string")
+                        source.usageWindow = { from: w.from, to: w.to };
+                    sources.push(source);
+                }
+                else {
+                    degrade("source-malformed", "sources", `sources[${i}] has no resource; ignored`);
+                }
+            });
+        }
+        else {
+            degrade("source-malformed", "sources", `sources is ${kindOf(s)}, not a list; ignored`);
+        }
+    }
+    let usageWindow;
+    const w = data.usage_window;
+    if (isRecord(w) && typeof w.from === "string" && typeof w.to === "string")
+        usageWindow = { from: w.from, to: w.to };
+    const resource = text("resource");
+    const links = facts.links.map((l) => {
+        const resolved = resolveLink(l.url, path, ctx.linkIndex);
+        const link = { raw: l.url, kind: resolved.kind };
+        if (resolved.target !== undefined)
+            link.target = resolved.target;
+        return link;
+    });
+    // GFM lower-cases footnote identifiers; the join with sources[].id is case-insensitive, and each missing id is reported once.
+    const sourceIds = new Set(sources.map((s) => s.id?.toLowerCase()).filter((id) => id !== undefined));
+    const missing = new Set();
+    for (const id of facts.footnoteReferences)
+        if (!sourceIds.has(id.toLowerCase()))
+            missing.add(id);
+    for (const id of missing)
+        degrade("footnote-without-source", "sources", `footnote ${id} has no matching sources entry`);
+    if (facts.htmlBlocks > 0 || facts.inlineHtml > 0 || facts.hasScriptLike) {
+        const parts = [];
+        if (facts.htmlBlocks > 0)
+            parts.push(`${facts.htmlBlocks} HTML block${facts.htmlBlocks === 1 ? "" : "s"}`);
+        if (facts.inlineHtml > 0)
+            parts.push(`${facts.inlineHtml} inline HTML element${facts.inlineHtml === 1 ? "" : "s"}`);
+        if (facts.hasScriptLike)
+            parts.push("including a script-like element");
+        degrade("body-html", "body", `${parts.join(", ")}; indexed as text, never executed, kept out of the description`);
+    }
+    const page = {
+        path,
+        folder: folderOf(path),
+        hash: sha256Hex(file.bytes),
+        type,
+        title,
+        titleSource,
+        descriptionSource,
+        tags,
+        status: statusResult.status,
+        statusSource: statusResult.source,
+        verified,
+        trust,
+        sources,
+        links,
+        footnoteReferences: facts.footnoteReferences,
+        frontmatter: data,
+        body: split.body,
+        ...(facts.prose === undefined ? {} : { prose: facts.prose }),
+        degradations,
+    };
+    if (description !== undefined)
+        page.description = description;
+    if (statusResult.raw !== undefined)
+        page.statusRaw = statusResult.raw;
+    if (staleAfter !== undefined)
+        page.staleAfter = staleAfter;
+    if (generated !== undefined)
+        page.generated = generated;
+    const latest = latestVerification(verified);
+    if (latest !== undefined)
+        page.latestVerification = latest;
+    if (usageWindow !== undefined)
+        page.usageWindow = usageWindow;
+    if (resource !== undefined)
+        page.resource = resource;
+    return { ok: true, page };
+}
+function normaliseStatus(value, present, degrade) {
+    if (!present || value === undefined || value === null)
+        return { status: "stable", source: "default" };
+    const raw = typeof value === "string" ? value : JSON.stringify(value);
+    const normalised = raw.trim().toLowerCase();
+    if (normalised.length === 0)
+        return { status: "stable", source: "default" };
+    if (STATUSES.has(normalised))
+        return { status: normalised, source: "frontmatter", raw };
+    degrade("status-unknown", "status", `status "${raw}" is not draft, stable or deprecated; treated as draft`);
+    return { status: "draft", source: "frontmatter", raw };
+}
+const ARTICLE = {
+    folder: "a folder",
+    reserved: "an index or log file",
+    attachment: "an attachment",
+};
+/**
+ * The replacement for a deprecated page: the first body link that is not a same-page anchor decides. It is the
+ * replacement when it resolves to an admitted page other than this one; otherwise there is none, and the reason is
+ * recorded. Nothing is decided for a page that is not deprecated.
+ */
+export function decideReplacement(page, admitted) {
+    if (page.status !== "deprecated")
+        return {};
+    const reason = (code, detail) => ({
+        degradation: { path: page.path, code, field: "replacement", detail },
+    });
+    const first = page.links.find((l) => l.kind !== "anchor");
+    if (first === undefined)
+        return reason("replacement-missing", "a deprecated page should link to the page that replaced it; no link was found");
+    switch (first.kind) {
+        case "external":
+            return reason("replacement-external", `the first link, ${first.raw}, points outside the bundle`);
+        case "broken":
+            return reason("replacement-broken", `the first link, ${first.raw}, points at nothing in the bundle`);
+        case "page": {
+            const target = first.target ?? "";
+            if (target === page.path)
+                return reason("replacement-self", "the first link points at the page itself");
+            if (!admitted.has(target))
+                return reason("replacement-not-served", `the first link points at ${target}, which is not served`);
+            return { replacement: target };
+        }
+        default:
+            return reason("replacement-not-a-page", `the first link, ${first.raw}, points at ${ARTICLE[first.kind] ?? first.kind}, not a page`);
+    }
+}
+//# sourceMappingURL=page.js.map