feat: standardize mdtodo rxx parser

Refs #2157
This commit is contained in:
lyon
2026-06-25 21:38:08 +08:00
parent 590b662317
commit fc601f4d29
3 changed files with 513 additions and 28 deletions
+92 -3
View File
@@ -1,13 +1,51 @@
import { expect, test } from "bun:test";
import { parseMdtodoDocument } from "./mdtodo.ts";
import { mutateMdtodoDocument, parseMdtodoDocument } from "./mdtodo.ts";
test("parseMdtodoDocument projects markdown tasks without raw body leakage", () => {
const document = parseMdtodoDocument(`# Demo\n\n- [ ] Root task\n - [x] Child done\n- [-] Blocked item\n`, {
test("parseMdtodoDocument projects Rxx hierarchy without raw body leakage", () => {
const document = parseMdtodoDocument(`# Demo
## R1 Root task [Processing]
Root notes with [context](https://example.test/context).
### R1.1 Child done [Finished]
- implementation notes stay in source markdown
## R2 Backlog item
`, {
sourceId: "demo",
relativePath: "project-management/MDTODO.md"
});
expect(document.parserMode).toBe("rxx");
expect(document.taskCount).toBe(3);
expect(document.tasks.map((task) => task.taskId)).toEqual(["R1", "R1.1", "R2"]);
expect(document.tasks.map((task) => task.status)).toEqual(["in_progress", "done", "open"]);
expect(document.tasks[1].parentTaskRef).toBe(document.tasks[0].taskRef);
expect(document.tasks[0].depth).toBe(0);
expect(document.tasks[1].depth).toBe(1);
expect(document.tasks[0].linkCount).toBe(1);
expect(document.tasks[0].bodyPreview).toBe("Root notes with [context](https://example.test/context).");
expect(document.tasks[0].body).toBeUndefined();
expect(document.tasks[0].rawContent).toBeUndefined();
expect(document.document.parserMode).toBe("rxx");
expect(document.document.valuesRedacted).toBe(true);
});
test("parseMdtodoDocument keeps legacy checkbox parsing explicit", () => {
const document = parseMdtodoDocument(`# Demo
- [ ] Root task
- [x] Child done
- [-] Blocked item
`, {
sourceId: "demo",
relativePath: "project-management/MDTODO.md"
});
expect(document.parserMode).toBe("legacy-checkbox");
expect(document.taskCount).toBe(3);
expect(document.tasks[0].status).toBe("open");
expect(document.tasks[1].status).toBe("done");
@@ -15,3 +53,54 @@ test("parseMdtodoDocument projects markdown tasks without raw body leakage", ()
expect(document.tasks[2].status).toBe("blocked");
expect(document.document.valuesRedacted).toBe(true);
});
test("mutateMdtodoDocument supports Rxx add update delete operations", () => {
const initial = `# Demo
## R1 Root task
Original body.
### R1.1 Existing child
`;
let result = mutateMdtodoDocument(initial, { type: "addTask", title: "Second root" });
expect(result.ok).toBe(true);
expect(result.changedTaskId).toBe("R2");
expect(result.markdown).toContain("## R2 Second root");
result = mutateMdtodoDocument(result.markdown, { type: "addTask", parentRxxId: "R1", title: "Second child", body: "Child body" });
expect(result.changedTaskId).toBe("R1.2");
expect(result.markdown).toContain("### R1.2 Second child\n\nChild body");
result = mutateMdtodoDocument(result.markdown, { type: "addTask", afterRxxId: "R1.1", title: "Continue child" });
expect(result.changedTaskId).toBe("R1.3");
expect(result.markdown).toContain("### R1.3 Continue child");
result = mutateMdtodoDocument(result.markdown, { type: "updateStatus", rxxId: "R1", status: "done" });
expect(result.markdown).toContain("## R1 Root task [completed]");
result = mutateMdtodoDocument(result.markdown, { type: "updateTitle", rxxId: "R1", title: "Renamed root" });
expect(result.markdown).toContain("## R1 Renamed root [completed]");
result = mutateMdtodoDocument(result.markdown, { type: "updateBody", rxxId: "R1", body: "Rewritten body." });
expect(result.markdown).toContain("## R1 Renamed root [completed]\n\nRewritten body.\n\n### R1.1 Existing child");
result = mutateMdtodoDocument(result.markdown, { type: "deleteTask", rxxId: "R1.2" });
expect(result.affectedTaskIds).toEqual(["R1.2"]);
expect(result.markdown).not.toContain("R1.2");
expect(result.document.tasks.map((task) => task.taskId)).toEqual(["R1", "R1.1", "R1.3", "R2"]);
});
test("mutateMdtodoDocument reports optimistic revision conflicts", () => {
const conflict = mutateMdtodoDocument("# Demo\n\n## R1 Root\n", {
type: "updateTitle",
rxxId: "R1",
title: "No write",
expectedFingerprint: "stale"
});
expect(conflict.ok).toBe(false);
expect(conflict.conflict).toBe(true);
expect(conflict.code).toBe("revision_conflict");
});
+419 -24
View File
@@ -1,3 +1,7 @@
/**
* SPEC: PJ2026-010404 项目管理 draft-2026-06-25-p0-mdtodo-web-active-editing-hwpod-source.
* Projects MDTODO Markdown source-of-truth into stable Rxx task projections and mutation helpers.
*/
import { createHash } from "node:crypto";
import { promises as fs } from "node:fs";
import path from "node:path";
@@ -15,6 +19,27 @@ const excludedDirectories = new Set([
]);
const taskLinePattern = /^(\s*)(?:[-*+]|\d+[.)])\s+\[([ xX-])\]\s+(.*)$/u;
const rxxHeadingPattern = /^(#{2,})\s+(R\d+(?:\.\d+)*)(?:\s+(.*?))?\s*$/iu;
const rxxIdPattern = /^R\d+(?:\.\d+)*$/iu;
const markdownLinkPattern = /!?\[[^\]]*\]\([^)]+\)|https?:\/\/\S+/giu;
const statusAliases = new Map([
["open", "open"],
["todo", "open"],
["blocked", "blocked"],
["in_progress", "in_progress"],
["processing", "in_progress"],
["completed", "done"],
["finished", "done"],
["done", "done"]
]);
const statusMarkers = new Map([
["open", null],
["blocked", "blocked"],
["in_progress", "in_progress"],
["done", "completed"]
]);
export async function discoverMdtodoDocuments(options = {}) {
const root = path.resolve(String(options.root || process.cwd()));
@@ -45,6 +70,162 @@ export function parseMdtodoDocument(markdown, options = {}) {
const fingerprint = sha256(markdown);
const projectId = safeToken(options.projectId, projectIdForRelativePath(relativePath));
const lines = String(markdown ?? "").split(/\r?\n/u);
const context = { sourceId, relativePath, fileRef, fingerprint, projectId };
const rxxProjection = parseRxxTasks(lines, context);
const projection = rxxProjection.tasks.length > 0 ? rxxProjection : parseLegacyCheckboxTasks(lines, context);
const parserMode = rxxProjection.tasks.length > 0 ? "rxx" : "legacy-checkbox";
return {
sourceId,
fileRef,
relativePath,
fingerprint,
revision: fingerprint,
projectId,
title: documentTitle(markdown, relativePath),
parserMode,
taskCount: projection.tasks.length,
tasks: projection.tasks,
document: {
source: "markdown-file",
relativePath,
firstHeading: documentTitle(markdown, relativePath),
parserMode,
revision: fingerprint,
parseDiagnostics: projection.diagnostics,
valuesRedacted: true
}
};
}
export function mutateMdtodoDocument(markdown, operation = {}, options = {}) {
const originalMarkdown = String(markdown ?? "");
const previousFingerprint = sha256(originalMarkdown);
const expectedFingerprint = operation.expectedFingerprint ?? options.expectedFingerprint;
if (expectedFingerprint && expectedFingerprint !== previousFingerprint) {
return {
ok: false,
conflict: true,
code: "revision_conflict",
message: "MDTODO document fingerprint does not match expectedFingerprint",
expectedFingerprint,
actualFingerprint: previousFingerprint
};
}
const type = String(operation.type ?? "").trim();
const lines = originalMarkdown.split(/\r?\n/u);
const blocks = collectRxxBlocks(lines);
let nextLines = lines.slice();
let changedTaskId = null;
let affectedTaskIds = [];
if (type === "updateTitle") {
const block = requireRxxBlock(blocks, operation.rxxId);
const title = cleanTaskTitle(operation.title);
nextLines[block.lineIndex] = formatRxxHeading(block.hashes, block.rxxId, block.status, title);
changedTaskId = block.rxxId;
affectedTaskIds = [block.rxxId];
} else if (type === "updateStatus") {
const block = requireRxxBlock(blocks, operation.rxxId);
const status = requireStatus(operation.status);
nextLines[block.lineIndex] = formatRxxHeading(block.hashes, block.rxxId, status, block.title);
changedTaskId = block.rxxId;
affectedTaskIds = [block.rxxId];
} else if (type === "updateBody") {
const block = requireRxxBlock(blocks, operation.rxxId);
const bodyLines = bodyReplacementLines(operation.body);
nextLines.splice(block.lineIndex + 1, Math.max(0, block.endLine - block.lineIndex - 1), ...bodyLines);
changedTaskId = block.rxxId;
affectedTaskIds = [block.rxxId];
} else if (type === "addTask") {
const add = buildAddTaskMutation(lines, blocks, operation);
nextLines = insertLines(nextLines, add.insertPos, add.lines);
changedTaskId = add.rxxId;
affectedTaskIds = [add.rxxId];
} else if (type === "deleteTask") {
const block = requireRxxBlock(blocks, operation.rxxId);
const endLine = subtreeEndLine(blocks, block.rxxId);
affectedTaskIds = blocks
.filter((candidate) => candidate.rxxId === block.rxxId || isDescendantRxxId(candidate.rxxId, block.rxxId))
.map((candidate) => candidate.rxxId);
nextLines.splice(block.lineIndex, Math.max(1, endLine - block.lineIndex));
changedTaskId = block.rxxId;
} else {
throw mutationError("unsupported_mutation", `Unsupported MDTODO mutation type: ${type || "<empty>"}`);
}
const nextMarkdown = normalizeMarkdownOutput(nextLines);
const fingerprint = sha256(nextMarkdown);
const document = parseMdtodoDocument(nextMarkdown, options);
return {
ok: true,
conflict: false,
markdown: nextMarkdown,
previousFingerprint,
fingerprint,
revision: fingerprint,
changedTaskId,
affectedTaskIds,
taskCount: document.taskCount,
document
};
}
export function documentFileRef(relativePath) {
return `file_${shortHash(normalizeRelativePath(relativePath), 16)}`;
}
function parseRxxTasks(lines, context) {
const diagnostics = [];
const blocks = collectRxxBlocks(lines);
const seen = new Set();
const uniqueBlocks = [];
for (const block of blocks) {
if (seen.has(block.rxxId)) {
diagnostics.push({ code: "duplicate_rxx_id", rxxId: block.rxxId, lineNumber: block.lineNumber });
continue;
}
seen.add(block.rxxId);
uniqueBlocks.push(block);
}
const taskIds = new Set(uniqueBlocks.map((block) => block.rxxId));
const tasks = uniqueBlocks.map((block, index) => {
const parentRxxId = parentRxxIdFor(block.rxxId);
const parentTaskRef = parentRxxId && taskIds.has(parentRxxId) ? taskRefForRxx(context.sourceId, context.fileRef, parentRxxId) : null;
if (parentRxxId && !parentTaskRef) diagnostics.push({ code: "missing_parent_rxx_id", rxxId: block.rxxId, parentRxxId, lineNumber: block.lineNumber });
const bodyText = lines.slice(block.lineIndex + 1, block.endLine).join("\n");
const bodyPreview = bodyPreviewFromLines(lines, block.lineIndex + 1, block.endLine);
const title = cleanTaskTitle(block.title || bodyPreview || block.rxxId);
return {
taskRef: taskRefForRxx(context.sourceId, context.fileRef, block.rxxId),
taskId: block.rxxId,
rxxId: block.rxxId,
projectId: context.projectId,
sourceId: context.sourceId,
fileRef: context.fileRef,
relativePath: context.relativePath,
title,
status: block.status,
parentTaskRef,
parentRxxId: parentTaskRef ? parentRxxId : null,
depth: rxxDepth(block.rxxId),
lineNumber: block.lineNumber,
ordinal: index + 1,
linkCount: countMarkdownLinks(bodyText),
sourceFingerprint: context.fingerprint,
parserMode: "rxx",
headingLevel: block.hashes.length,
statusMarker: block.statusMarker,
bodyPreview
};
});
return { tasks, diagnostics };
}
function parseLegacyCheckboxTasks(lines, context) {
const tasks = [];
const parentStack = [];
let ordinal = 0;
@@ -56,16 +237,16 @@ export function parseMdtodoDocument(markdown, options = {}) {
const depth = Math.max(0, Math.floor(match[1].replace(/\t/gu, " ").length / 2));
const checkbox = match[2];
const title = cleanTaskTitle(match[3]);
const taskId = `t${index + 1}_${shortHash(`${relativePath}:${index + 1}:${title}`)}`;
const taskId = `t${index + 1}_${shortHash(`${context.relativePath}:${index + 1}:${title}`)}`;
const parentTaskRef = parentStack[depth - 1] ?? null;
const taskRef = `mdtodo:${sourceId}:${fileRef}:${taskId}`;
const taskRef = `mdtodo:${context.sourceId}:${context.fileRef}:${taskId}`;
const task = {
taskRef,
taskId,
projectId,
sourceId,
fileRef,
relativePath,
projectId: context.projectId,
sourceId: context.sourceId,
fileRef: context.fileRef,
relativePath: context.relativePath,
title,
status: checkbox.toLowerCase() === "x" ? "done" : checkbox === "-" ? "blocked" : "open",
parentTaskRef,
@@ -73,38 +254,175 @@ export function parseMdtodoDocument(markdown, options = {}) {
lineNumber: index + 1,
ordinal,
linkCount: 0,
sourceFingerprint: fingerprint
sourceFingerprint: context.fingerprint,
parserMode: "legacy-checkbox"
};
tasks.push(task);
parentStack[depth] = taskRef;
parentStack.length = depth + 1;
}
return { tasks, diagnostics: [] };
}
function collectRxxBlocks(lines) {
const headings = [];
for (let index = 0; index < lines.length; index += 1) {
const heading = parseRxxHeading(lines[index]);
if (heading) headings.push({ ...heading, lineIndex: index, lineNumber: index + 1 });
}
return headings.map((heading, index) => ({
...heading,
endLine: headings[index + 1]?.lineIndex ?? lines.length
}));
}
function parseRxxHeading(line) {
const match = rxxHeadingPattern.exec(String(line ?? ""));
if (!match) return null;
const parsed = parseHeadingRest(match[3] ?? "");
return {
sourceId,
fileRef,
relativePath,
fingerprint,
projectId,
title: documentTitle(markdown, relativePath),
taskCount: tasks.length,
tasks,
document: {
source: "markdown-file",
relativePath,
firstHeading: documentTitle(markdown, relativePath),
valuesRedacted: true
}
hashes: match[1],
rxxId: normalizeRxxId(match[2]),
status: parsed.status,
statusMarker: parsed.statusMarker,
title: parsed.title
};
}
export function documentFileRef(relativePath) {
return `file_${shortHash(normalizeRelativePath(relativePath), 16)}`;
function parseHeadingRest(value) {
const rest = String(value ?? "").trim();
let status = "open";
let statusMarker = null;
for (const match of rest.matchAll(/\[([^\]]+)\]/gu)) {
const candidate = normalizeStatus(match[1]);
if (!candidate) continue;
if (statusRank(candidate) >= statusRank(status)) {
status = candidate;
statusMarker = match[1];
}
}
return {
status,
statusMarker,
title: stripStatusMarkers(rest)
};
}
function buildAddTaskMutation(lines, blocks, operation) {
const title = cleanTaskTitle(operation.title);
const status = requireStatus(operation.status ?? "open");
const parentRxxId = normalizeOptionalRxxId(operation.parentRxxId);
const afterRxxId = normalizeOptionalRxxId(operation.afterRxxId);
let rxxId;
let headingLevel;
let insertPos;
if (parentRxxId) {
const parent = requireRxxBlock(blocks, parentRxxId);
rxxId = nextChildRxxId(blocks, parent.rxxId);
headingLevel = headingLevelForRxx(rxxId);
insertPos = subtreeEndLine(blocks, parent.rxxId);
} else if (afterRxxId) {
const after = requireRxxBlock(blocks, afterRxxId);
const siblingParent = parentRxxIdFor(after.rxxId);
rxxId = siblingParent ? nextChildRxxId(blocks, siblingParent) : nextRootRxxId(blocks);
headingLevel = headingLevelForRxx(rxxId);
insertPos = subtreeEndLine(blocks, after.rxxId);
} else {
rxxId = nextRootRxxId(blocks);
headingLevel = headingLevelForRxx(rxxId);
insertPos = logicalDocumentEnd(lines);
}
const bodyLines = normalizeBodyLines(operation.body);
const taskLines = [formatRxxHeading("#".repeat(headingLevel), rxxId, status, title)];
if (bodyLines.length > 0) taskLines.push("", ...bodyLines);
return { rxxId, insertPos, lines: taskLines };
}
function requireRxxBlock(blocks, value) {
const rxxId = normalizeOptionalRxxId(value);
if (!rxxId) throw mutationError("rxx_id_required", "A valid Rxx id is required");
const block = blocks.find((candidate) => candidate.rxxId === rxxId);
if (!block) throw mutationError("rxx_id_not_found", `MDTODO task ${rxxId} was not found`, { rxxId });
return block;
}
function insertLines(lines, insertPos, newLines) {
const next = lines.slice();
const insertion = [];
if (insertPos > 0 && String(next[insertPos - 1] ?? "").trim() !== "") insertion.push("");
insertion.push(...newLines);
if (insertPos < next.length && String(next[insertPos] ?? "").trim() !== "") insertion.push("");
next.splice(insertPos, 0, ...insertion);
return next;
}
function bodyReplacementLines(value) {
const lines = normalizeBodyLines(value);
if (lines.length === 0) return [];
return ["", ...lines, ""];
}
function normalizeBodyLines(value) {
const text = String(value ?? "").replace(/\r\n/gu, "\n").trim();
if (!text) return [];
return text.split("\n");
}
function formatRxxHeading(hashes, rxxId, status, title) {
const marker = statusMarkers.get(normalizeStatus(status) ?? "open");
const parts = [hashes, rxxId];
if (title) parts.push(cleanTaskTitle(title));
if (marker) parts.push(`[${marker}]`);
return parts.join(" ");
}
function nextRootRxxId(blocks) {
const next = Math.max(0, ...blocks
.map((block) => rxxSegments(block.rxxId))
.filter((segments) => segments.length === 1)
.map((segments) => segments[0])) + 1;
return `R${next}`;
}
function nextChildRxxId(blocks, parentRxxId) {
const parentSegments = rxxSegments(parentRxxId);
const prefix = `${parentRxxId}.`;
const next = Math.max(0, ...blocks
.filter((block) => block.rxxId.startsWith(prefix))
.map((block) => rxxSegments(block.rxxId))
.filter((segments) => segments.length === parentSegments.length + 1)
.map((segments) => segments.at(-1))) + 1;
return `${parentRxxId}.${next}`;
}
function subtreeEndLine(blocks, rxxId) {
const startIndex = blocks.findIndex((block) => block.rxxId === rxxId);
if (startIndex < 0) return 0;
let endLine = blocks[startIndex].endLine;
for (let index = startIndex + 1; index < blocks.length; index += 1) {
if (!isDescendantRxxId(blocks[index].rxxId, rxxId)) break;
endLine = blocks[index].endLine;
}
return endLine;
}
function logicalDocumentEnd(lines) {
let end = lines.length;
while (end > 0 && String(lines[end - 1] ?? "").trim() === "") end -= 1;
return end;
}
function normalizeMarkdownOutput(lines) {
const text = lines.join("\n").replace(/[\t ]+$/gmu, "").replace(/\n{4,}/gu, "\n\n\n").trimEnd();
return text ? `${text}\n` : "";
}
function isMdtodoCandidate(filePath, markdown) {
const basename = path.basename(filePath).toLowerCase();
return basename.includes("todo") || basename.includes("mdtodo") || String(markdown ?? "").split(/\r?\n/u).some((line) => taskLinePattern.test(line));
return basename.includes("todo") || basename.includes("mdtodo") || String(markdown ?? "").split(/\r?\n/u).some((line) => taskLinePattern.test(line) || rxxHeadingPattern.test(line));
}
async function walkMarkdown(root, current, files, maxFiles) {
@@ -140,6 +458,76 @@ function projectIdForRelativePath(relativePath) {
return `project_${safeToken(candidate.toLowerCase().replace(/[^a-z0-9_.:-]+/gu, "_"), "default")}`;
}
function taskRefForRxx(sourceId, fileRef, rxxId) {
return `mdtodo:${sourceId}:${fileRef}:${rxxId}`;
}
function parentRxxIdFor(rxxId) {
const segments = String(rxxId ?? "").split(".");
if (segments.length <= 1) return null;
return segments.slice(0, -1).join(".");
}
function isDescendantRxxId(candidate, parent) {
return candidate.startsWith(`${parent}.`);
}
function rxxDepth(rxxId) {
return Math.max(0, rxxSegments(rxxId).length - 1);
}
function headingLevelForRxx(rxxId) {
return Math.min(6, rxxDepth(rxxId) + 2);
}
function rxxSegments(rxxId) {
return String(rxxId ?? "").replace(/^R/iu, "").split(".").map((segment) => Number(segment)).filter((segment) => Number.isInteger(segment) && segment > 0);
}
function normalizeRxxId(value) {
return String(value ?? "").replace(/^r/iu, "R");
}
function normalizeOptionalRxxId(value) {
const text = normalizeRxxId(value).trim();
return rxxIdPattern.test(text) ? text : null;
}
function normalizeStatus(value) {
const key = String(value ?? "open").trim().toLowerCase().replace(/[\s-]+/gu, "_");
return statusAliases.get(key) ?? null;
}
function requireStatus(value) {
const status = normalizeStatus(value);
if (!status) throw mutationError("invalid_status", `Unsupported MDTODO status: ${String(value ?? "")}`);
return status;
}
function statusRank(status) {
if (status === "done") return 3;
if (status === "in_progress") return 2;
if (status === "blocked") return 1;
return 0;
}
function stripStatusMarkers(value) {
return String(value ?? "").replace(/\s*\[([^\]]+)\]/gu, (match, label) => normalizeStatus(label) ? "" : match).trim();
}
function countMarkdownLinks(value) {
return [...String(value ?? "").matchAll(markdownLinkPattern)].length;
}
function bodyPreviewFromLines(lines, start, end) {
for (const line of lines.slice(start, end)) {
const text = String(line ?? "").trim();
if (!text || rxxHeadingPattern.test(text)) continue;
return text.replace(/^[-*+]\s+/u, "").slice(0, 240);
}
return "";
}
function cleanTaskTitle(value) {
return String(value ?? "").replace(/\s+<!--.*?-->\s*$/u, "").trim().slice(0, 500) || "Untitled task";
}
@@ -165,3 +553,10 @@ function positiveInteger(value, fallback) {
const number = Number(value);
return Number.isInteger(number) && number > 0 ? number : fallback;
}
function mutationError(code, message, details = {}) {
const error = new Error(message);
error.code = code;
Object.assign(error, details);
return error;
}
+2 -1
View File
@@ -10,7 +10,7 @@ import { createProjectManagementStore } from "./store.ts";
test("project management app exposes MDTODO read model and Workbench link writes", async () => {
const root = await mkdtemp(join(tmpdir(), "hwlab-project-management-"));
try {
await writeFile(join(root, "MDTODO.md"), "# Demo\n\n- [ ] Launch project nav\n- [x] Wire read model\n", "utf8");
await writeFile(join(root, "MDTODO.md"), "# Demo\n\n## R1 Launch project nav\n\n## R2 Wire read model [completed]\n", "utf8");
const app = createProjectManagementApp({
env: {},
store: createProjectManagementStore({ kind: "memory" }),
@@ -32,6 +32,7 @@ test("project management app exposes MDTODO read model and Workbench link writes
expect(files.files[0].taskCount).toBe(2);
const tasks = await json(app, "/v1/project-management/mdtodo/tasks?sourceId=test-mdtodo");
expect(tasks.tasks.map((task) => task.taskId)).toEqual(["R1", "R2"]);
expect(tasks.tasks.map((task) => task.status)).toEqual(["open", "done"]);
expect(tasks.tasks.every((task) => task.projectId === "project_test")).toBe(true);