big cmap refactoring

This commit is contained in:
2026-09-03 08:40:15 +02:00
parent f0562a06cc
commit 2c141b6d3f
32 changed files with 5616 additions and 4216 deletions
+89 -1
View File
@@ -1,4 +1,13 @@
/* Versioned JSON interchange for the Racket Wiki CMap model. */
/**
* Versioned JSON interchange for the Racket Wiki CMap model.
*
* The bundle separates shared concept content from map-local placement data.
* `buildBundle` follows map and page references, embeds the referenced page
* attachments, and validates the resulting object. Import code calls
* `validateBundle` first and then uses `preparedMapDocument` to join shared
* concepts back into a map document. The helpers in this module deliberately
* return detached JSON values so callers cannot mutate an input map or bundle.
*/
const FORMAT = "racket-wiki-cmap-bundle";
const FORMAT_VERSION = 1;
@@ -12,10 +21,15 @@
];
const PLACEMENT_CONTENT_KEYS = new Set(CONCEPT_KEYS.filter((key) => key !== "id"));
/** Return a detached JSON-compatible copy while preserving undefined. */
function clone(value) {
return value === undefined ? undefined : JSON.parse(JSON.stringify(value));
}
/**
* Check a wiki page reference in either `slug` or `namespace:slug` form.
* The check is shared by concept links, map metadata and exported pages.
*/
function validPageReference(value) {
if (typeof value !== "string") return false;
const separator = value.indexOf(":");
@@ -25,6 +39,7 @@
return namespace.length <= 80 && SLUG.test(namespace) && SLUG.test(slug);
}
/** Accept only complete HTTP(S) URLs for external concept links. */
function validExternalUrl(value) {
try {
const url = new URL(String(value));
@@ -34,6 +49,11 @@
}
}
/**
* Decode the one or two historical JSON string wrappers used by storage.
* Invalid or non-object values become an empty document; bundle validation
* remains responsible for rejecting malformed imported data.
*/
function decodedDocument(value) {
let documentValue = value;
for (let attempt = 0; attempt < 2 && typeof documentValue === "string"; attempt += 1) {
@@ -43,6 +63,7 @@
documentValue : {};
}
/** Copy only shared concept fields into the bundle-level concept table. */
function conceptContent(value) {
const result = {};
for (const key of CONCEPT_KEYS) {
@@ -51,18 +72,21 @@
return result;
}
/** Index valid document concepts by their stable identity. */
function conceptsById(documentValue) {
return new Map((Array.isArray(documentValue.concepts) ? documentValue.concepts : [])
.filter((concept) => concept && typeof concept.id === "string" && concept.id)
.map((concept) => [concept.id, concept]));
}
/** Return unique concept ids used by non-phrase placements. */
function itemConceptIds(documentValue) {
return [...new Set((Array.isArray(documentValue.items) ? documentValue.items : [])
.filter((item) => item && item.kind !== "phrase" && typeof item.conceptId === "string" && item.conceptId)
.map((item) => item.conceptId))];
}
/** Find linked CMap slugs in the concepts actually placed on a document. */
function linkedMapSlugs(documentValue) {
const concepts = conceptsById(documentValue);
return [...new Set(itemConceptIds(documentValue)
@@ -70,6 +94,11 @@
.filter(Boolean))];
}
/**
* Collect page references from map metadata and placed concepts.
* Unplaced concept records are intentionally ignored because they are not
* part of the visible map dependency graph.
*/
function linkedPageReferences(documentValue) {
const references = new Set();
const metadata = documentValue.metadata && typeof documentValue.metadata === "object" ?
@@ -87,6 +116,11 @@
return references;
}
/**
* Extract local upload URLs from Markdown and HTML-like markup.
* A Set removes duplicates; the final prefix check removes a shorter URL
* accidentally captured from a URL containing a space.
*/
function attachmentUrls(markdown) {
const source = String(markdown || "");
const urls = new Set();
@@ -102,6 +136,7 @@
other !== url && other.startsWith(`${url} `)));
}
/** Derive a readable fallback filename from an upload URL. */
function attachmentName(url) {
const encoded = String(url || "").split("/").at(-1) || "attachment.bin";
try {
@@ -111,6 +146,7 @@
}
}
/** Replace upload URLs after imported attachments receive new server URLs. */
function replaceAttachmentUrls(markdown, replacements) {
let result = String(markdown || "");
const entries = replacements instanceof Map ? [...replacements.entries()] :
@@ -123,6 +159,11 @@
return result;
}
/**
* Strip shared concept fields from placements while preserving layout data.
* It also keeps only connectors whose two local item endpoints still exist,
* because export must not emit dangling layout relations.
*/
function placementDocument(documentValue) {
const documentCopy = clone(decodedDocument(documentValue));
const ids = itemConceptIds(documentCopy);
@@ -149,6 +190,10 @@
return documentCopy;
}
/**
* Convert one wiki page and all uploads referenced by its Markdown into a
* bundle page record. The attachment loader is called once per unique URL.
*/
async function pageRecord(page, requestedReference, loadAttachment) {
const markdown = String(page.markdown || "");
const attachments = [];
@@ -176,6 +221,24 @@
};
}
/**
* Build and validate a complete export bundle.
*
* @param {object} options Export options and asynchronous repository loaders.
* @param {object} options.rootMap Root map with `slug`, `title` and `document`.
* @param {Function} options.loadConceptMap Loads a linked map by slug.
* @param {Function} options.loadWikiPage Loads a linked wiki page by reference.
* @param {Function} [options.loadAttachment] Loads base64 upload content.
* @param {number} [options.maxDepth=0] Maximum depth for linked CMaps.
* @returns {Promise<object>} A validated, detached interchange bundle.
* @throws {Error} When required loaders are absent or a loaded record is invalid.
* @sideeffects Calls the supplied map, page and attachment loaders.
*
* `collectMap` traverses map dependencies and uses `placementDocument` for
* layout-only map records. It collects shared concepts and page references
* separately; page records are then built and the final aggregate is checked
* by `validateBundle` before it is returned.
*/
async function buildBundle(options) {
if (!options?.rootMap?.slug) throw new Error("A root CMap is required.");
if (typeof options.loadConceptMap !== "function") throw new Error("loadConceptMap is required.");
@@ -248,6 +311,18 @@
return bundle;
}
/**
* Validate the complete bundle contract and return the original bundle.
*
* @param {object} bundle Candidate bundle to validate.
* @returns {object} The same bundle object after successful validation.
* @throws {Error} With `validationErrors` when one or more contract checks fail.
*
* Validation also builds the concept, map, item and page-reference indexes
* needed to detect duplicate identities, dangling references and missing
* linked pages. Up to twenty errors are included in the message while the
* complete list remains available on `error.validationErrors`.
*/
function validateBundle(bundle) {
const errors = [];
const issue = (path, message) => errors.push(`${path}: ${message}`);
@@ -260,6 +335,7 @@
if (!Array.isArray(bundle.concepts)) issue("concepts", "must be an array");
if (!Array.isArray(bundle.pages)) issue("pages", "must be an array");
// Validate the shared concept table and collect its page dependencies.
const conceptIds = new Set();
const conceptLabels = new Set();
const usedConceptIds = new Set();
@@ -297,6 +373,7 @@
}
}
// Validate map identities, placements, connectors and map-level page links.
const mapSlugs = new Set();
for (const [mapIndex, cmap] of (Array.isArray(bundle.cmaps) ? bundle.cmaps : []).entries()) {
const path = `cmaps[${mapIndex}]`;
@@ -362,6 +439,7 @@
if (!usedConceptIds.has(id)) issue(`concepts[id=${id}]`, "must occur as a diagram placement");
}
// Validate page records and their embedded attachment payloads.
const pageReferences = new Set();
for (const [index, page] of (Array.isArray(bundle.pages) ? bundle.pages : []).entries()) {
const path = `pages[${index}]`;
@@ -405,6 +483,7 @@
}
}
}
// Reconcile all collected page dependencies with present or declared-missing pages.
const explicitlyMissingPages = new Set(Array.isArray(bundle.missing?.pages) ? bundle.missing.pages : []);
for (const reference of linkedPageReferences) {
if (!pageReferences.has(reference) && !explicitlyMissingPages.has(reference)) {
@@ -412,6 +491,7 @@
}
}
// Report all discovered issues together so import callers can repair a bundle in one pass.
if (errors.length) {
const error = new Error(`Invalid CMap bundle:\n${errors.slice(0, 20).join("\n")}`);
error.validationErrors = errors;
@@ -420,6 +500,14 @@
return bundle;
}
/**
* Rejoin bundle-level concept content with one placement-only map document.
*
* @param {object} bundle A bundle that satisfies `validateBundle`.
* @param {object} cmap One entry from `bundle.cmaps`.
* @returns {object} A detached document suitable for repository import.
* @throws {Error} When the bundle is invalid or a referenced concept is absent.
*/
function preparedMapDocument(bundle, cmap) {
validateBundle(bundle);
const byId = new Map(bundle.concepts.map((concept) => [concept.id, concept]));