@stll/folio-core 0.11.0 → 0.13.0

package/dist/version-comparison.js

@@ -160,12 +160,14 @@ const pairByStableId = (base, revised) => {
 const MAX_LCS_CELLS = 4e6;
 /** True when an `unpairedBaseCount * unpairedRevisedCount` LCS table would exceed {@link MAX_LCS_CELLS}. */
 const exceedsLcsBudget = (unpairedBaseCount, unpairedRevisedCount) => unpairedBaseCount * unpairedRevisedCount > MAX_LCS_CELLS;
+const createLcsBudget = () => ({ remainingCells: MAX_LCS_CELLS });
 /** Pass 2: order-preserving LCS by exact text equality over the blocks pass 1 left unpaired. */
-const pairByExactText = (base, revised) => {
+const pairByExactText = (base, revised, lcsBudget) => {
 	const m = base.length;
 	const n = revised.length;
 	if (m === 0 || n === 0) return [];
-	if (exceedsLcsBudget(m, n)) return [];
+	if (exceedsLcsBudget(m, n) || lcsBudget.remainingCells <= 0) return [];
+	lcsBudget.remainingCells -= m * n;
 	const baseTexts = base.map(({ block }) => block.text);
 	const revisedTexts = revised.map(({ block }) => block.text);
 	const stride = n + 1;
@@ -205,8 +207,14 @@ const pairByExactText = (base, revised) => {
 * snapshots and flatten it into an ordered event stream. Shared by
 * {@link compareDocxVersions} and the redline generator so both interpret
 * one document walk instead of re-deriving it.
+*
+* `lcsBudget` defaults to a fresh, single-call budget so a caller comparing
+* one block pair in isolation (the redline generator) behaves exactly as
+* before. {@link compareDocxVersions} passes one budget object shared across
+* every story pair instead, so pass 2's cell allowance is aggregate across
+* the whole comparison rather than reset per story.
 */
-const alignFolioBlocks = (baseBlocks, revisedBlocks) => {
+const alignFolioBlocks = (baseBlocks, revisedBlocks, lcsBudget = createLcsBudget()) => {
 	const stableIdAnchors = pairByStableId(baseBlocks, revisedBlocks);
 	const usedBaseIndexes = new Set(stableIdAnchors.map((anchor) => anchor.baseIndex));
 	const usedRevisedIndexes = new Set(stableIdAnchors.map((anchor) => anchor.revisedIndex));
@@ -224,7 +232,7 @@ const alignFolioBlocks = (baseBlocks, revisedBlocks) => {
 			index: blockIndex
 		});
 	});
-	const exactTextAnchors = pairByExactText(baseRemaining, revisedRemaining);
+	const exactTextAnchors = pairByExactText(baseRemaining, revisedRemaining, lcsBudget);
 	const anchors = longestIncreasingByRevisedIndex([...stableIdAnchors, ...exactTextAnchors].toSorted((a, b) => a.baseIndex - b.baseIndex));
 	const events = [];
 	/** Pass 3: positionally zip the leftover blocks in one gap between anchors. */
@@ -331,6 +339,20 @@ const meetsMoveWordCount = (text) => {
 	return false;
 };
 /**
+* Cap on how many same-text `deleted` candidates {@link detectMoves} queues
+* per distinct text. Duplicated boilerplate at or above the word floor
+* (e.g. a repeated long clause) could otherwise grow one text's candidate
+* list without bound; past this cap, further same-text deletions are simply
+* left as `deleted` (never matched to a move) instead of queued.
+*/
+const MAX_MOVE_CANDIDATES_PER_TEXT = 1e4;
+const dequeueDeletedIndex = (queue) => {
+	if (!queue || queue.head >= queue.items.length) return;
+	const index = queue.items[queue.head];
+	queue.head += 1;
+	return index;
+};
+/**
 * Re-classify `deleted` + `added` pairs with identical text as
 * `movedFrom` / `movedTo` entries sharing a `moveGroupId`, in place, so each
 * side keeps its slot in the revised-side document order. Matching is FIFO
@@ -340,17 +362,22 @@ const meetsMoveWordCount = (text) => {
 const detectMoves = (changes, counts, firstMoveGroupId) => {
 	const deletedIndexesByText = /* @__PURE__ */ new Map();
 	changes.forEach((change, index) => {
-		if (change.type === "deleted" && meetsMoveWordCount(change.text)) {
-			const queue = deletedIndexesByText.get(change.text) ?? [];
-			queue.push(index);
-			deletedIndexesByText.set(change.text, queue);
+		if (change.type !== "deleted" || !meetsMoveWordCount(change.text)) return;
+		const queue = deletedIndexesByText.get(change.text);
+		if (!queue) {
+			deletedIndexesByText.set(change.text, {
+				items: [index],
+				head: 0
+			});
+			return;
 		}
+		if (queue.items.length < MAX_MOVE_CANDIDATES_PER_TEXT) queue.items.push(index);
 	});
 	if (deletedIndexesByText.size === 0) return;
 	let moveGroupId = firstMoveGroupId - 1;
 	changes.forEach((change, index) => {
 		if (change.type !== "added") return;
-		const deletedIndex = deletedIndexesByText.get(change.text)?.shift();
+		const deletedIndex = dequeueDeletedIndex(deletedIndexesByText.get(change.text));
 		if (deletedIndex === void 0) return;
 		const deleted = changes[deletedIndex];
 		if (!deleted || deleted.type !== "deleted") return;
@@ -394,10 +421,10 @@ const addSummaryCounts = (target, source) => {
 	target.metadataChanged += source.metadataChanged;
 	target.unchanged += source.unchanged;
 };
-const compareStoryBlocks = ({ baseStory, revisedStory, baseBlocks, revisedBlocks, firstMoveGroupId, includeText, includeFormatting }) => {
+const compareStoryBlocks = ({ baseStory, revisedStory, baseBlocks, revisedBlocks, firstMoveGroupId, includeText, includeFormatting, lcsBudget }) => {
 	const changes = [];
 	const counts = createSummaryCounts();
-	for (const event of alignFolioBlocks(baseBlocks, revisedBlocks)) {
+	for (const event of alignFolioBlocks(baseBlocks, revisedBlocks, lcsBudget)) {
 		if (event.type === "pair") {
 			if (!baseStory || !revisedStory) panic("A paired comparison event requires both story handles");
 			const { baseBlock, revisedBlock } = event;
@@ -554,6 +581,7 @@ const compareDocxVersions = async (base, revised, options = {}) => {
 	const baseStories = baseReviewer.listStories().map(({ handle }) => handle);
 	const revisedStories = revisedReviewer.listStories().map(({ handle }) => handle);
 	let nextMoveGroupId = 1;
+	const lcsBudget = createLcsBudget();
 	for (const pair of pairFolioDocumentStories(baseStories, revisedStories)) {
 		const baseBlocks = pair.baseStory ? baseReviewer.readReviewedStory({
 			story: pair.baseStory,
@@ -569,7 +597,8 @@ const compareDocxVersions = async (base, revised, options = {}) => {
 			revisedBlocks,
 			firstMoveGroupId: nextMoveGroupId,
 			includeText: scopes.has("text"),
-			includeFormatting: scopes.has("formatting")
+			includeFormatting: scopes.has("formatting"),
+			lcsBudget
 		});
 		stories.push(storyDiff);
 		for (const change of storyDiff.changes) changes.push(change);

package/dist/watermark/index.d.ts

@@ -1,6 +1,16 @@
 import { document_d_exports } from "../types/document.js";
 
 //#region src/watermark/index.d.ts
+/**
+ * Whether `target` is safe to save as a watermark picture's external image
+ * relationship. A watermark dialog lets an author type an arbitrary string
+ * for this field; without a scheme check it becomes a `TargetMode="External"`
+ * relationship verbatim (see `docx/rezip.ts`), which would let a `file:` URL
+ * or UNC path (`\\host\share\...`) into the exported `.docx` — a resource
+ * whatever later opens the file (e.g. Word) would try to resolve. Only
+ * `http:`/`https:` targets are allowed.
+ */
+declare function isAllowedExternalWatermarkImageUrl(target: string): boolean;
 /**
  * Read the document's watermark. Walks every header part and returns
  * the first watermark encountered (header insertion order). Returns
@@ -36,4 +46,4 @@ declare function ensureWatermarkHeaderCoverage(doc: document_d_exports.Document,
 type PictureWatermark = document_d_exports.PictureWatermark;
 type TextWatermark = document_d_exports.TextWatermark;
 type Watermark = document_d_exports.Watermark;
-export { type PictureWatermark, type TextWatermark, type Watermark, ensureWatermarkHeaderCoverage, getDocumentWatermark, setDocumentWatermark };
+export { type PictureWatermark, type TextWatermark, type Watermark, ensureWatermarkHeaderCoverage, getDocumentWatermark, isAllowedExternalWatermarkImageUrl, setDocumentWatermark };

package/dist/watermark/index.js

@@ -1,5 +1,26 @@
 //#region src/watermark/index.ts
 /**
+* Schemes allowed for a picture watermark's external image target
+* (`TargetMode="External"` in the saved package's relationships).
+*/
+const ALLOWED_EXTERNAL_IMAGE_SCHEMES = /* @__PURE__ */ new Set(["http:", "https:"]);
+/**
+* Whether `target` is safe to save as a watermark picture's external image
+* relationship. A watermark dialog lets an author type an arbitrary string
+* for this field; without a scheme check it becomes a `TargetMode="External"`
+* relationship verbatim (see `docx/rezip.ts`), which would let a `file:` URL
+* or UNC path (`\\host\share\...`) into the exported `.docx` — a resource
+* whatever later opens the file (e.g. Word) would try to resolve. Only
+* `http:`/`https:` targets are allowed.
+*/
+function isAllowedExternalWatermarkImageUrl(target) {
+	try {
+		return ALLOWED_EXTERNAL_IMAGE_SCHEMES.has(new URL(target).protocol);
+	} catch {
+		return false;
+	}
+}
+/**
 * Read the document's watermark. Walks every header part and returns
 * the first watermark encountered (header insertion order). Returns
 * `undefined` when no header carries one.
@@ -146,4 +167,4 @@ function ensureWatermarkHeaderCoverage(doc, watermark) {
 	};
 }
 //#endregion
-export { ensureWatermarkHeaderCoverage, getDocumentWatermark, setDocumentWatermark };
+export { ensureWatermarkHeaderCoverage, getDocumentWatermark, isAllowedExternalWatermarkImageUrl, setDocumentWatermark };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@stll/folio-core",
-  "version": "0.11.0",
+  "version": "0.13.0",
   "description": "Headless, framework-neutral core of folio: the OOXML (.docx) parser, document model, ProseMirror integration, and page-layout engine. No React.",
   "keywords": [
     "document-model",
@@ -107,11 +107,12 @@
     "perf": "bun scripts/profile-editor.ts"
   },
   "dependencies": {
-    "@stll/docx-core": "^0.1.0",
+    "@stll/docx-core": "^0.4.0",
     "@stll/docx-utils": "^0.1.0",
     "@stll/template-conditions": "^0.1.0",
     "better-result": "2.9.2",
     "csstype": "^3.1.3",
+    "dompurify": "^3.4.12",
     "fast-xml-parser": "^5.9.3",
     "hyphen": "1.14.1",
     "jszip": "3.10.1",