@docx4j/core-ts 0.2.1 → 0.2.2

package/README.md

@@ -231,7 +231,9 @@ body.acceptAll();     // or rejectAll(), or accept()/reject() one at a time
 ```
 
 `getTrackedChanges()` is also on `Paragraph` and `Range`. Text and searches read the accepted
-view throughout, so an agent sees the document as it will read once the changes are taken.
+view throughout, so an agent sees the document as it will read once the changes are taken. A moved
+paragraph is listed as its two halves, and accepting or rejecting either resolves the whole move,
+as Word's Review tab does.
 
 To record one caller's edits without touching the document's own setting - an agent working in a
 document whose `w:trackRevisions` belongs to its owner - use `withTracking`. Nothing is written

package/dist/model/content/Range.d.mts

@@ -118,6 +118,9 @@ export declare class Range {
      * nothing.
      */
     private touches;
+    private touchesSpan;
+    /** The text a hyperlink holds, as [start, end) in the paragraph, or undefined when it holds none. */
+    private spanOf;
     /**
      * The `Hyperlink` character style on every run of a hyperlink, in place of any character style
      * and with the direct colour removed, or off again when the link goes: what Word does (CR-002

package/dist/model/content/Range.mjs

@@ -271,6 +271,22 @@ export class Range {
     set hyperlink(value) {
         var _a, _b, _c;
         const paragraph = this.paragraph;
+        // A range that crosses the edge of a link takes the whole of that link into the new one; a range
+        // inside a link replaces it with a link over the range alone. Measured in Word through Office JS
+        // (test/README.md check 16, cases F and G; CR-002 section 25): "alpha beta" linked and then
+        // "beta gamma" is one link over "alpha beta gamma", where "alpha beta gamma" linked and then
+        // "beta" leaves "alpha " and " gamma" unlinked.
+        let start = this.start;
+        let end = this.end;
+        for (const holder of hyperlinkHolders(paragraph)) {
+            const span = this.spanOf(holder);
+            if (span === undefined || !this.touchesSpan(span))
+                continue;
+            if (span[0] <= this.start && this.end <= span[1])
+                continue;
+            start = Math.min(start, span[0]);
+            end = Math.max(end, span[1]);
+        }
         this.removeHyperlinks();
         if (value === '')
             return;
@@ -287,13 +303,13 @@ export class Range {
             holder.value.anchor = location;
         if (address !== '')
             holder.value.id = this.addHyperlinkRelationship(address);
-        if (this.start === this.end) {
-            paragraph.insertItemsAt(this.start, [holder]);
+        if (start === end) {
+            paragraph.insertItemsAt(start, [holder]);
             return;
         }
-        paragraph.splitAt(this.start);
-        paragraph.splitAt(this.end);
-        const segments = paragraph.segments().filter((seg) => seg.start >= this.start && seg.end <= this.end);
+        paragraph.splitAt(start);
+        paragraph.splitAt(end);
+        const segments = paragraph.segments().filter((seg) => seg.start >= start && seg.end <= end);
         if (segments.length === 0)
             throw new Docx4JException('This range covers no run');
         const owner = segments[0].runOwner;
@@ -342,6 +358,14 @@ export class Range {
      * nothing.
      */
     touches(holder) {
+        const span = this.spanOf(holder);
+        return span !== undefined && this.touchesSpan(span);
+    }
+    touchesSpan([start, end]) {
+        return this.start === this.end ? start <= this.start && this.start < end : start < this.end && end > this.start;
+    }
+    /** The text a hyperlink holds, as [start, end) in the paragraph, or undefined when it holds none. */
+    spanOf(holder) {
         let start;
         let end = 0;
         for (const segment of this.paragraph.segments()) {
@@ -350,9 +374,7 @@ export class Range {
             start !== null && start !== void 0 ? start : (start = segment.start);
             end = segment.end;
         }
-        if (start === undefined)
-            return false;
-        return this.start === this.end ? start <= this.start && this.start < end : start < this.end && end > this.start;
+        return start === undefined ? undefined : [start, end];
     }
     /**
      * The `Hyperlink` character style on every run of a hyperlink, in place of any character style

package/dist/model/content/TrackedChange.d.mts

@@ -70,9 +70,15 @@ export declare class TrackedChange {
     get text(): string;
     /** The range the change covers; a deletion is a collapsed range where it sits in the accepted text. */
     getRange(): Range | undefined;
-    /** Keeps the change: a `w:ins` is unwrapped, a `w:del` removed, a deleted mark joins the paragraphs. */
+    /**
+     * Keeps the change: a `w:ins` is unwrapped, a `w:del` removed, a deleted mark joins the paragraphs.
+     * Either half of a move keeps the whole move, as Word's Review tab does (CR-002 section 29).
+     */
     accept(): void;
-    /** Puts back what was there: a `w:ins` is removed, a `w:del` restored, a `w:rPrChange` re-applied. */
+    /**
+     * Puts back what was there: a `w:ins` is removed, a `w:del` restored, a `w:rPrChange` re-applied.
+     * Either half of a move rejects the whole move, as Word's Review tab does (CR-002 section 29).
+     */
     reject(): void;
     private requireParagraph;
 }
@@ -86,7 +92,7 @@ export declare class TrackedChange {
  * with: the paragraph then goes, its content joining the one before. With neither neighbour the
  * mark is simply dropped.
  */
-export declare function joinWithNext(paragraph: Paragraph, fallbackToPrevious?: boolean): void;
+export declare function joinWithNext(paragraph: Pick<Paragraph, 'container' | 'element' | 'p'>, fallbackToPrevious?: boolean): void;
 /** The tracked changes of one paragraph, in document order. */
 export declare function trackedChangesOfParagraph(paragraph: Paragraph): TrackedChange[];
 /**

package/dist/model/content/TrackedChange.mjs

@@ -73,12 +73,17 @@ export class TrackedChange {
         }
         return p.getRange('Whole');
     }
-    /** Keeps the change: a `w:ins` is unwrapped, a `w:del` removed, a deleted mark joins the paragraphs. */
+    /**
+     * Keeps the change: a `w:ins` is unwrapped, a `w:del` removed, a deleted mark joins the paragraphs.
+     * Either half of a move keeps the whole move, as Word's Review tab does (CR-002 section 29).
+     */
     accept() {
         var _a, _b, _c;
         const t = this.target;
         switch (t.kind) {
             case 'run':
+                if ((t.revision === 'moveFrom' || t.revision === 'moveTo') && resolveMove(this, 'accept'))
+                    return;
                 if (t.revision === 'ins' || t.revision === 'moveTo')
                     unwrap(t);
                 else
@@ -111,12 +116,17 @@ export class TrackedChange {
                 return;
         }
     }
-    /** Puts back what was there: a `w:ins` is removed, a `w:del` restored, a `w:rPrChange` re-applied. */
+    /**
+     * Puts back what was there: a `w:ins` is removed, a `w:del` restored, a `w:rPrChange` re-applied.
+     * Either half of a move rejects the whole move, as Word's Review tab does (CR-002 section 29).
+     */
     reject() {
         var _a;
         const t = this.target;
         switch (t.kind) {
             case 'run':
+                if ((t.revision === 'moveFrom' || t.revision === 'moveTo') && resolveMove(this, 'reject'))
+                    return;
                 if (t.revision === 'ins' || t.revision === 'moveTo')
                     remove(t.owner, t.element);
                 else
@@ -298,6 +308,129 @@ function acceptedOffsetOf(p, target) {
     visit(runItemsOf(p));
     return found >= 0 ? found : pos;
 }
+const RANGE_MARKERS = {
+    moveFromRangeStart: { kind: 'start', side: 'from' }, moveFromRangeEnd: { kind: 'end', side: 'from' },
+    moveToRangeStart: { kind: 'start', side: 'to' }, moveToRangeEnd: { kind: 'end', side: 'to' },
+};
+/**
+ * Accepts or rejects the whole move a run-level half belongs to, as Word's Review tab does from
+ * either half (`test/README.md` check 18, Word 16.0.20326.20158; Office JS cannot list a move at
+ * all there, check 17). Word pairs the halves by the `w:name` on their range starts
+ * (`w:moveFromRangeStart`, `w:moveToRangeStart`); `w:moveFrom` and `w:moveTo` carry none, so a half
+ * belongs to the move whose range of its side is open where it stands. Accepting keeps the
+ * destination and removes the source, paragraph mark and all; rejecting the other way round; the
+ * four range markers go either way. Returns false for a half in no named range - markup without
+ * range markers, or a half already resolved - which the caller then treats on its own, as before.
+ */
+function resolveMove(change, action) {
+    var _a, _b, _c;
+    const t = change.target;
+    if (t.kind !== 'run' || !change.paragraph)
+        return false;
+    const pieces = movePiecesOf(storyOf(change.paragraph.p));
+    const half = pieces.find((piece) => piece.kind === 'half' && piece.element === t.element);
+    if (!half || half.kind !== 'half' || half.names.length === 0)
+        return false;
+    const name = half.names[0];
+    const ids = new Set(pieces.filter((piece) => piece.kind === 'start' && piece.name === name).map((piece) => piece.id));
+    const mine = pieces.filter((piece) => (piece.kind === 'start' && piece.name === name) || (piece.kind === 'end' && ids.has(piece.id))
+        || ((piece.kind === 'half' || piece.kind === 'mark') && piece.names.includes(name)));
+    const kept = action === 'accept' ? 'to' : 'from';
+    // the markers first: a range end between two paragraphs would stop the paragraph join below
+    for (const piece of mine)
+        if (piece.kind === 'start' || piece.kind === 'end')
+            remove(piece.owner, piece.element);
+    for (const piece of mine) {
+        if (piece.kind !== 'half')
+            continue;
+        const target = { kind: 'run', revision: piece.side === 'from' ? 'moveFrom' : 'moveTo', element: piece.element, owner: piece.owner, value: piece.element.value };
+        if (piece.side === kept) {
+            if (piece.side === 'from')
+                restoreDeleted(target);
+            else
+                unwrap(target);
+        }
+        else
+            remove(piece.owner, piece.element);
+    }
+    // the paragraph marks last, last first, so that a join never disturbs one still to do
+    const marks = mine.filter((piece) => piece.kind === 'mark');
+    for (let i = marks.length - 1; i >= 0; i--) {
+        const mark = marks[i];
+        const p = mark.paragraph.value;
+        if (mark.side === kept) {
+            const rPr = (_a = p.pPr) === null || _a === void 0 ? void 0 : _a.rPr;
+            if (rPr)
+                delete rPr[mark.side === 'from' ? 'moveFrom' : 'moveTo'];
+            pruneParagraphProperties(p);
+        }
+        else if (((_b = p.content) !== null && _b !== void 0 ? _b : []).length === 0) {
+            remove(mark.container, mark.paragraph); // the moved paragraph, emptied: gone, as in Word
+        }
+        else {
+            const rPr = (_c = p.pPr) === null || _c === void 0 ? void 0 : _c.rPr;
+            if (rPr)
+                delete rPr[mark.side === 'from' ? 'moveFrom' : 'moveTo'];
+            joinWithNext({ container: mark.container, element: mark.paragraph, p }, mark.side === 'to');
+        }
+    }
+    return true;
+}
+/** The story a paragraph is in: the outermost array of block content above it (a body, a header, a note). */
+function storyOf(p) {
+    let story;
+    for (let o = p; o !== undefined; o = o.PARENT) {
+        const content = o.content;
+        if (Array.isArray(content) && o.TYPE_NAME !== 'org_docx4j_wml.P')
+            story = content;
+    }
+    return story !== null && story !== void 0 ? story : [];
+}
+/** Every piece of move markup in a story, in document order, each half and mark with the names of the ranges open over it. */
+function movePiecesOf(story) {
+    const out = [];
+    const open = { from: new Map(), to: new Map() };
+    const visit = (items) => {
+        var _a, _b, _c, _d, _e, _f;
+        for (const el of items) {
+            const marker = RANGE_MARKERS[(_b = (_a = el.name) === null || _a === void 0 ? void 0 : _a.localPart) !== null && _b !== void 0 ? _b : ''];
+            const value = el.value;
+            if (marker && value) {
+                const id = Number(value.id);
+                if (marker.kind === 'start') {
+                    open[marker.side].set(id, (_c = value.name) !== null && _c !== void 0 ? _c : '');
+                    out.push({ kind: 'start', side: marker.side, id, name: (_d = value.name) !== null && _d !== void 0 ? _d : '', element: el, owner: items });
+                }
+                else {
+                    open[marker.side].delete(id);
+                    out.push({ kind: 'end', side: marker.side, id, element: el, owner: items });
+                }
+                continue;
+            }
+            const revision = revisionKindOf(el);
+            if (revision === 'moveFrom' || revision === 'moveTo') {
+                const side = revision === 'moveFrom' ? 'from' : 'to';
+                out.push({ kind: 'half', side, element: el, owner: items, names: [...open[side].values()] });
+            }
+            const v = el.value;
+            if (typeof v !== 'object' || v === null)
+                continue;
+            const children = (_e = childrenOf(v)) !== null && _e !== void 0 ? _e : runItemsOf(v);
+            if (children)
+                visit(children);
+            if (typeNameOf(el) === 'org_docx4j_wml.P') {
+                const rPr = (_f = v.pPr) === null || _f === void 0 ? void 0 : _f.rPr;
+                for (const side of ['from', 'to']) {
+                    if (rPr === null || rPr === void 0 ? void 0 : rPr[side === 'from' ? 'moveFrom' : 'moveTo']) {
+                        out.push({ kind: 'mark', side, paragraph: el, container: items, names: [...open[side].values()] });
+                    }
+                }
+            }
+        }
+    };
+    visit(story);
+    return out;
+}
 // --- collecting ----------------------------------------------------------------------------
 /** The tracked changes of one paragraph, in document order. */
 export function trackedChangesOfParagraph(paragraph) {

package/dist/model/content/index.d.mts

@@ -13,7 +13,7 @@ export { contentOf, isFlatOpc, rewriteRelationshipIds, type OoxmlOptions } from
 export { type SearchOptions, searchPattern, findAll, matchesOf, expandReplacement } from './search.mjs';
 export { segmentsOf, runsOf, runItemsOf, childrenOf, rowsOf, cellsOf, BLOCK_LEVEL_TYPES, SDT_TYPES, runOf, paragraphOf, textOfView, revisionKindOf, type TextSegment, type Located, type TextView, type TextViewOptions, type RevisionKind, type RevisionHolder } from './tree.mjs';
 export { TrackedChange, trackedChangesOfParagraph, trackedChangesOfRow, joinWithNext, type TrackedChangeType, type TrackedChangeTarget } from './TrackedChange.mjs';
-export { ChangeTracker, trackerOf, calendarOf, dateOf, copyRPr, markDeleted, markInserted, toDeletedText, toRestoredText, restoreRPr, restorePPr, trackInsertedParagraph, trackInsertedTable, wrapNewRuns, paraRPrOf, rowPrOf, pruneParagraphProperties, type ChangeTrackingMode, type TrackingHost, type TrackingScope, } from './tracking.mjs';
+export { ChangeTracker, trackerOf, calendarOf, dateOf, copyRPr, markDeleted, markInserted, highestAnnotationId, MARKUP_TYPES, toDeletedText, toRestoredText, restoreRPr, restorePPr, trackInsertedParagraph, trackInsertedTable, wrapNewRuns, paraRPrOf, rowPrOf, pruneParagraphProperties, type ChangeTrackingMode, type TrackingHost, type TrackingScope, } from './tracking.mjs';
 export { wml, wmlOne, p, r, t, br, tab, tbl, tr, tc, sdt, sdtPr, nextSdtId, sdtProperty, sdtKindOf, inlinePicture, rPrToElements, rPrFromElements, textOf, walk, walkAll, find, linkParents, applyRunOptions, readRunOptions, UNDERLINE_TO_WML, isElement, typeNameOf, W_NS, type Element, type Wrapper, type WmlOptions, type Raw, type Interpolated, type RunOptions, type RunFormatting, type ParagraphOptions, type TableOptions, type CellOptions, type SdtKind, type SdtForm, type SdtOptions } from '@docx4j/generated-objects-ts/builders/wml';
 export { StylesFacade, type StylePackageLike } from './stylesFacade.mjs';
 export { bindSnapshot, boundsOf, pathOf, type Bound, type BoundSnapshot } from './binder.mjs';

package/dist/model/content/index.mjs

@@ -16,7 +16,7 @@ export { searchPattern, findAll, matchesOf, expandReplacement } from './search.m
 export { segmentsOf, runsOf, runItemsOf, childrenOf, rowsOf, cellsOf, BLOCK_LEVEL_TYPES, SDT_TYPES, runOf, paragraphOf, textOfView, revisionKindOf } from './tree.mjs';
 // change tracking (CR-002 phase F)
 export { TrackedChange, trackedChangesOfParagraph, trackedChangesOfRow, joinWithNext } from './TrackedChange.mjs';
-export { ChangeTracker, trackerOf, calendarOf, dateOf, copyRPr, markDeleted, markInserted, toDeletedText, toRestoredText, restoreRPr, restorePPr, trackInsertedParagraph, trackInsertedTable, wrapNewRuns, paraRPrOf, rowPrOf, pruneParagraphProperties, } from './tracking.mjs';
+export { ChangeTracker, trackerOf, calendarOf, dateOf, copyRPr, markDeleted, markInserted, highestAnnotationId, MARKUP_TYPES, toDeletedText, toRestoredText, restoreRPr, restorePPr, trackInsertedParagraph, trackInsertedTable, wrapNewRuns, paraRPrOf, rowPrOf, pruneParagraphProperties, } from './tracking.mjs';
 // The objects package's builders, re-exported so that one import serves the content API and the tree it works on.
 export { wml, wmlOne, p, r, t, br, tab, tbl, tr, tc, sdt, sdtPr, nextSdtId, sdtProperty, sdtKindOf, inlinePicture, rPrToElements, rPrFromElements, textOf, walk, walkAll, find, linkParents, applyRunOptions, readRunOptions, UNDERLINE_TO_WML, isElement, typeNameOf, W_NS } from '@docx4j/generated-objects-ts/builders/wml';
 export { StylesFacade } from './stylesFacade.mjs';

package/dist/model/content/search.d.mts

@@ -26,6 +26,14 @@ export declare function findAll(text: string, pattern: RegExp): [number, number]
  * Every match of a pattern in a text, with its capture groups: what `replaceText` needs to expand
  * `$1` and Word's `\1`. An empty match advances by one rather than looping forever, as `findAll`
  * has always done.
+ *
+ * A match that starts or ends inside a grapheme cluster is dropped: half of an emoji's surrogate
+ * pair (a wildcard `?` or class matches one UTF-16 unit), or an Arabic letter without the
+ * combining mark after it (`\p{L}+` stops before the mark). Such a range is not text a person can
+ * see, and replacing it splits a character - half a surrogate pair is saved as U+FFFD (CR-002
+ * section 32 item 1, found by the editor's agreement suite). The clusters are `Intl.Segmenter`'s,
+ * computed only for a text with a code unit at or above U+0300, below which every code unit is a
+ * cluster of its own, so Latin text pays nothing; a runtime without `Intl.Segmenter` keeps every match.
  */
 export declare function matchesOf(text: string, pattern: RegExp): RegExpExecArray[];
 /**

package/dist/model/content/search.mjs

@@ -94,6 +94,14 @@ export function findAll(text, pattern) {
  * Every match of a pattern in a text, with its capture groups: what `replaceText` needs to expand
  * `$1` and Word's `\1`. An empty match advances by one rather than looping forever, as `findAll`
  * has always done.
+ *
+ * A match that starts or ends inside a grapheme cluster is dropped: half of an emoji's surrogate
+ * pair (a wildcard `?` or class matches one UTF-16 unit), or an Arabic letter without the
+ * combining mark after it (`\p{L}+` stops before the mark). Such a range is not text a person can
+ * see, and replacing it splits a character - half a surrogate pair is saved as U+FFFD (CR-002
+ * section 32 item 1, found by the editor's agreement suite). The clusters are `Intl.Segmenter`'s,
+ * computed only for a text with a code unit at or above U+0300, below which every code unit is a
+ * cluster of its own, so Latin text pays nothing; a runtime without `Intl.Segmenter` keeps every match.
  */
 export function matchesOf(text, pattern) {
     const out = [];
@@ -105,7 +113,20 @@ export function matchesOf(text, pattern) {
         }
         out.push(m);
     }
-    return out;
+    const boundaries = out.length > 0 ? graphemeBoundaries(text) : undefined;
+    return boundaries === undefined ? out : out.filter((m) => boundaries.has(m.index) && boundaries.has(m.index + m[0].length));
+}
+/** The code-unit offsets where a grapheme cluster starts, and the text's end; undefined where every unit is its own cluster. */
+function graphemeBoundaries(text) {
+    if (!/[\u0300-\uffff]/.test(text))
+        return undefined;
+    const Segmenter = Intl.Segmenter;
+    if (Segmenter === undefined)
+        return undefined;
+    const boundaries = new Set([text.length]);
+    for (const segment of new Segmenter(undefined, { granularity: 'grapheme' }).segment(text))
+        boundaries.add(segment.index);
+    return boundaries;
 }
 /**
  * A replacement with its group references expanded against one match: `$1` to `$9`, `$&` and `$$`

package/dist/packages/WordprocessingMLPackage.d.mts

@@ -232,6 +232,18 @@ export declare class WordprocessingMLPackage extends OpcPackage implements Track
      * bytes cannot change; bytes that were set are read again when they are set again.
      */
     seedAnnotationIds(): Promise<number>;
+    /**
+     * Records the highest annotation id of a part the caller has read itself, so that
+     * `seedAnnotationIds()` does not read it again. An editor that imports the main document part into
+     * a model of its own knows that part's ids already, and seeding would parse it a second time -
+     * 1.06 s for a 255-page document (CR-002 section 34). The floor rises to `highest` at once, and
+     * the part counts as read for the bytes it holds now: set its bytes again and the next seed reads
+     * it afresh; unmarshal it and the counter's own scan of it covers it.
+     *
+     * The caller vouches for the number: one lower than the part's highest id lets a new revision
+     * reuse an id. Asynchronous only because a part whose bytes were set is keyed on those bytes.
+     */
+    noteAnnotationIds(part: XmlPart<unknown>, highest: number): Promise<void>;
     /** TrackingHost: the trees a new revision id must be above, which is every part already unmarshalled. */
     markupRoots(): object[];
     /**

package/dist/packages/WordprocessingMLPackage.mjs

@@ -448,6 +448,28 @@ export class WordprocessingMLPackage extends OpcPackage {
         }
         return this.annotationIdFloor;
     }
+    /**
+     * Records the highest annotation id of a part the caller has read itself, so that
+     * `seedAnnotationIds()` does not read it again. An editor that imports the main document part into
+     * a model of its own knows that part's ids already, and seeding would parse it a second time -
+     * 1.06 s for a 255-page document (CR-002 section 34). The floor rises to `highest` at once, and
+     * the part counts as read for the bytes it holds now: set its bytes again and the next seed reads
+     * it afresh; unmarshal it and the counter's own scan of it covers it.
+     *
+     * The caller vouches for the number: one lower than the part's highest id lets a new revision
+     * reuse an id. Asynchronous only because a part whose bytes were set is keyed on those bytes.
+     */
+    async noteAnnotationIds(part, highest) {
+        if (!Number.isInteger(highest) || highest < 0) {
+            throw new Docx4JException(`The highest annotation id must be a non-negative integer, not ${String(highest)}`);
+        }
+        if (this.parts.get(part.partName.name) !== part) {
+            throw new Docx4JException(`${part.partName.name} is not a part of this package`);
+        }
+        this.annotationIdFloor = Math.max(this.annotationIdFloor, highest);
+        if (!part.isUnmarshalled)
+            this.seededParts.set(part, part.isUntouched ? true : await part.getBytes());
+    }
     /** TrackingHost: the trees a new revision id must be above, which is every part already unmarshalled. */
     markupRoots() {
         const out = [];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@docx4j/core-ts",
-  "version": "0.2.1",
+  "version": "0.2.2",
   "description": "docx4j-core for TypeScript: Open Packaging (docx/pptx/xlsx zip and flat OPC), typed parts and relationships, style, numbering and font resolution over @docx4j/generated-objects-ts. Node, browsers, Word add-ins.",
   "type": "module",
   "main": "./dist/index.mjs",