@beyondwork/docx-react-component 1.0.19 → 1.0.21

package/src/io/ooxml/parse-main-document.ts

@@ -26,6 +26,8 @@ import {
 import { toCanonicalNumberingInstanceId } from "./parse-numbering.ts";
 import { parseComplexContentXml } from "./parse-complex-content.ts";
 import { parseShapeXml, parseVmlXml } from "./parse-shapes.ts";
+import { classifyFieldInstruction } from "./parse-fields.ts";
+import { resolveHighlightColor } from "./highlight-colors.ts";
 
 export interface ParsedMainDocument {
   blocks: ParsedBlockNode[];
@@ -56,6 +58,7 @@ export interface ParsedParagraphNode {
   };
   alignment?: "left" | "center" | "right" | "both" | "distribute";
   spacing?: ParagraphSpacing;
+  contextualSpacing?: boolean;
   indentation?: ParagraphIndentation;
   tabStops?: TabStop[];
   keepNext?: boolean;
@@ -90,7 +93,10 @@ export type ParsedInlineNode =
   | ParsedVmlShapeInlineNode
   | ParsedBookmarkStartInlineNode
   | ParsedBookmarkEndInlineNode
-  | ParsedFieldInlineNode;
+  | ParsedFootnoteRefInlineNode
+  | ParsedFieldInlineNode
+  | ParsedPermStartInlineNode
+  | ParsedPermEndInlineNode;
 
 export interface ParsedTextNode {
   type: "text";
@@ -125,6 +131,8 @@ export interface ParsedImageNode {
   contentType?: string;
   filename?: string;
   altText?: string;
+  widthEmu?: number;
+  heightEmu?: number;
   placementXml?: string;
   display?: "inline" | "floating";
   floating?: {
@@ -204,11 +212,32 @@ export interface ParsedBookmarkEndInlineNode {
   rawXml: string;
 }
 
+export interface ParsedFootnoteRefInlineNode {
+  type: "footnote_ref";
+  noteId: string;
+  noteKind: "footnote" | "endnote";
+}
+
 export interface ParsedFieldInlineNode {
   type: "field";
-  fieldType: "simple";
+  fieldType: "simple" | "complex";
   instruction: string;
-  contentXml: string;
+  contentXml?: string;
+  children?: Array<ParsedTextNode | ParsedBreakNode | ParsedTabNode>;
+  rawXml: string;
+}
+
+export interface ParsedPermStartInlineNode {
+  type: "perm_start";
+  rangeId: string;
+  editorGroup?: string;
+  editor?: string;
+  rawXml: string;
+}
+
+export interface ParsedPermEndInlineNode {
+  type: "perm_end";
+  rangeId: string;
   rawXml: string;
 }
 
@@ -412,6 +441,7 @@ function parseBodyChild(
   let numbering: ParsedParagraphNode["numbering"];
   let alignment: ParsedParagraphNode["alignment"];
   let spacing: ParsedParagraphNode["spacing"];
+  let contextualSpacing: ParsedParagraphNode["contextualSpacing"];
   let indentation: ParsedParagraphNode["indentation"];
   let tabStops: ParsedParagraphNode["tabStops"];
   let keepNext: ParsedParagraphNode["keepNext"];
@@ -428,6 +458,11 @@ function parseBodyChild(
   let sectionPropertiesXml: string | undefined;
   let paragraphSupported = true;
   const children: ParsedInlineNode[] = [];
+  let activeComplexField: {
+    instruction: string;
+    children: Array<ParsedTextNode | ParsedBreakNode | ParsedTabNode>;
+    mode: "instruction" | "result";
+  } | null = null;
 
   for (const child of node.children) {
     if (child.type !== "element") {
@@ -440,6 +475,7 @@ function parseBodyChild(
         numbering = readParagraphNumbering(child);
         alignment = readParagraphAlignment(child);
         spacing = readParagraphSpacing(child);
+        contextualSpacing = readOptionalOnOffParagraphProperty(child, "contextualSpacing");
         indentation = readParagraphIndentation(child);
         tabStops = readParagraphTabStops(child);
         keepNext = readOnOffParagraphProperty(child, "keepNext");
@@ -457,33 +493,148 @@ function parseBodyChild(
         paragraphSupported = paragraphSupported && supportsParagraphProperties(child);
         break;
       case "r":
-        children.push(...parseRun(child, sourceXml, relationships, mediaParts, sourcePartPath));
+        activeComplexField = appendParagraphRunNodes(
+          child,
+          children,
+          activeComplexField,
+          sourceXml,
+          relationships,
+          mediaParts,
+          sourcePartPath,
+        );
         break;
       case "hyperlink": {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
         const hyperlink = parseHyperlink(child, sourceXml, relationshipMap);
         children.push(hyperlink);
         break;
       }
       case "ins":
       case "del": {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
         children.push(...parseRevisionContainer(child, sourceXml, relationshipMap));
         break;
       }
       case "commentRangeStart":
       case "commentRangeEnd":
         break;
-      case "bookmarkStart":
-      case "bookmarkEnd":
-      case "fldSimple":
-      case "permStart":
-      case "permEnd":
+      case "bookmarkStart": {
+        const bkId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        const bkName = child.attributes["w:name"] ?? child.attributes.name ?? "";
+        if (bkId) {
+          const bookmarkNode = {
+            type: "bookmark_start",
+            bookmarkId: bkId,
+            name: bkName,
+            rawXml: sourceXml.slice(child.start, child.end),
+          } as ParsedBookmarkStartInlineNode;
+          flushActiveComplexField(children, () => {
+            activeComplexField = null;
+          }, activeComplexField);
+          children.push(bookmarkNode);
+        } else {
+          flushActiveComplexField(children, () => {
+            activeComplexField = null;
+          }, activeComplexField);
+          children.push({ type: "opaque_inline", rawXml: sourceXml.slice(child.start, child.end) });
+        }
+        break;
+      }
+      case "bookmarkEnd": {
+        const bkEndId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        if (bkEndId) {
+          const bookmarkNode = {
+            type: "bookmark_end",
+            bookmarkId: bkEndId,
+            rawXml: sourceXml.slice(child.start, child.end),
+          } as ParsedBookmarkEndInlineNode;
+          flushActiveComplexField(children, () => {
+            activeComplexField = null;
+          }, activeComplexField);
+          children.push(bookmarkNode);
+        } else {
+          flushActiveComplexField(children, () => {
+            activeComplexField = null;
+          }, activeComplexField);
+          children.push({ type: "opaque_inline", rawXml: sourceXml.slice(child.start, child.end) });
+        }
+        break;
+      }
+      case "fldSimple": {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+        const fieldInstr = child.attributes["w:instr"] ?? child.attributes.instr ?? "";
+        const fieldContentXml = child.children
+          .filter((c): c is XmlElementNode => c.type === "element")
+          .map((c) => sourceXml.slice(c.start, c.end))
+          .join("");
+        children.push({
+          type: "field",
+          fieldType: "simple",
+          instruction: fieldInstr,
+          contentXml: fieldContentXml,
+          rawXml: sourceXml.slice(child.start, child.end),
+        } as ParsedFieldInlineNode);
+        break;
+      }
       case "proofErr":
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
         children.push({
           type: "opaque_inline",
           rawXml: sourceXml.slice(child.start, child.end),
         });
         break;
+      case "permStart":
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+        children.push(parsePermStartNode(child, sourceXml));
+        break;
+      case "permEnd":
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+        children.push(parsePermEndNode(child, sourceXml));
+        break;
+      case "footnoteReference": {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+        const noteId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        if (noteId) {
+          children.push({
+            type: "footnote_ref",
+            noteId,
+            noteKind: "footnote",
+          });
+        }
+        break;
+      }
+      case "endnoteReference": {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+        const noteId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        if (noteId) {
+          children.push({
+            type: "footnote_ref",
+            noteId,
+            noteKind: "endnote",
+          });
+        }
+        break;
+      }
       default:
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
         children.push({
           type: "opaque_inline",
           rawXml: sourceXml.slice(child.start, child.end),
@@ -492,6 +643,10 @@ function parseBodyChild(
     }
   }
 
+  flushActiveComplexField(children, () => {
+    activeComplexField = null;
+  }, activeComplexField);
+
   if (!paragraphSupported) {
     return {
       type: "opaque_block",
@@ -505,6 +660,7 @@ function parseBodyChild(
     ...(numbering ? { numbering } : {}),
     ...(alignment ? { alignment } : {}),
     ...(spacing ? { spacing } : {}),
+    ...(contextualSpacing !== undefined ? { contextualSpacing } : {}),
     ...(indentation ? { indentation } : {}),
     ...(tabStops && tabStops.length > 0 ? { tabStops } : {}),
     ...(keepNext ? { keepNext } : {}),
@@ -524,6 +680,114 @@ function parseBodyChild(
   };
 }
 
+function appendParagraphRunNodes(
+  node: XmlElementNode,
+  children: ParsedInlineNode[],
+  activeComplexField: {
+    instruction: string;
+    children: Array<ParsedTextNode | ParsedBreakNode | ParsedTabNode>;
+    mode: "instruction" | "result";
+  } | null,
+  sourceXml: string,
+  relationships: readonly OpcRelationship[],
+  mediaParts: ReadonlyMap<string, InlineMediaPart>,
+  sourcePartPath: string,
+): {
+  instruction: string;
+  children: Array<ParsedTextNode | ParsedBreakNode | ParsedTabNode>;
+  mode: "instruction" | "result";
+} | null {
+  const hasFieldMarkers = node.children.some(
+    (child) =>
+      child.type === "element" &&
+      (localName(child.name) === "fldChar" || localName(child.name) === "instrText"),
+  );
+
+  if (activeComplexField?.mode === "result" && !hasFieldMarkers) {
+    const run = parseRunContentOnly(node, sourceXml);
+    if (
+      run.supported &&
+      run.nodes.every(
+        (child) =>
+          child.type === "text" || child.type === "hard_break" || child.type === "tab",
+      )
+    ) {
+      activeComplexField.children.push(...run.nodes);
+      return activeComplexField;
+    }
+  }
+
+  if (!hasFieldMarkers) {
+    children.push(...parseRun(node, sourceXml, relationships, mediaParts, sourcePartPath));
+    return activeComplexField;
+  }
+
+  for (const child of node.children) {
+    if (child.type !== "element") {
+      continue;
+    }
+
+    const name = localName(child.name);
+    if (name === "fldChar") {
+      const fldType = child.attributes["w:fldCharType"] ?? child.attributes.fldCharType;
+      if (fldType === "begin") {
+        activeComplexField = { instruction: "", children: [], mode: "instruction" };
+      } else if (fldType === "separate" && activeComplexField) {
+        activeComplexField.mode = "result";
+      } else if (fldType === "end" && activeComplexField) {
+        flushActiveComplexField(children, () => {
+          activeComplexField = null;
+        }, activeComplexField);
+      }
+      continue;
+    }
+
+    if (name === "instrText") {
+      const instruction = child.children
+        .filter((entry): entry is XmlTextNode => entry.type === "text")
+        .map((entry) => entry.text)
+        .join("");
+      if (activeComplexField) {
+        activeComplexField.instruction += instruction;
+      } else if (instruction.trim().length > 0) {
+        children.push({
+          type: "field",
+          fieldType: "complex",
+          instruction,
+          children: [],
+          rawXml: sourceXml.slice(node.start, node.end),
+        });
+      }
+      continue;
+    }
+  }
+
+  return activeComplexField;
+}
+
+function flushActiveComplexField(
+  children: ParsedInlineNode[],
+  reset: () => void,
+  activeComplexField: {
+    instruction: string;
+    children: Array<ParsedTextNode | ParsedBreakNode | ParsedTabNode>;
+    mode: "instruction" | "result";
+  } | null,
+): void {
+  if (!activeComplexField || activeComplexField.instruction.trim().length === 0) {
+    return;
+  }
+
+  children.push({
+    type: "field",
+    fieldType: "complex",
+    instruction: activeComplexField.instruction,
+    children: activeComplexField.children,
+    rawXml: "",
+  });
+  reset();
+}
+
 function parseTableElement(
   node: XmlElementNode,
   sourceXml: string,
@@ -918,9 +1182,35 @@ function readTableGridColumns(node: XmlElementNode): number[] {
  */
 function tableRequiresOpaquePreservation(rawXml: string): boolean {
   // Safe table-local content now includes hyperlinks, bookmarks, comments,
-  // nested tables, floating images, and VML preview atoms because the parser
-  // and serializer can preserve them without degrading the whole table.
-  return /<w:(ins|del|rPrChange|pPrChange|tblPrChange|trPrChange|tcPrChange|sectPrChange|cellIns|cellDel|cellMerge|fldChar|fldSimple|smartTag|gridAfter|gridBefore|tblCellSpacing)\b/.test(rawXml);
+  // nested tables, floating images, VML preview atoms, and bounded field
+  // families already owned by the current field slice. Risky table-local
+  // semantics still fail closed to preserve-only.
+  if (/<w:(ins|del|rPrChange|pPrChange|tblPrChange|trPrChange|tcPrChange|sectPrChange|cellIns|cellDel|cellMerge|smartTag|gridAfter|gridBefore|tblCellSpacing)\b/.test(rawXml)) {
+    return true;
+  }
+
+  const fldSimpleInstructions = [...rawXml.matchAll(/\bw:instr="([^"]*)"/g)].map((match) => match[1] ?? "");
+  const complexInstructions = [...rawXml.matchAll(/<(?:\w+:)?instrText\b[^>]*>([\s\S]*?)<\/(?:\w+:)?instrText>/gu)]
+    .map((match) => decodeXmlEntities(match[1] ?? ""));
+  for (const instruction of [...fldSimpleInstructions, ...complexInstructions]) {
+    const classification = classifyFieldInstruction(instruction);
+    if (!isSafeMainStoryTableFieldFamily(classification.family)) {
+      return true;
+    }
+  }
+
+  return false;
+}
+
+function isSafeMainStoryTableFieldFamily(family: string): boolean {
+  return (
+    family === "REF" ||
+    family === "PAGEREF" ||
+    family === "NOTEREF" ||
+    family === "TOC" ||
+    family === "PAGE" ||
+    family === "NUMPAGES"
+  );
 }
 
 function readCellGridSpan(node: XmlElementNode): number | undefined {
@@ -1091,6 +1381,18 @@ function readOnOffParagraphProperty(node: XmlElementNode, name: string): boolean
   return val !== "false" && val !== "0" && val !== "off" ? true : undefined;
 }
 
+function readOptionalOnOffParagraphProperty(
+  node: XmlElementNode,
+  name: string,
+): boolean | undefined {
+  const propNode = node.children.find(
+    (child): child is XmlElementNode => child.type === "element" && localName(child.name) === name,
+  );
+  if (!propNode) return undefined;
+  const val = (propNode.attributes["w:val"] ?? propNode.attributes.val ?? "true").toLowerCase();
+  return val !== "false" && val !== "0" && val !== "off";
+}
+
 function readParagraphOutlineLevel(node: XmlElementNode): number | undefined {
   const propNode = node.children.find(
     (child): child is XmlElementNode => child.type === "element" && localName(child.name) === "outlineLvl",
@@ -1355,6 +1657,8 @@ function parseRun(
             contentType: media.contentType,
             filename: media.filename,
             ...(media.altText ? { altText: media.altText } : {}),
+            ...(media.widthEmu !== undefined ? { widthEmu: media.widthEmu } : {}),
+            ...(media.heightEmu !== undefined ? { heightEmu: media.heightEmu } : {}),
             placementXml,
             ...(media.display ? { display: media.display } : {}),
             ...(media.floating ? { floating: media.floating } : {}),
@@ -1387,6 +1691,28 @@ function parseRun(
       }
       case "commentReference":
         break;
+      case "footnoteReference": {
+        const noteId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        if (noteId) {
+          result.push({
+            type: "footnote_ref",
+            noteId,
+            noteKind: "footnote",
+          });
+        }
+        break;
+      }
+      case "endnoteReference": {
+        const noteId = child.attributes["w:id"] ?? child.attributes.id ?? "";
+        if (noteId) {
+          result.push({
+            type: "footnote_ref",
+            noteId,
+            noteKind: "endnote",
+          });
+        }
+        break;
+      }
       case "lastRenderedPageBreak":
       case "proofErr":
         result.push({
@@ -1466,8 +1792,6 @@ function parseRevisionContainer(
       case "commentRangeEnd":
       case "bookmarkStart":
       case "bookmarkEnd":
-      case "permStart":
-      case "permEnd":
       case "proofErr":
       case "lastRenderedPageBreak":
         return [
@@ -1476,6 +1800,10 @@ function parseRevisionContainer(
             rawXml: sourceXml.slice(node.start, node.end),
           },
         ];
+      case "permStart":
+        return [parsePermStartNode(node, sourceXml)];
+      case "permEnd":
+        return [parsePermEndNode(node, sourceXml)];
       default:
         return [
           {
@@ -1697,6 +2025,11 @@ function readRunMarks(node: XmlElementNode, sourceXml: string): MarksParseResult
     marks.push(backgroundColorMark);
   }
 
+  const highlightMark = readRunHighlight(properties);
+  if (highlightMark) {
+    marks.push(highlightMark);
+  }
+
   const charSpacingMark = readNumericRunMark(properties, "spacing", "charSpacing");
   if (charSpacingMark) {
     marks.push(charSpacingMark);
@@ -1809,6 +2142,28 @@ function readRunBackgroundColor(properties: XmlElementNode): TextMark | undefine
   return { type: "backgroundColor", color: fill };
 }
 
+function readRunHighlight(properties: XmlElementNode): TextMark | undefined {
+  const highlightNode = properties.children.find(
+    (child): child is XmlElementNode =>
+      child.type === "element" && localName(child.name) === "highlight",
+  );
+  if (!highlightNode) {
+    return undefined;
+  }
+
+  const highlightValue = highlightNode.attributes["w:val"] ?? highlightNode.attributes.val;
+  const resolvedHighlight = resolveHighlightColor(highlightValue);
+  if (!resolvedHighlight) {
+    return undefined;
+  }
+
+  return {
+    type: "highlight",
+    color: resolvedHighlight.color,
+    val: resolvedHighlight.val,
+  };
+}
+
 function readNumericRunMark(
   properties: XmlElementNode,
   elementName: "spacing" | "kern" | "position",
@@ -2412,3 +2767,30 @@ function parseBorderSpec(node: XmlElementNode): BorderSpec | undefined {
 
   return Object.keys(border).length > 0 ? border : undefined;
 }
+
+function parsePermStartNode(
+  node: XmlElementNode,
+  sourceXml: string,
+): ParsedPermStartInlineNode {
+  const rangeId = node.attributes["w:id"] ?? node.attributes.id ?? "";
+  const edGrp = node.attributes["w:edGrp"] ?? node.attributes.edGrp;
+  const ed = node.attributes["w:ed"] ?? node.attributes.ed;
+  return {
+    type: "perm_start",
+    rangeId,
+    ...(edGrp ? { editorGroup: edGrp } : {}),
+    ...(ed ? { editor: ed } : {}),
+    rawXml: sourceXml.slice(node.start, node.end),
+  };
+}
+
+function parsePermEndNode(
+  node: XmlElementNode,
+  sourceXml: string,
+): ParsedPermEndInlineNode {
+  return {
+    type: "perm_end",
+    rangeId: node.attributes["w:id"] ?? node.attributes.id ?? "",
+    rawXml: sourceXml.slice(node.start, node.end),
+  };
+}