@cj-tech-master/excelts 9.5.4 → 9.5.5

package/dist/cjs/modules/word/reader/docx-reader.js

@@ -0,0 +1,2560 @@
+"use strict";
+/**
+ * DOCX Module - Reader / Parser
+ *
+ * Reads a DOCX ZIP file and parses it into a DocxDocument model.
+ * Uses the archive module for ZIP reading and XML module for parsing.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.readDocx = readDocx;
+const read_archive_1 = require("../../archive/read-archive.js");
+const dom_1 = require("../../xml/dom.js");
+const constants_1 = require("../constants");
+const internal_utils_1 = require("../core/internal-utils");
+const text_utils_1 = require("../core/text-utils");
+const errors_1 = require("../errors");
+const encryption_1 = require("../security/encryption");
+const policy_1 = require("../security/policy");
+const chart_parser_1 = require("./chart-parser");
+const comments_parser_1 = require("./comments-parser");
+const doc_props_parsers_1 = require("./doc-props-parsers");
+const form_field_parser_1 = require("./form-field-parser");
+const image_parsers_1 = require("./image-parsers");
+const math_parser_1 = require("./math-parser");
+const metadata_parsers_1 = require("./metadata-parsers");
+const numbering_parser_1 = require("./numbering-parser");
+const paragraph_section_parsers_1 = require("./paragraph-section-parsers");
+const parse_utils_1 = require("./parse-utils");
+const properties_parsers_1 = require("./properties-parsers");
+const reader_context_1 = require("./reader-context");
+const sdt_helpers_1 = require("./sdt-helpers");
+const styles_parser_1 = require("./styles-parser");
+const table_properties_parsers_1 = require("./table-properties-parsers");
+const watermark_parser_1 = require("./watermark-parser");
+// =============================================================================
+// Run Content Parser
+// =============================================================================
+function parseRunContent(el) {
+    const content = [];
+    for (const child of el.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        const name = child.name.replace(/^w:/, "");
+        switch (name) {
+            case "t":
+                content.push({ type: "text", text: (0, dom_1.textContent)(child) });
+                break;
+            case "br": {
+                const brType = (0, parse_utils_1.attrVal)(child, "type");
+                content.push({
+                    type: "break",
+                    breakType: brType
+                });
+                break;
+            }
+            case "tab":
+                content.push({ type: "tab" });
+                break;
+            case "ptab": {
+                const alignment = (0, parse_utils_1.attrVal)(child, "alignment") ?? "left";
+                const relativeTo = (0, parse_utils_1.attrVal)(child, "relativeTo") ?? "margin";
+                const leader = (0, parse_utils_1.attrVal)(child, "leader");
+                const ptab = {
+                    type: "ptab",
+                    alignment: alignment,
+                    relativeTo: relativeTo
+                };
+                if (leader) {
+                    ptab.leader = leader;
+                }
+                content.push(ptab);
+                break;
+            }
+            case "ruby": {
+                const ruby = {
+                    type: "ruby",
+                    rubyText: [],
+                    baseText: []
+                };
+                const rubyPrEl = (0, parse_utils_1.findChildNs)(child, "rubyPr");
+                if (rubyPrEl) {
+                    const props = {};
+                    const alignEl = (0, parse_utils_1.findChildNs)(rubyPrEl, "rubyAlign");
+                    if (alignEl) {
+                        props.align = (0, parse_utils_1.attrVal)(alignEl, "val");
+                    }
+                    const hpsEl = (0, parse_utils_1.findChildNs)(rubyPrEl, "hps");
+                    if (hpsEl) {
+                        props.fontSize = (0, parse_utils_1.attrInt)(hpsEl, "val");
+                    }
+                    const hpsRaiseEl = (0, parse_utils_1.findChildNs)(rubyPrEl, "hpsRaise");
+                    if (hpsRaiseEl) {
+                        props.raise = (0, parse_utils_1.attrInt)(hpsRaiseEl, "val");
+                    }
+                    const hpsBaseTextEl = (0, parse_utils_1.findChildNs)(rubyPrEl, "hpsBaseText");
+                    if (hpsBaseTextEl) {
+                        props.baseFontSize = (0, parse_utils_1.attrInt)(hpsBaseTextEl, "val");
+                    }
+                    const lidEl = (0, parse_utils_1.findChildNs)(rubyPrEl, "lid");
+                    if (lidEl) {
+                        props.language = (0, parse_utils_1.attrVal)(lidEl, "val");
+                    }
+                    if (Object.keys(props).length > 0) {
+                        ruby.properties = props;
+                    }
+                }
+                // Parse w:rt (ruby text)
+                const rtEl = (0, parse_utils_1.findChildNs)(child, "rt");
+                if (rtEl) {
+                    for (const rtChild of rtEl.children) {
+                        if (rtChild.type === "element" && rtChild.name.replace(/^w:/, "") === "r") {
+                            ruby.rubyText.push(parseRun(rtChild));
+                        }
+                    }
+                }
+                // Parse w:rubyBase
+                const baseEl = (0, parse_utils_1.findChildNs)(child, "rubyBase");
+                if (baseEl) {
+                    for (const bChild of baseEl.children) {
+                        if (bChild.type === "element" && bChild.name.replace(/^w:/, "") === "r") {
+                            ruby.baseText.push(parseRun(bChild));
+                        }
+                    }
+                }
+                content.push(ruby);
+                break;
+            }
+            case "sym":
+                content.push({
+                    type: "symbol",
+                    font: (0, parse_utils_1.attrVal)(child, "font") ?? "",
+                    char: (0, parse_utils_1.attrVal)(child, "char") ?? ""
+                });
+                break;
+            case "footnoteReference": {
+                const fr = {
+                    type: "footnoteRef",
+                    id: (0, parse_utils_1.attrInt)(child, "id") ?? 0
+                };
+                const cmf = (0, parse_utils_1.attrVal)(child, "customMarkFollows");
+                if (cmf === "1" || cmf === "true") {
+                    fr.customMarkFollows = true;
+                }
+                content.push(fr);
+                break;
+            }
+            case "endnoteReference": {
+                const er = {
+                    type: "endnoteRef",
+                    id: (0, parse_utils_1.attrInt)(child, "id") ?? 0
+                };
+                const cmf = (0, parse_utils_1.attrVal)(child, "customMarkFollows");
+                if (cmf === "1" || cmf === "true") {
+                    er.customMarkFollows = true;
+                }
+                content.push(er);
+                break;
+            }
+            case "drawing":
+                (0, image_parsers_1.parseDrawingContent)(child, content);
+                break;
+            case "cr":
+                content.push({ type: "carriageReturn" });
+                break;
+            case "noBreakHyphen":
+                content.push({ type: "noBreakHyphen" });
+                break;
+            case "softHyphen":
+                content.push({ type: "softHyphen" });
+                break;
+            case "lastRenderedPageBreak":
+                content.push({ type: "lastRenderedPageBreak" });
+                break;
+            case "annotationRef":
+                content.push({ type: "annotationReference", id: (0, parse_utils_1.attrInt)(child, "id") ?? 0 });
+                break;
+            case "commentReference":
+                // This is annotationReference for comments inside runs
+                content.push({ type: "annotationReference", id: (0, parse_utils_1.attrInt)(child, "id") ?? 0 });
+                break;
+            case "rPr":
+            case "fldChar":
+            case "instrText":
+            case "delText":
+                // Known structural elements handled elsewhere — skip silently
+                break;
+            default:
+                // Unknown run child: preserve as opaque for round-trip fidelity
+                content.push({
+                    type: "opaqueRun",
+                    rawXml: (0, parse_utils_1.serializeElement)(child)
+                });
+                break;
+        }
+    }
+    return content;
+}
+// =============================================================================
+// DrawingML Shape Parser
+// =============================================================================
+function parseDrawingShape(anchorEl, wspEl, ctx) {
+    const docPrEl = (0, dom_1.findChild)(anchorEl, "wp:docPr");
+    const extentEl = (0, dom_1.findChild)(anchorEl, "wp:extent");
+    const cx = parseInt(extentEl?.attributes["cx"] ?? "0", 10);
+    const cy = parseInt(extentEl?.attributes["cy"] ?? "0", 10);
+    // Parse preset shape type from wps:spPr > a:prstGeom
+    const spPrEl = (0, dom_1.findChild)(wspEl, "wps:spPr") ?? (0, parse_utils_1.findChildNs)(wspEl, "spPr");
+    const prstGeomEl = spPrEl
+        ? ((0, dom_1.findChild)(spPrEl, "a:prstGeom") ?? (0, parse_utils_1.findChildNs)(spPrEl, "prstGeom"))
+        : undefined;
+    const shapeType = prstGeomEl?.attributes["prst"] ?? "rect";
+    const shape = {
+        type: "drawingShape",
+        shapeType: shapeType,
+        width: cx,
+        height: cy,
+        altText: docPrEl?.attributes["descr"],
+        name: docPrEl?.attributes["name"]
+    };
+    // Parse fill
+    if (spPrEl) {
+        const solidFill = (0, dom_1.findChild)(spPrEl, "a:solidFill") ?? (0, parse_utils_1.findChildNs)(spPrEl, "solidFill");
+        if (solidFill) {
+            const srgb = (0, dom_1.findChild)(solidFill, "a:srgbClr") ?? (0, parse_utils_1.findChildNs)(solidFill, "srgbClr");
+            if (srgb) {
+                shape.fillColor = srgb.attributes["val"];
+            }
+        }
+        const noFill = (0, dom_1.findChild)(spPrEl, "a:noFill") ?? (0, parse_utils_1.findChildNs)(spPrEl, "noFill");
+        if (noFill) {
+            shape.noFill = true;
+        }
+        // Parse outline
+        const lnEl = (0, dom_1.findChild)(spPrEl, "a:ln") ?? (0, parse_utils_1.findChildNs)(spPrEl, "ln");
+        if (lnEl) {
+            const w = lnEl.attributes["w"];
+            if (w) {
+                shape.outlineWidth = parseInt(w, 10);
+            }
+            const lnFill = (0, dom_1.findChild)(lnEl, "a:solidFill") ?? (0, parse_utils_1.findChildNs)(lnEl, "solidFill");
+            if (lnFill) {
+                const srgb = (0, dom_1.findChild)(lnFill, "a:srgbClr") ?? (0, parse_utils_1.findChildNs)(lnFill, "srgbClr");
+                if (srgb) {
+                    shape.outlineColor = srgb.attributes["val"];
+                }
+            }
+            const noLn = (0, dom_1.findChild)(lnEl, "a:noFill") ?? (0, parse_utils_1.findChildNs)(lnEl, "noFill");
+            if (noLn) {
+                shape.noOutline = true;
+            }
+        }
+    }
+    // Parse text content (wps:txbx > w:txbxContent)
+    const txbxEl = (0, dom_1.findChild)(wspEl, "wps:txbx") ?? (0, parse_utils_1.findChildNs)(wspEl, "txbx");
+    const txbxContentEl = txbxEl
+        ? ((0, dom_1.findChild)(txbxEl, "w:txbxContent") ?? (0, parse_utils_1.findChildNs)(txbxEl, "txbxContent"))
+        : undefined;
+    if (txbxContentEl) {
+        const paras = [];
+        for (const child of txbxContentEl.children) {
+            if (child.type === "element" && child.name.replace(/^w:/, "") === "p") {
+                paras.push(parseParagraph(child, ctx));
+            }
+        }
+        if (paras.length > 0) {
+            shape.textContent = paras;
+        }
+    }
+    // Parse positioning
+    const posH = (0, dom_1.findChild)(anchorEl, "wp:positionH");
+    if (posH) {
+        const hp = {
+            relativeTo: posH.attributes["relativeFrom"]
+        };
+        const offsetEl = (0, dom_1.findChild)(posH, "wp:posOffset");
+        if (offsetEl) {
+            hp.offset = parseInt((0, dom_1.textContent)(offsetEl), 10);
+        }
+        const alignEl = (0, dom_1.findChild)(posH, "wp:align");
+        if (alignEl) {
+            hp.align = (0, dom_1.textContent)(alignEl);
+        }
+        shape.horizontalPosition = hp;
+    }
+    const posV = (0, dom_1.findChild)(anchorEl, "wp:positionV");
+    if (posV) {
+        const vp = {
+            relativeTo: posV.attributes["relativeFrom"]
+        };
+        const offsetEl = (0, dom_1.findChild)(posV, "wp:posOffset");
+        if (offsetEl) {
+            vp.offset = parseInt((0, dom_1.textContent)(offsetEl), 10);
+        }
+        const alignEl = (0, dom_1.findChild)(posV, "wp:align");
+        if (alignEl) {
+            vp.align = (0, dom_1.textContent)(alignEl);
+        }
+        shape.verticalPosition = vp;
+    }
+    // Wrap
+    for (const wrapChild of anchorEl.children) {
+        if (wrapChild.type !== "element") {
+            continue;
+        }
+        const wn = wrapChild.name;
+        if (wn === "wp:wrapSquare") {
+            shape.wrap = {
+                style: "square",
+                side: wrapChild.attributes["wrapText"]
+            };
+        }
+        else if (wn === "wp:wrapTight") {
+            shape.wrap = {
+                style: "tight",
+                side: wrapChild.attributes["wrapText"]
+            };
+        }
+        else if (wn === "wp:wrapTopAndBottom") {
+            shape.wrap = { style: "topAndBottom" };
+        }
+        else if (wn === "wp:wrapNone") {
+            shape.wrap = { style: "none" };
+        }
+    }
+    // Behind doc
+    if (anchorEl.attributes["behindDoc"] === "1") {
+        shape.behindDoc = true;
+    }
+    // Rotation
+    if (spPrEl) {
+        const xfrmEl = (0, dom_1.findChild)(spPrEl, "a:xfrm") ?? (0, parse_utils_1.findChildNs)(spPrEl, "xfrm");
+        if (xfrmEl?.attributes["rot"]) {
+            shape.rotation = parseInt(xfrmEl.attributes["rot"], 10);
+        }
+    }
+    return shape;
+}
+// =============================================================================
+// TextBox Parser
+// =============================================================================
+function parseTextBox(pictEl, ctx) {
+    // Look for v:shape > v:textbox > w:txbxContent
+    let txbxContentEl;
+    let shapeEl;
+    for (const child of pictEl.children) {
+        if (child.type === "element" && (child.name === "v:shape" || child.name === "v:rect")) {
+            shapeEl = child;
+            for (const sc of child.children) {
+                if (sc.type === "element" && sc.name === "v:textbox") {
+                    for (const tc of sc.children) {
+                        if (tc.type === "element" &&
+                            (tc.name === "w:txbxContent" || tc.name === "txbxContent")) {
+                            txbxContentEl = tc;
+                        }
+                    }
+                }
+            }
+        }
+    }
+    if (!txbxContentEl) {
+        return undefined;
+    }
+    const paragraphs = [];
+    for (const c of txbxContentEl.children) {
+        if (c.type === "element" && c.name.replace(/^w:/, "") === "p") {
+            paragraphs.push(parseParagraph(c, ctx));
+        }
+    }
+    const tb = { type: "textBox", content: paragraphs };
+    if (shapeEl) {
+        const style = shapeEl.attributes["style"];
+        if (style) {
+            tb.style = style;
+        }
+        const sc = shapeEl.attributes["strokecolor"];
+        if (sc) {
+            tb.strokeColor = sc;
+        }
+        const fc = shapeEl.attributes["fillcolor"];
+        if (fc) {
+            tb.fillColor = fc;
+        }
+        if (shapeEl.attributes["stroked"] === "f") {
+            tb.stroke = false;
+        }
+        if (shapeEl.attributes["filled"] === "f") {
+            tb.fill = false;
+        }
+    }
+    return tb;
+}
+// =============================================================================
+// SDT / CheckBox / TOC Parser
+// =============================================================================
+function parseSdt(sdtEl, ctx) {
+    const sdtPrEl = (0, parse_utils_1.findChildNs)(sdtEl, "sdtPr");
+    const sdtContentEl = (0, parse_utils_1.findChildNs)(sdtEl, "sdtContent");
+    // Check for checkbox (w14:checkbox)
+    if (sdtPrEl) {
+        const checkBoxEl = (0, dom_1.findChild)(sdtPrEl, "w14:checkbox");
+        if (checkBoxEl) {
+            return (0, sdt_helpers_1.parseCheckBox)(checkBoxEl);
+        }
+    }
+    // Check for TOC (contains docPartObj with docPartGallery "Table of Contents")
+    if (sdtPrEl) {
+        const docPartObjEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "docPartObj");
+        if (docPartObjEl) {
+            const galleryEl = (0, parse_utils_1.findChildNs)(docPartObjEl, "docPartGallery");
+            const galleryVal = galleryEl ? (0, parse_utils_1.attrVal)(galleryEl, "val") : undefined;
+            if (galleryVal === "Table of Contents") {
+                return parseTocFromSdt(sdtContentEl, ctx);
+            }
+        }
+    }
+    // Generic SDT
+    const props = {};
+    if (sdtPrEl) {
+        const tagEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "tag");
+        if (tagEl) {
+            props.tag = (0, parse_utils_1.attrVal)(tagEl, "val");
+        }
+        const aliasEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "alias");
+        if (aliasEl) {
+            props.alias = (0, parse_utils_1.attrVal)(aliasEl, "val");
+        }
+        const lockEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "lock");
+        if (lockEl) {
+            const v = (0, parse_utils_1.attrVal)(lockEl, "val");
+            if (v === "contentLocked" || v === "sdtContentLocked") {
+                props.lockContent = true;
+            }
+            if (v === "sdtLocked" || v === "sdtContentLocked") {
+                props.lockSdt = true;
+            }
+        }
+        // Plain text
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "text")) {
+            props.plainText = true;
+        }
+        // showingPlcHdr is a toggle, not a property with a val
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "showingPlcHdr")) {
+            const v = (0, parse_utils_1.boolToggle)(sdtPrEl, "showingPlcHdr");
+            if (v !== false) {
+                props.showingPlaceholder = true;
+            }
+        }
+        // w15:appearance (replaces the old misused showingPlcHdr)
+        const appearanceEl = (0, dom_1.findChild)(sdtPrEl, "w15:appearance");
+        if (appearanceEl) {
+            const v = appearanceEl.attributes["w15:val"] ?? appearanceEl.attributes["val"];
+            if (v === "boundingBox" || v === "tags" || v === "hidden") {
+                props.appearance = v;
+            }
+        }
+        // Dropdown list
+        const ddlEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "dropDownList");
+        if (ddlEl) {
+            const items = [];
+            for (const li of (0, parse_utils_1.findChildrenNs)(ddlEl, "listItem")) {
+                const item = { value: (0, parse_utils_1.attrVal)(li, "value") ?? "" };
+                const dt = (0, parse_utils_1.attrVal)(li, "displayText");
+                if (dt) {
+                    item.displayText = dt;
+                }
+                items.push(item);
+            }
+            props.dropdownList = items;
+        }
+        // ComboBox
+        const cbEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "comboBox");
+        if (cbEl) {
+            const items = [];
+            for (const li of (0, parse_utils_1.findChildrenNs)(cbEl, "listItem")) {
+                const item = { value: (0, parse_utils_1.attrVal)(li, "value") ?? "" };
+                const dt = (0, parse_utils_1.attrVal)(li, "displayText");
+                if (dt) {
+                    item.displayText = dt;
+                }
+                items.push(item);
+            }
+            props.comboBox = items;
+        }
+        // Date picker
+        const dateEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "date");
+        if (dateEl) {
+            const dateProp = {};
+            const fullDate = (0, parse_utils_1.attrVal)(dateEl, "fullDate");
+            if (fullDate) {
+                dateProp.fullDate = fullDate;
+            }
+            const dfEl = (0, parse_utils_1.findChildNs)(dateEl, "dateFormat");
+            if (dfEl) {
+                dateProp.dateFormat = (0, parse_utils_1.attrVal)(dfEl, "val");
+            }
+            const lidEl = (0, parse_utils_1.findChildNs)(dateEl, "lid");
+            if (lidEl) {
+                dateProp.lid = (0, parse_utils_1.attrVal)(lidEl, "val");
+            }
+            const storeEl = (0, parse_utils_1.findChildNs)(dateEl, "storeMappedDataAs");
+            if (storeEl) {
+                dateProp.storeMappedDataAs = (0, parse_utils_1.attrVal)(storeEl, "val");
+            }
+            props.date = dateProp;
+        }
+        // ID
+        const idEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "id");
+        if (idEl) {
+            const v = (0, parse_utils_1.attrInt)(idEl, "val");
+            if (v !== undefined) {
+                props.id = v;
+            }
+        }
+        // Data binding
+        const dbEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "dataBinding");
+        if (dbEl) {
+            const xpath = (0, parse_utils_1.attrVal)(dbEl, "xpath");
+            const storeItemId = (0, parse_utils_1.attrVal)(dbEl, "storeItemID");
+            if (xpath && storeItemId) {
+                const binding = {
+                    xpath,
+                    storeItemId
+                };
+                const prefixMappings = (0, parse_utils_1.attrVal)(dbEl, "prefixMappings");
+                if (prefixMappings) {
+                    binding.prefixMappings = prefixMappings;
+                }
+                props.dataBinding = binding;
+            }
+        }
+        // Placeholder
+        const phEl = (0, parse_utils_1.findChildNs)(sdtPrEl, "placeholder");
+        if (phEl) {
+            const docPartEl = (0, parse_utils_1.findChildNs)(phEl, "docPart");
+            if (docPartEl) {
+                props.placeholder = (0, parse_utils_1.attrVal)(docPartEl, "val");
+            }
+        }
+        // Boolean marker elements
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "richText")) {
+            props.richText = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "picture")) {
+            props.picture = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "group")) {
+            props.group = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "equation")) {
+            props.equation = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "citation")) {
+            props.citation = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "bibliography")) {
+            props.bibliography = true;
+        }
+        if ((0, parse_utils_1.findChildNs)(sdtPrEl, "temporary")) {
+            props.temporary = true;
+        }
+        // w15: repeating section
+        const rsEl = (0, dom_1.findChild)(sdtPrEl, "w15:repeatingSection");
+        if (rsEl) {
+            const rs = {};
+            // Read from child elements (correct per schema)
+            const titleEl = (0, dom_1.findChild)(rsEl, "w15:sectionTitle");
+            if (titleEl) {
+                const v = titleEl.attributes["w15:val"] ?? titleEl.attributes["val"];
+                if (v !== undefined) {
+                    rs.sectionTitle = v;
+                }
+            }
+            if ((0, dom_1.findChild)(rsEl, "w15:doNotAllowInsertDeleteSection")) {
+                rs.allowInsertDelete = false;
+            }
+            // Also accept attribute form for backwards compatibility
+            const stAttr = rsEl.attributes["w15:sectionTitle"];
+            if (stAttr !== undefined && rs.sectionTitle === undefined) {
+                rs.sectionTitle = stAttr;
+            }
+            const noInsDelAttr = rsEl.attributes["w15:doNotAllowInsertDeleteSection"];
+            if (noInsDelAttr !== undefined && rs.allowInsertDelete === undefined) {
+                rs.allowInsertDelete = noInsDelAttr === "0";
+            }
+            props.repeatingSection = rs;
+        }
+        if ((0, dom_1.findChild)(sdtPrEl, "w15:repeatingSectionItem")) {
+            props.repeatingSectionItem = true;
+        }
+    }
+    const content = [];
+    if (sdtContentEl) {
+        for (const child of sdtContentEl.children) {
+            if (child.type !== "element") {
+                continue;
+            }
+            const n = child.name.replace(/^w:/, "");
+            if (n === "p") {
+                content.push(parseParagraph(child, ctx));
+            }
+            else if (n === "tbl") {
+                content.push(parseTable(child, ctx));
+            }
+            else if (n === "r") {
+                content.push(parseRun(child));
+            }
+            else if (n === "sdt") {
+                // Nested SDT (e.g. repeating section item SDTs). Preserve the
+                // inner SDT verbatim — including its own properties — so data
+                // binding, alias, lock and similar metadata round-trip correctly.
+                const inner = parseSdt(child, ctx);
+                if (inner && inner.type === "sdt") {
+                    content.push(inner);
+                }
+            }
+        }
+    }
+    return { type: "sdt", properties: props, content };
+}
+function parseTocFromSdt(sdtContentEl, ctx) {
+    const toc = { type: "tableOfContents" };
+    const cachedParagraphs = [];
+    if (sdtContentEl) {
+        // Collect all instrText to assemble the complete TOC field instruction
+        let instrText = "";
+        const collectInstr = (el) => {
+            for (const child of el.children) {
+                if (child.type !== "element") {
+                    continue;
+                }
+                const name = child.name.replace(/^w:/, "");
+                if (name === "instrText") {
+                    instrText += (0, dom_1.textContent)(child);
+                }
+                else {
+                    collectInstr(child);
+                }
+            }
+        };
+        collectInstr(sdtContentEl);
+        if (instrText.trim()) {
+            (0, sdt_helpers_1.parseTocInstruction)(instrText, toc);
+        }
+        for (const child of sdtContentEl.children) {
+            if (child.type !== "element") {
+                continue;
+            }
+            const n = child.name.replace(/^w:/, "");
+            if (n === "p") {
+                cachedParagraphs.push(parseParagraph(child, ctx));
+            }
+        }
+    }
+    if (cachedParagraphs.length > 0) {
+        toc.cachedParagraphs = cachedParagraphs;
+    }
+    return toc;
+}
+/** Parse a TOC field instruction string (e.g. `TOC \o "1-3" \h \t "Style,1" \c "Figure"`). */
+// =============================================================================
+// Paragraph Parser
+// =============================================================================
+function parseRun(el) {
+    const rPrEl = (0, parse_utils_1.findChildNs)(el, "rPr");
+    return {
+        properties: rPrEl ? (0, properties_parsers_1.parseRunProperties)(rPrEl) : undefined,
+        content: parseRunContent(el)
+    };
+}
+function parseParagraph(pEl, ctx) {
+    const pPrEl = (0, parse_utils_1.findChildNs)(pEl, "pPr");
+    const children = [];
+    // Field state machine lives on ctx so that complex fields (TOC, INDEX,
+    // long REF/SEQ chains) can span paragraph boundaries — the matching
+    // `<w:fldChar fldCharType="end">` may occur in a later paragraph than the
+    // `begin`. Storing state on ctx is also safe because part-scoped parsers
+    // (header/footer/footnote/endnote/comment) save and reset it on entry.
+    const field = ctx.field;
+    for (const child of pEl.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        // Handle mc:AlternateContent — pick mc:Choice, fall back to mc:Fallback
+        let resolved = child;
+        if (child.name === "mc:AlternateContent") {
+            const choice = (0, dom_1.findChild)(child, "mc:Choice");
+            const fallback = (0, dom_1.findChild)(child, "mc:Fallback");
+            const chosen = choice ?? fallback;
+            if (chosen && chosen.children.length > 0) {
+                // The first element child inside Choice/Fallback is the real element
+                const inner = chosen.children.find(c => c.type === "element");
+                if (inner) {
+                    resolved = inner;
+                }
+                else {
+                    continue;
+                }
+            }
+            else {
+                continue;
+            }
+        }
+        const name = resolved.name.replace(/^w:/, "");
+        switch (name) {
+            case "r": {
+                // Check for fldChar and instrText inside the run
+                let hasFldChar = false;
+                for (const rc of resolved.children) {
+                    if (rc.type !== "element") {
+                        continue;
+                    }
+                    const rcName = rc.name.replace(/^w:/, "");
+                    if (rcName === "fldChar") {
+                        hasFldChar = true;
+                        const fldCharType = (0, parse_utils_1.attrVal)(rc, "fldCharType");
+                        if (fldCharType === "begin") {
+                            field.state = "instrText";
+                            field.instr = "";
+                            field.cached = "";
+                            // Capture run properties from this run for the field
+                            const rPrEl = (0, parse_utils_1.findChildNs)(resolved, "rPr");
+                            field.runProps = rPrEl ? (0, properties_parsers_1.parseRunProperties)(rPrEl) : undefined;
+                            // Parse ffData for legacy form fields
+                            const ffDataEl = (0, parse_utils_1.findChildNs)(rc, "ffData");
+                            field.formField = ffDataEl ? (0, form_field_parser_1.parseFfData)(ffDataEl) : undefined;
+                        }
+                        else if (fldCharType === "separate") {
+                            field.state = "cached";
+                        }
+                        else if (fldCharType === "end") {
+                            // Emit the assembled field as a Run with FieldContent
+                            const fc = {
+                                type: "field",
+                                instruction: field.instr.trim(),
+                                cachedValue: field.cached || undefined,
+                                formField: field.formField
+                            };
+                            children.push({
+                                properties: field.runProps,
+                                content: [fc]
+                            });
+                            field.state = "none";
+                            field.instr = "";
+                            field.cached = "";
+                            field.runProps = undefined;
+                            field.formField = undefined;
+                        }
+                    }
+                    else if (rcName === "instrText" && field.state === "instrText") {
+                        hasFldChar = true;
+                        field.instr += (0, dom_1.textContent)(rc);
+                    }
+                }
+                if (field.state === "cached") {
+                    // Collect cached text from this run
+                    for (const rc of resolved.children) {
+                        if (rc.type !== "element") {
+                            continue;
+                        }
+                        const rcName = rc.name.replace(/^w:/, "");
+                        if (rcName === "t") {
+                            field.cached += (0, dom_1.textContent)(rc);
+                        }
+                        else if (rcName === "fldChar") {
+                            // Already handled above
+                        }
+                    }
+                    if (!hasFldChar) {
+                        continue; // Skip adding this run normally
+                    }
+                }
+                if (field.state === "instrText" && hasFldChar) {
+                    continue; // Don't add begin/instrText runs as normal content
+                }
+                if (field.state === "none" && !hasFldChar) {
+                    // Detect a degenerate `<w:r>` whose only meaningful child is
+                    // `<w:commentReference>`. The OOXML schema requires the leaf
+                    // to live inside a w:r, but at the model level we represent
+                    // it as a paragraph-child `commentReference`. Hoisting here
+                    // means a round-trip preserves the model shape instead of
+                    // collapsing to `annotationReference`.
+                    let onlyCommentRefId;
+                    let onlyCommentRefSeen = false;
+                    let hasOtherMeaningfulChild = false;
+                    for (const rcc of resolved.children) {
+                        if (rcc.type !== "element") {
+                            continue;
+                        }
+                        const rccName = rcc.name.replace(/^w:/, "");
+                        if (rccName === "rPr") {
+                            continue;
+                        }
+                        if (rccName === "commentReference") {
+                            if (onlyCommentRefSeen) {
+                                // Multiple commentReferences in one run is malformed;
+                                // fall through to the generic run parser.
+                                hasOtherMeaningfulChild = true;
+                                break;
+                            }
+                            onlyCommentRefSeen = true;
+                            const idAttr = rcc.attributes["w:id"] ?? rcc.attributes["id"];
+                            const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                            if (!Number.isNaN(id)) {
+                                onlyCommentRefId = id;
+                            }
+                        }
+                        else {
+                            hasOtherMeaningfulChild = true;
+                            break;
+                        }
+                    }
+                    if (onlyCommentRefSeen && !hasOtherMeaningfulChild && onlyCommentRefId !== undefined) {
+                        children.push({ type: "commentReference", id: onlyCommentRefId });
+                    }
+                    else {
+                        children.push(parseRun(resolved));
+                    }
+                }
+                break;
+            }
+            case "fldSimple": {
+                // Simple field: <w:fldSimple w:instr=" PAGE "><w:r>...</w:r></w:fldSimple>
+                const instr = (0, parse_utils_1.attrVal)(resolved, "instr") ?? "";
+                let cached = "";
+                for (const fc of resolved.children) {
+                    if (fc.type === "element" && fc.name.replace(/^w:/, "") === "r") {
+                        for (const rc of fc.children) {
+                            if (rc.type === "element" && rc.name.replace(/^w:/, "") === "t") {
+                                cached += (0, dom_1.textContent)(rc);
+                            }
+                        }
+                    }
+                }
+                const fc = {
+                    type: "field",
+                    instruction: instr.trim(),
+                    cachedValue: cached || undefined
+                };
+                children.push({
+                    properties: undefined,
+                    content: [fc]
+                });
+                break;
+            }
+            case "hyperlink": {
+                const rId = resolved.attributes["r:id"];
+                const anchor = resolved.attributes["w:anchor"] ?? resolved.attributes["anchor"];
+                const tooltip = resolved.attributes["w:tooltip"] ?? resolved.attributes["tooltip"];
+                const historyAttr = resolved.attributes["w:history"] ?? resolved.attributes["history"];
+                const tgtFrame = resolved.attributes["w:tgtFrame"] ?? resolved.attributes["tgtFrame"];
+                const docLocation = resolved.attributes["w:docLocation"] ?? resolved.attributes["docLocation"];
+                const hRuns = [];
+                for (const hChild of resolved.children) {
+                    if (hChild.type === "element" && hChild.name.replace(/^w:/, "") === "r") {
+                        hRuns.push(parseRun(hChild));
+                    }
+                }
+                // Resolve URL from relMap. If the security policy disallows
+                // external targets, skip URL resolution entirely so the resulting
+                // Hyperlink only carries an anchor (or becomes a plain non-link
+                // wrapper). Internal anchor-only hyperlinks are unaffected.
+                let url;
+                if (rId && ctx.securityPolicy.allowExternalTargets) {
+                    const rel = ctx.relMap.get(rId);
+                    if (rel && rel.targetMode === "External") {
+                        url = rel.target;
+                    }
+                }
+                const hyperlink = {
+                    type: "hyperlink",
+                    rId,
+                    anchor,
+                    url,
+                    tooltip,
+                    children: hRuns
+                };
+                if (historyAttr === "1" || historyAttr === "true") {
+                    hyperlink.history = true;
+                }
+                if (tgtFrame) {
+                    hyperlink.tgtFrame = tgtFrame;
+                }
+                if (docLocation) {
+                    hyperlink.docLocation = docLocation;
+                }
+                children.push(hyperlink);
+                break;
+            }
+            case "bookmarkStart": {
+                const idAttr = resolved.attributes["w:id"] ?? resolved.attributes["id"];
+                const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                if (Number.isNaN(id)) {
+                    // Without a valid id we can't pair this with a bookmarkEnd; drop it
+                    // rather than fabricate id=0 (which would collide with every other
+                    // bookmark missing an id and corrupt cross-references on round-trip).
+                    break;
+                }
+                const bm = {
+                    type: "bookmarkStart",
+                    id,
+                    name: resolved.attributes["w:name"] ?? resolved.attributes["name"] ?? ""
+                };
+                const colFirst = resolved.attributes["w:colFirst"] ?? resolved.attributes["colFirst"];
+                if (colFirst !== undefined) {
+                    bm.colFirst = parseInt(colFirst, 10);
+                }
+                const colLast = resolved.attributes["w:colLast"] ?? resolved.attributes["colLast"];
+                if (colLast !== undefined) {
+                    bm.colLast = parseInt(colLast, 10);
+                }
+                const dcx = resolved.attributes["w:displacedByCustomXml"] ??
+                    resolved.attributes["displacedByCustomXml"];
+                if (dcx === "next" || dcx === "prev") {
+                    bm.displacedByCustomXml = dcx;
+                }
+                children.push(bm);
+                break;
+            }
+            case "bookmarkEnd": {
+                const idAttr = resolved.attributes["w:id"] ?? resolved.attributes["id"];
+                const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                if (Number.isNaN(id)) {
+                    break;
+                }
+                children.push({ type: "bookmarkEnd", id });
+                break;
+            }
+            case "commentRangeStart": {
+                const idAttr = resolved.attributes["w:id"] ?? resolved.attributes["id"];
+                const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                if (Number.isNaN(id)) {
+                    break;
+                }
+                children.push({ type: "commentRangeStart", id });
+                break;
+            }
+            case "commentRangeEnd": {
+                const idAttr = resolved.attributes["w:id"] ?? resolved.attributes["id"];
+                const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                if (Number.isNaN(id)) {
+                    break;
+                }
+                children.push({ type: "commentRangeEnd", id });
+                break;
+            }
+            case "commentReference": {
+                const idAttr = resolved.attributes["w:id"] ?? resolved.attributes["id"];
+                const id = idAttr !== undefined ? parseInt(idAttr, 10) : NaN;
+                if (Number.isNaN(id)) {
+                    break;
+                }
+                children.push({ type: "commentReference", id });
+                break;
+            }
+            case "ins": {
+                // Inserted run (track changes)
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(resolved);
+                if (rev) {
+                    for (const insChild of resolved.children) {
+                        if (insChild.type === "element" && insChild.name.replace(/^w:/, "") === "r") {
+                            children.push({
+                                type: "insertedRun",
+                                revision: rev,
+                                run: parseRun(insChild)
+                            });
+                        }
+                    }
+                }
+                break;
+            }
+            case "del": {
+                // Deleted run (track changes)
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(resolved);
+                if (rev) {
+                    for (const delChild of resolved.children) {
+                        if (delChild.type === "element" && delChild.name.replace(/^w:/, "") === "r") {
+                            children.push({
+                                type: "deletedRun",
+                                revision: rev,
+                                run: parseDeletedRun(delChild)
+                            });
+                        }
+                    }
+                }
+                break;
+            }
+            case "moveFrom": {
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(resolved);
+                if (rev) {
+                    for (const mfChild of resolved.children) {
+                        if (mfChild.type === "element" && mfChild.name.replace(/^w:/, "") === "r") {
+                            children.push({
+                                type: "movedFromRun",
+                                revision: rev,
+                                run: parseRun(mfChild)
+                            });
+                        }
+                    }
+                }
+                break;
+            }
+            case "moveTo": {
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(resolved);
+                if (rev) {
+                    for (const mtChild of resolved.children) {
+                        if (mtChild.type === "element" && mtChild.name.replace(/^w:/, "") === "r") {
+                            children.push({
+                                type: "movedToRun",
+                                revision: rev,
+                                run: parseRun(mtChild)
+                            });
+                        }
+                    }
+                }
+                break;
+            }
+            case "moveFromRangeStart":
+            case "moveFromRangeEnd":
+            case "moveToRangeStart":
+            case "moveToRangeEnd": {
+                const id = (0, parse_utils_1.attrInt)(resolved, "id");
+                if (id !== undefined) {
+                    const marker = {
+                        type: name,
+                        id
+                    };
+                    const author = (0, parse_utils_1.attrVal)(resolved, "author");
+                    if (author) {
+                        marker.author = author;
+                    }
+                    const date = (0, parse_utils_1.attrVal)(resolved, "date");
+                    if (date) {
+                        marker.date = date;
+                    }
+                    const mName = (0, parse_utils_1.attrVal)(resolved, "name");
+                    if (mName) {
+                        marker.name = mName;
+                    }
+                    children.push(marker);
+                }
+                break;
+            }
+            case "customXmlInsRangeStart":
+            case "customXmlInsRangeEnd":
+            case "customXmlDelRangeStart":
+            case "customXmlDelRangeEnd":
+            case "customXmlMoveFromRangeStart":
+            case "customXmlMoveFromRangeEnd":
+            case "customXmlMoveToRangeStart":
+            case "customXmlMoveToRangeEnd": {
+                const id = (0, parse_utils_1.attrInt)(resolved, "id");
+                if (id !== undefined) {
+                    const marker = {
+                        type: name,
+                        id
+                    };
+                    const author = (0, parse_utils_1.attrVal)(resolved, "author");
+                    if (author) {
+                        marker.author = author;
+                    }
+                    const date = (0, parse_utils_1.attrVal)(resolved, "date");
+                    if (date) {
+                        marker.date = date;
+                    }
+                    children.push(marker);
+                }
+                break;
+            }
+            case "smartTag":
+            case "customXml":
+            case "dir": {
+                // Semantic wrappers: flatten their children into the current
+                // paragraph. The wrapper's own properties element (smartTagPr,
+                // customXmlPr, …) is not a paragraph child and would otherwise
+                // fall through to the `default` branch below and be emitted as a
+                // bogus `opaqueParagraphChild` containing the properties XML —
+                // poisoning the paragraph on round-trip. Build a synthetic element
+                // that excludes those `*Pr` siblings before recursing.
+                const filteredChildren = resolved.children.filter(c => {
+                    if (c.type !== "element") {
+                        return true;
+                    }
+                    const ln = c.name.replace(/^w:/, "");
+                    return ln !== "smartTagPr" && ln !== "customXmlPr";
+                });
+                const surrogate = {
+                    ...resolved,
+                    children: filteredChildren
+                };
+                const subPara = parseParagraph(surrogate, ctx);
+                for (const sub of subPara.children) {
+                    children.push(sub);
+                }
+                break;
+            }
+            case "proofErr":
+            case "permStart":
+            case "permEnd":
+            case "lastRenderedPageBreak":
+                // Non-semantic markers; safely ignored
+                break;
+            default:
+                // Unknown paragraph child: preserve as opaque for round-trip fidelity
+                children.push({
+                    type: "opaqueParagraphChild",
+                    rawXml: (0, parse_utils_1.serializeElement)(resolved)
+                });
+                break;
+        }
+    }
+    const paraId = pEl.attributes["w14:paraId"];
+    const textId = pEl.attributes["w14:textId"];
+    const result = {
+        type: "paragraph",
+        properties: pPrEl ? (0, paragraph_section_parsers_1.parseParagraphProperties)(pPrEl) : undefined,
+        children
+    };
+    if (paraId) {
+        result.paraId = paraId;
+    }
+    if (textId) {
+        result.textId = textId;
+    }
+    return result;
+}
+/** Parse a deleted run (w:delText instead of w:t). */
+function parseDeletedRun(el) {
+    const rPrEl = (0, parse_utils_1.findChildNs)(el, "rPr");
+    const content = [];
+    for (const child of el.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        const name = child.name.replace(/^w:/, "");
+        if (name === "delText") {
+            content.push({ type: "text", text: (0, dom_1.textContent)(child) });
+        }
+        else if (name === "t") {
+            content.push({ type: "text", text: (0, dom_1.textContent)(child) });
+        }
+        else if (name === "br") {
+            content.push({
+                type: "break",
+                breakType: (0, parse_utils_1.attrVal)(child, "type")
+            });
+        }
+        else if (name === "tab") {
+            content.push({ type: "tab" });
+        }
+    }
+    return {
+        properties: rPrEl ? (0, properties_parsers_1.parseRunProperties)(rPrEl) : undefined,
+        content
+    };
+}
+// =============================================================================
+// Table Parser
+// =============================================================================
+function parseTableCell(el, ctx) {
+    const tcPrEl = (0, parse_utils_1.findChildNs)(el, "tcPr");
+    const content = [];
+    for (const child of el.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        const name = child.name.replace(/^w:/, "");
+        if (name === "p") {
+            content.push(parseParagraph(child, ctx));
+        }
+        else if (name === "tbl") {
+            content.push(parseTable(child, ctx));
+        }
+        else if (name === "sdt") {
+            // SDT inside a table cell. The TableCell.content union does not
+            // include StructuredDocumentTag, so we flatten the SDT's inner
+            // paragraphs/tables into the cell. SDT-level metadata (data binding,
+            // alias, repeating section, …) is lost on round-trip but visible
+            // content is preserved — better than dropping the runs entirely.
+            const sdt = parseSdt(child, ctx);
+            if (sdt && sdt.type === "sdt") {
+                for (const c of sdt.content) {
+                    if (c.type === "paragraph") {
+                        content.push(c);
+                    }
+                    else if (c.type === "table") {
+                        content.push(c);
+                    }
+                    // Run-only and nested-SDT children cannot live as direct
+                    // siblings of <w:p>/<w:tbl> in a <w:tc>, so they are dropped.
+                }
+            }
+        }
+    }
+    let props;
+    if (tcPrEl) {
+        const p = {};
+        const wEl = (0, parse_utils_1.findChildNs)(tcPrEl, "tcW");
+        if (wEl) {
+            p.width = (0, properties_parsers_1.parseTableWidth)(wEl);
+        }
+        const gsEl = (0, parse_utils_1.findChildNs)(tcPrEl, "gridSpan");
+        if (gsEl) {
+            p.gridSpan = (0, parse_utils_1.attrInt)(gsEl, "val");
+        }
+        const vmEl = (0, parse_utils_1.findChildNs)(tcPrEl, "vMerge");
+        if (vmEl) {
+            p.verticalMerge = ((0, parse_utils_1.attrVal)(vmEl, "val") ??
+                "continue");
+        }
+        const bordersEl = (0, parse_utils_1.findChildNs)(tcPrEl, "tcBorders");
+        if (bordersEl) {
+            p.borders = (0, table_properties_parsers_1.parseTableBorders)(bordersEl);
+        }
+        const shdEl = (0, parse_utils_1.findChildNs)(tcPrEl, "shd");
+        if (shdEl) {
+            p.shading = (0, properties_parsers_1.parseShading)(shdEl);
+        }
+        const vAlignEl = (0, parse_utils_1.findChildNs)(tcPrEl, "vAlign");
+        if (vAlignEl) {
+            p.verticalAlign = (0, parse_utils_1.attrVal)(vAlignEl, "val");
+        }
+        if ((0, parse_utils_1.findChildNs)(tcPrEl, "noWrap")) {
+            p.noWrap = true;
+        }
+        const textDirEl = (0, parse_utils_1.findChildNs)(tcPrEl, "textDirection");
+        if (textDirEl) {
+            p.textDirection = (0, parse_utils_1.attrVal)(textDirEl, "val");
+        }
+        const marginsEl = (0, parse_utils_1.findChildNs)(tcPrEl, "tcMar");
+        if (marginsEl) {
+            p.margins = (0, table_properties_parsers_1.parseTableCellMargins)(marginsEl);
+        }
+        // Conditional formatting
+        const cnfEl = (0, parse_utils_1.findChildNs)(tcPrEl, "cnfStyle");
+        if (cnfEl) {
+            p.cnfStyle = (0, parse_utils_1.attrVal)(cnfEl, "val");
+        }
+        // Hide cell end-of-cell marker
+        if ((0, parse_utils_1.findChildNs)(tcPrEl, "hideMark")) {
+            p.hideMark = true;
+        }
+        // Fit text
+        if ((0, parse_utils_1.findChildNs)(tcPrEl, "tcFitText")) {
+            p.fitText = true;
+        }
+        // Cell-level revisions
+        const cellInsEl = (0, parse_utils_1.findChildNs)(tcPrEl, "cellIns");
+        if (cellInsEl) {
+            const rev = (0, properties_parsers_1.parseRevisionInfo)(cellInsEl);
+            if (rev) {
+                p.inserted = { revision: rev };
+            }
+        }
+        const cellDelEl = (0, parse_utils_1.findChildNs)(tcPrEl, "cellDel");
+        if (cellDelEl) {
+            const rev = (0, properties_parsers_1.parseRevisionInfo)(cellDelEl);
+            if (rev) {
+                p.deleted = { revision: rev };
+            }
+        }
+        const cellMergeEl = (0, parse_utils_1.findChildNs)(tcPrEl, "cellMerge");
+        if (cellMergeEl) {
+            const vMerge = (0, parse_utils_1.attrVal)(cellMergeEl, "vMerge");
+            const rev = (0, properties_parsers_1.parseRevisionInfo)(cellMergeEl);
+            if (rev && (vMerge === "cont" || vMerge === "rest")) {
+                p.cellMerge = { vMerge, revision: rev };
+            }
+        }
+        // tcPrChange
+        const tcPrChangeEl = (0, parse_utils_1.findChildNs)(tcPrEl, "tcPrChange");
+        if (tcPrChangeEl) {
+            const rev = (0, properties_parsers_1.parseRevisionInfo)(tcPrChangeEl);
+            if (rev) {
+                const prev = (0, parse_utils_1.findChildNs)(tcPrChangeEl, "tcPr");
+                p.propertyChange = { revision: rev };
+                if (prev) {
+                    // Minimal: previousProperties won't recurse (avoid infinite recursion).
+                    // Just capture the presence of the change marker here.
+                }
+            }
+        }
+        props = p;
+    }
+    return { properties: props, content };
+}
+function parseTableRow(el, ctx) {
+    const trPrEl = (0, parse_utils_1.findChildNs)(el, "trPr");
+    const tblPrExEl = (0, parse_utils_1.findChildNs)(el, "tblPrEx");
+    const cells = [];
+    for (const child of el.children) {
+        if (child.type === "element" && child.name.replace(/^w:/, "") === "tc") {
+            cells.push(parseTableCell(child, ctx));
+        }
+    }
+    let props;
+    if (trPrEl || tblPrExEl) {
+        const p = {};
+        if (tblPrExEl) {
+            p.tblPrEx = (0, table_properties_parsers_1.parseTableProperties)(tblPrExEl);
+        }
+        if (trPrEl) {
+            const heightEl = (0, parse_utils_1.findChildNs)(trPrEl, "trHeight");
+            if (heightEl) {
+                p.height = {
+                    value: (0, parse_utils_1.attrInt)(heightEl, "val") ?? 0,
+                    rule: (0, parse_utils_1.attrVal)(heightEl, "hRule")
+                };
+            }
+            if ((0, parse_utils_1.findChildNs)(trPrEl, "tblHeader")) {
+                p.tableHeader = true;
+            }
+            if ((0, parse_utils_1.findChildNs)(trPrEl, "cantSplit")) {
+                p.cantSplit = true;
+            }
+            if ((0, parse_utils_1.findChildNs)(trPrEl, "hidden")) {
+                p.hidden = true;
+            }
+            const csEl = (0, parse_utils_1.findChildNs)(trPrEl, "tblCellSpacing");
+            if (csEl) {
+                p.cellSpacing = (0, properties_parsers_1.parseTableWidth)(csEl);
+            }
+            const insEl = (0, parse_utils_1.findChildNs)(trPrEl, "ins");
+            if (insEl) {
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(insEl);
+                if (rev) {
+                    p.inserted = { revision: rev };
+                }
+            }
+            const delEl = (0, parse_utils_1.findChildNs)(trPrEl, "del");
+            if (delEl) {
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(delEl);
+                if (rev) {
+                    p.deleted = { revision: rev };
+                }
+            }
+            const gbEl = (0, parse_utils_1.findChildNs)(trPrEl, "gridBefore");
+            if (gbEl) {
+                p.gridBefore = (0, parse_utils_1.attrInt)(gbEl, "val");
+            }
+            const gaEl = (0, parse_utils_1.findChildNs)(trPrEl, "gridAfter");
+            if (gaEl) {
+                p.gridAfter = (0, parse_utils_1.attrInt)(gaEl, "val");
+            }
+            const wbEl = (0, parse_utils_1.findChildNs)(trPrEl, "wBefore");
+            if (wbEl) {
+                p.widthBefore = (0, properties_parsers_1.parseTableWidth)(wbEl);
+            }
+            const waEl = (0, parse_utils_1.findChildNs)(trPrEl, "wAfter");
+            if (waEl) {
+                p.widthAfter = (0, properties_parsers_1.parseTableWidth)(waEl);
+            }
+            const cnfEl = (0, parse_utils_1.findChildNs)(trPrEl, "cnfStyle");
+            if (cnfEl) {
+                p.cnfStyle = (0, parse_utils_1.attrVal)(cnfEl, "val");
+            }
+            const trPrChangeEl = (0, parse_utils_1.findChildNs)(trPrEl, "trPrChange");
+            if (trPrChangeEl) {
+                const rev = (0, properties_parsers_1.parseRevisionInfo)(trPrChangeEl);
+                if (rev) {
+                    const prevTrPr = (0, parse_utils_1.findChildNs)(trPrChangeEl, "trPr");
+                    p.propertyChange = {
+                        revision: rev,
+                        previousProperties: prevTrPr ? parseRowPrInner(prevTrPr) : undefined
+                    };
+                }
+            }
+        }
+        props = p;
+    }
+    return { properties: props, cells };
+}
+/** Inner parse for row properties content (used by propertyChange recursion). */
+function parseRowPrInner(trPrEl) {
+    const p = {};
+    const heightEl = (0, parse_utils_1.findChildNs)(trPrEl, "trHeight");
+    if (heightEl) {
+        p.height = {
+            value: (0, parse_utils_1.attrInt)(heightEl, "val") ?? 0,
+            rule: (0, parse_utils_1.attrVal)(heightEl, "hRule")
+        };
+    }
+    if ((0, parse_utils_1.findChildNs)(trPrEl, "tblHeader")) {
+        p.tableHeader = true;
+    }
+    if ((0, parse_utils_1.findChildNs)(trPrEl, "cantSplit")) {
+        p.cantSplit = true;
+    }
+    return p;
+}
+function parseTable(tblEl, ctx) {
+    const tblPrEl = (0, parse_utils_1.findChildNs)(tblEl, "tblPr");
+    const gridEl = (0, parse_utils_1.findChildNs)(tblEl, "tblGrid");
+    const rows = [];
+    for (const child of tblEl.children) {
+        if (child.type === "element" && child.name.replace(/^w:/, "") === "tr") {
+            rows.push(parseTableRow(child, ctx));
+        }
+    }
+    let columnWidths;
+    if (gridEl) {
+        columnWidths = [];
+        for (const col of (0, parse_utils_1.findChildrenNs)(gridEl, "gridCol")) {
+            columnWidths.push(parseInt(col.attributes["w:w"] ?? col.attributes["w"] ?? "0", 10));
+        }
+    }
+    return {
+        type: "table",
+        properties: tblPrEl ? (0, table_properties_parsers_1.parseTableProperties)(tblPrEl) : undefined,
+        columnWidths,
+        rows
+    };
+}
+// =============================================================================
+// Footnotes/Endnotes Parser
+// =============================================================================
+function parseNotesXml(xmlStr, elementName, ctx) {
+    // Each note part is self-contained. Save and reset the field state so an
+    // unterminated complex field in the document body cannot bleed into a
+    // footnote/endnote and swallow its runs.
+    const savedField = ctx.field;
+    ctx.field = (0, reader_context_1.createFieldState)();
+    try {
+        const doc = (0, dom_1.parseXml)(xmlStr);
+        const root = doc.root;
+        const notes = [];
+        for (const noteEl of (0, parse_utils_1.findChildrenNs)(root, elementName)) {
+            const id = (0, parse_utils_1.attrInt)(noteEl, "id");
+            const type = (0, parse_utils_1.attrVal)(noteEl, "type");
+            // Skip auto-generated separator entries (default IDs -1 and 0)
+            // Real separators/continuationSeparators are regenerated by the writer.
+            if (type === "separator" || type === "continuationSeparator") {
+                continue;
+            }
+            if (id === undefined) {
+                continue;
+            }
+            const content = [];
+            for (const child of noteEl.children) {
+                if (child.type !== "element") {
+                    continue;
+                }
+                const ln = child.name.replace(/^w:/, "");
+                if (ln === "p") {
+                    content.push(parseParagraph(child, ctx));
+                }
+                else if (ln === "sdt") {
+                    // SDT inside a footnote/endnote: the model's content type is
+                    // `Paragraph[]`, so flatten the SDT's inner paragraphs (and their
+                    // descendants reachable as paragraphs). SDT-level metadata is
+                    // dropped here on round-trip — better than losing the visible
+                    // text completely.
+                    const sdt = parseSdt(child, ctx);
+                    if (sdt && sdt.type === "sdt") {
+                        for (const c of sdt.content) {
+                            if (c.type === "paragraph") {
+                                content.push(c);
+                            }
+                        }
+                    }
+                }
+            }
+            const note = { id, content };
+            if (type === "continuationNotice" || type === "normal") {
+                note.type = type;
+            }
+            notes.push(note);
+        }
+        return notes;
+    }
+    finally {
+        ctx.field = savedField;
+    }
+}
+// =============================================================================
+// Header/Footer Parser
+// =============================================================================
+function parseHeaderFooterXml(xmlStr, ctx) {
+    return parseHeaderFooterRoot((0, dom_1.parseXml)(xmlStr).root, ctx);
+}
+function parseHeaderFooterRoot(root, ctx) {
+    // Header/footer parts are self-contained: reset field state on entry so an
+    // unterminated complex field in the body does not consume header/footer runs.
+    const savedField = ctx.field;
+    ctx.field = (0, reader_context_1.createFieldState)();
+    try {
+        const children = [];
+        for (const child of root.children) {
+            if (child.type !== "element") {
+                continue;
+            }
+            const name = child.name.replace(/^w:/, "");
+            if (name === "p") {
+                children.push(parseParagraph(child, ctx));
+            }
+            else if (name === "tbl") {
+                children.push(parseTable(child, ctx));
+            }
+            else if (name === "sdt") {
+                // Flatten SDT children. HeaderFooterContent.children is
+                // `(Paragraph | Table)[]` so we hoist the inner paragraphs/tables;
+                // SDT-level metadata is dropped on round-trip but visible content
+                // is preserved (better than losing the runs entirely).
+                const sdt = parseSdt(child, ctx);
+                if (sdt && sdt.type === "sdt") {
+                    for (const c of sdt.content) {
+                        if (c.type === "paragraph") {
+                            children.push(c);
+                        }
+                        else if (c.type === "table") {
+                            children.push(c);
+                        }
+                    }
+                }
+            }
+        }
+        return { children };
+    }
+    finally {
+        ctx.field = savedField;
+    }
+}
+/** Detect watermark from a header's parsed XML root element. */
+// =============================================================================
+// Comments Parser
+// =============================================================================
+function parseCommentsXmlFromCtx(xmlStr, ctx) {
+    return (0, comments_parser_1.parseCommentsXml)(xmlStr, ctx, parseParagraph);
+}
+// =============================================================================
+// Main Document Parser
+// =============================================================================
+/** Recursively extract floating images, drawing shapes, and opaque drawings from an element tree. */
+function extractFloatingContent(el, images, shapes, opaqueDrawings, ctx) {
+    for (const child of el.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        if (child.name === "wp:anchor") {
+            // Check if this is a pic (image) or wsp (shape)
+            const graphicEl = (0, dom_1.findChild)(child, "a:graphic");
+            const graphicDataEl = graphicEl ? (0, dom_1.findChild)(graphicEl, "a:graphicData") : undefined;
+            const wspEl = graphicDataEl
+                ? ((0, dom_1.findChild)(graphicDataEl, "wps:wsp") ?? (0, parse_utils_1.findChildNs)(graphicDataEl, "wsp"))
+                : undefined;
+            if (wspEl) {
+                const shape = parseDrawingShape(child, wspEl, ctx);
+                if (shape) {
+                    shapes.push(shape);
+                }
+            }
+            else {
+                const fi = (0, image_parsers_1.parseFloatingImage)(child);
+                if (fi) {
+                    images.push(fi);
+                }
+                else {
+                    // Unknown anchor content (chart, diagram, etc.) — preserve as opaque
+                    const drawingEl = findDrawingParent(child);
+                    if (drawingEl) {
+                        const rids = new Set();
+                        (0, parse_utils_1.collectRIds)(drawingEl, rids);
+                        opaqueDrawings.push({
+                            type: "opaqueDrawing",
+                            rawXml: (0, parse_utils_1.serializeElement)(drawingEl),
+                            referencedRIds: [...rids]
+                        });
+                    }
+                }
+            }
+        }
+        else if (child.name === "wp:inline") {
+            // Inline drawings that aren't images — check for chart etc.
+            const graphicEl = (0, dom_1.findChild)(child, "a:graphic");
+            const graphicDataEl = graphicEl ? (0, dom_1.findChild)(graphicEl, "a:graphicData") : undefined;
+            if (graphicDataEl) {
+                const picEl = (0, dom_1.findChild)(graphicDataEl, "pic:pic") ?? (0, parse_utils_1.findChildNs)(graphicDataEl, "pic");
+                if (!picEl) {
+                    // Not an image — opaque inline drawing. We deliberately keep this
+                    // path even though parseDrawingContent also emits an `opaqueRun`
+                    // for the same drawing: the body-level pass below removes the
+                    // duplicate opaqueRun once we know this OpaqueDrawing has been
+                    // captured. Inside table cells / headers / footers / SDTs (where
+                    // this extractor is not invoked) the opaqueRun is the only
+                    // representation, so the drawing still survives a round-trip.
+                    const rids = new Set();
+                    (0, parse_utils_1.collectRIds)(child, rids);
+                    // Serialize the wp:inline element wrapped in w:drawing
+                    const rawXml = `<w:drawing>${(0, parse_utils_1.serializeElement)(child)}</w:drawing>`;
+                    opaqueDrawings.push({
+                        type: "opaqueDrawing",
+                        rawXml,
+                        referencedRIds: [...rids]
+                    });
+                }
+            }
+        }
+        else {
+            extractFloatingContent(child, images, shapes, opaqueDrawings, ctx);
+        }
+    }
+}
+/** Find the w:drawing ancestor element for serialization. */
+function findDrawingParent(anchorEl) {
+    // We don't have parent refs, so we construct a synthetic w:drawing wrapper
+    return {
+        type: "element",
+        name: "w:drawing",
+        attributes: {},
+        children: [anchorEl]
+    };
+}
+/**
+ * A paragraph is considered "empty" for the purposes of synthetic-anchor
+ * detection if it has no children, or if every child is a run whose content
+ * is either absent or contains only zero-length text segments. Inline images,
+ * fields, hyperlinks etc. all count as non-empty content. Properties (style
+ * id, alignment, etc.) are intentionally ignored — a single floating drawing
+ * that the writer wrapped in its own paragraph would never have meaningful
+ * paragraph properties.
+ */
+function isEmptyParagraph(para) {
+    if (!para.children || para.children.length === 0) {
+        return true;
+    }
+    for (const child of para.children) {
+        if (!(0, text_utils_1.isRun)(child)) {
+            // Anything with a `type` (hyperlink, bookmark, insertedRun, etc.) is
+            // considered meaningful content.
+            return false;
+        }
+        const run = child;
+        for (const c of run.content) {
+            if (c.type === "text") {
+                if (c.text.length > 0) {
+                    return false;
+                }
+            }
+            else {
+                // Any non-text run content (image, field, break, tab, ruby, etc.)
+                // makes the paragraph non-empty.
+                return false;
+            }
+        }
+    }
+    return true;
+}
+/**
+ * Remove `opaqueRun` entries that wrap a non-picture `<wp:inline>` drawing.
+ *
+ * These are emitted by parseDrawingContent so the drawing survives a
+ * round-trip when its containing paragraph lives inside a table cell, header,
+ * footer or SDT (places where the body-level extractor never runs). At the
+ * body level, however, the same drawings are also captured as `OpaqueDrawing`
+ * entries by extractFloatingContent — keeping both would duplicate the
+ * drawing in the produced document. Mutates `para.children`/run content in
+ * place.
+ */
+function stripInlineDrawingOpaqueRuns(para) {
+    for (const child of para.children) {
+        if (!(0, text_utils_1.isRun)(child)) {
+            continue;
+        }
+        const run = child;
+        let i = 0;
+        while (i < run.content.length) {
+            const c = run.content[i];
+            if (c.type === "opaqueRun" &&
+                c.rawXml.includes("<wp:inline") &&
+                !c.rawXml.includes("<pic:pic")) {
+                run.content.splice(i, 1);
+            }
+            else {
+                i++;
+            }
+        }
+    }
+}
+function parseDocumentXml(xmlStr, ctx) {
+    const doc = (0, dom_1.parseXml)(xmlStr);
+    const root = doc.root;
+    // Parse background
+    let background;
+    const bgEl = (0, parse_utils_1.findChildNs)(root, "background");
+    if (bgEl) {
+        const bg = {};
+        const color = (0, parse_utils_1.attrVal)(bgEl, "color");
+        if (color) {
+            bg.color = color;
+        }
+        const themeColor = (0, parse_utils_1.attrVal)(bgEl, "themeColor");
+        if (themeColor) {
+            bg.themeColor = themeColor;
+        }
+        const themeShade = (0, parse_utils_1.attrVal)(bgEl, "themeShade");
+        if (themeShade) {
+            bg.themeShade = themeShade;
+        }
+        const themeTint = (0, parse_utils_1.attrVal)(bgEl, "themeTint");
+        if (themeTint) {
+            bg.themeTint = themeTint;
+        }
+        background = bg;
+    }
+    const bodyEl = (0, parse_utils_1.findChildNs)(root, "body") ?? (0, dom_1.findChild)(root, "w:body");
+    if (!bodyEl) {
+        throw new errors_1.DocxParseError("Missing w:body element in document.xml");
+    }
+    const body = [];
+    let sectionProperties;
+    // Instead of extracting floating content from the entire body tree and
+    // appending at the end (which loses positional information), we now extract
+    // floating content per-paragraph and insert it immediately after the
+    // paragraph it belongs to.
+    for (const child of bodyEl.children) {
+        if (child.type !== "element") {
+            continue;
+        }
+        const name = child.name.replace(/^w:/, "");
+        switch (name) {
+            case "p": {
+                // Per OOXML schema (CT_OMathPara is a member of EG_PContent), a
+                // body-level math block is encoded as a paragraph containing a
+                // single m:oMathPara child. Detect that shape and surface it as
+                // a top-level MathBlock so the document model stays flat — the
+                // writer reverses this by re-wrapping math blocks in <w:p>.
+                const mathParaChildren = child.children.filter(c => c.type === "element" && c.name === "m:oMathPara");
+                const otherChildren = child.children.filter(c => {
+                    if (c.type !== "element") {
+                        return false;
+                    }
+                    // pPr is allowed; everything else (runs, hyperlinks, etc.) means
+                    // we're NOT a synthetic math wrapper and must keep the paragraph.
+                    return c.name !== "w:pPr" && c.name !== "m:oMathPara";
+                });
+                if (mathParaChildren.length > 0 && otherChildren.length === 0) {
+                    for (const oMathPara of mathParaChildren) {
+                        if (oMathPara.type === "element") {
+                            body.push((0, math_parser_1.parseMathBlock)(oMathPara));
+                        }
+                    }
+                    break;
+                }
+                const para = parseParagraph(child, ctx);
+                // Extract floating content from this paragraph element and insert
+                // immediately after it to preserve document position.
+                const pFloatingImages = [];
+                const pDrawingShapes = [];
+                const pOpaqueDrawings = [];
+                extractFloatingContent(child, pFloatingImages, pDrawingShapes, pOpaqueDrawings, ctx);
+                // parseDrawingContent (called from parseRunContent) already preserved
+                // every non-picture inline drawing as an `opaqueRun` so the drawing
+                // survives a round-trip even inside cells/headers/footers/SDTs where
+                // this body-level extractor is not invoked. At the body level
+                // extractFloatingContent has now also captured those drawings as
+                // `OpaqueDrawing` entries — that is the form chart-parser is wired
+                // to look for when promoting them to `ChartContent`. To avoid
+                // duplicate output we strip any opaqueRun whose XML embeds a
+                // <wp:inline> drawing from the paragraph here.
+                if (pOpaqueDrawings.length > 0) {
+                    stripInlineDrawingOpaqueRuns(para);
+                }
+                // If the paragraph is otherwise empty AND we did extract anchored
+                // content out of it, treat the paragraph as a synthetic carrier for
+                // the floating drawing(s) and drop it. Otherwise keeping it would
+                // cause a phantom empty paragraph to accumulate on every round-trip
+                // (writer wraps floating images in their own <w:p>, reader pulls the
+                // anchor out, leaving an empty <w:p> behind).
+                const hasAnchoredContent = pFloatingImages.length > 0 || pDrawingShapes.length > 0 || pOpaqueDrawings.length > 0;
+                const paragraphIsEmpty = isEmptyParagraph(para);
+                if (!(hasAnchoredContent && paragraphIsEmpty)) {
+                    body.push(para);
+                }
+                for (const fi of pFloatingImages) {
+                    body.push(fi);
+                }
+                for (const ds of pDrawingShapes) {
+                    body.push(ds);
+                }
+                for (const od of pOpaqueDrawings) {
+                    body.push(od);
+                }
+                break;
+            }
+            case "tbl":
+                body.push(parseTable(child, ctx));
+                break;
+            case "sectPr":
+                // Final section properties at the body level
+                sectionProperties = (0, paragraph_section_parsers_1.parseSectionProperties)(child);
+                break;
+            case "sdt": {
+                const sdtResult = parseSdt(child, ctx);
+                if (sdtResult) {
+                    body.push(sdtResult);
+                }
+                break;
+            }
+            case "altChunk": {
+                const rId = child.attributes["r:id"] ?? child.attributes["id"];
+                if (rId) {
+                    body.push({ type: "altChunk", rId });
+                }
+                break;
+            }
+            default: {
+                // Check for math namespace
+                if (child.name === "m:oMathPara") {
+                    body.push((0, math_parser_1.parseMathBlock)(child));
+                }
+                else if (child.name === "m:oMath") {
+                    body.push({ type: "math", content: (0, math_parser_1.parseMathContent)(child) });
+                }
+                // Check for VML pict (textbox)
+                if (name === "pict" || child.name === "w:pict") {
+                    const tb = parseTextBox(child, ctx);
+                    if (tb) {
+                        body.push(tb);
+                    }
+                }
+                break;
+            }
+        }
+    }
+    return { body, sectionProperties, background };
+}
+/**
+ * Read a DOCX file from a Uint8Array buffer and parse it into a DocxDocument model.
+ *
+ * If the file is encrypted (CFB format), provide a password via the options parameter
+ * to decrypt it automatically.
+ */
+async function readDocx(buffer, options) {
+    const policy = (0, policy_1.resolveSecurityPolicy)(options?.securityPolicy);
+    // Defense-in-depth: reject obviously oversized packages up-front. The same
+    // limit is also enforced incrementally during entry decompression so a
+    // pathological deflate stream can't slip past this check.
+    if (buffer.length > policy.maxPackageSize) {
+        throw new errors_1.DocxLimitExceededError("packageSize", policy.maxPackageSize, buffer.length, "compressed input larger than maxPackageSize");
+    }
+    // Detect encrypted DOCX (CFB format) before attempting ZIP parse.
+    // CFB signature: D0 CF 11 E0 A1 B1 1A E1
+    if (buffer.length >= 8 &&
+        buffer[0] === 0xd0 &&
+        buffer[1] === 0xcf &&
+        buffer[2] === 0x11 &&
+        buffer[3] === 0xe0 &&
+        buffer[4] === 0xa1 &&
+        buffer[5] === 0xb1 &&
+        buffer[6] === 0x1a &&
+        buffer[7] === 0xe1) {
+        if (options?.password != null) {
+            // Pass the security policy's package-size cap so a hostile CFB cannot
+            // claim a multi-GiB decrypted size and force a huge buffer allocation
+            // before the unzip stage even runs.
+            const decryptedZip = await (0, encryption_1.decryptDocx)(buffer, options.password, policy.maxPackageSize);
+            return readDocx(decryptedZip, options);
+        }
+        throw new errors_1.DocxEncryptedError();
+    }
+    try {
+        return await _readDocxInner(buffer, policy);
+    }
+    catch (e) {
+        if (e instanceof errors_1.DocxError) {
+            throw e;
+        }
+        const msg = e instanceof Error ? e.message : String(e);
+        throw new errors_1.DocxParseError(`Failed to read DOCX: ${msg}`, { cause: e });
+    }
+}
+async function _readDocxInner(buffer, policy) {
+    const reader = (0, read_archive_1.unzip)(buffer);
+    const entries = new Map();
+    let totalUncompressed = 0;
+    let entryCount = 0;
+    for await (const entry of reader.entries()) {
+        entryCount++;
+        if (entryCount > policy.maxPartCount) {
+            throw new errors_1.DocxLimitExceededError("partCount", policy.maxPartCount, entryCount, "ZIP contains more entries than maxPartCount");
+        }
+        const data = await entry.bytes();
+        if (data.length > policy.maxPartSize) {
+            throw new errors_1.DocxLimitExceededError("partSize", policy.maxPartSize, data.length, `entry "${entry.path}" exceeds maxPartSize`);
+        }
+        totalUncompressed += data.length;
+        if (totalUncompressed > policy.maxPackageSize) {
+            throw new errors_1.DocxLimitExceededError("packageSize", policy.maxPackageSize, totalUncompressed, "cumulative uncompressed entry size exceeds maxPackageSize");
+        }
+        // Normalize path: remove leading slash, normalize separators
+        const path = entry.path.replace(/^\//, "").replace(/\\/g, "/");
+        entries.set(path, data);
+    }
+    const decoder = internal_utils_1.utf8Decoder;
+    const consumedPaths = new Set(["[Content_Types].xml"]);
+    // Best-effort parse for non-critical parts (settings, numbering, styles,
+    // theme, fontTable, comments, charts, headers, footers, notes, …). A
+    // malformed auxiliary part should not prevent us from returning the main
+    // document body. Only parse failures on document.xml itself are fatal.
+    const tryParse = (fn) => {
+        try {
+            return fn();
+        }
+        catch {
+            return undefined;
+        }
+    };
+    // Parse [Content_Types].xml for accurate opaque part content types
+    const contentTypesXml = entries.get("[Content_Types].xml");
+    const contentTypeOverrides = new Map();
+    const contentTypeDefaults = new Map();
+    if (contentTypesXml) {
+        const ctDoc = (0, dom_1.parseXml)(decoder.decode(contentTypesXml));
+        for (const child of ctDoc.root.children) {
+            if (child.type !== "element") {
+                continue;
+            }
+            if (child.name === "Override") {
+                const partName = child.attributes["PartName"] ?? "";
+                const ct = child.attributes["ContentType"] ?? "";
+                if (partName && ct) {
+                    // Normalize: remove leading slash
+                    contentTypeOverrides.set(partName.replace(/^\//, ""), ct);
+                }
+            }
+            else if (child.name === "Default") {
+                const ext = child.attributes["Extension"] ?? "";
+                const ct = child.attributes["ContentType"] ?? "";
+                if (ext && ct) {
+                    contentTypeDefaults.set(ext.toLowerCase(), ct);
+                }
+            }
+        }
+    }
+    const getText = (path) => {
+        const data = entries.get(path);
+        if (data) {
+            consumedPaths.add(path);
+        }
+        return data ? decoder.decode(data) : undefined;
+    };
+    // Parse document relationships (must be before parseDocumentXml for hyperlink resolution)
+    // First, try to discover document path via package rels (supports Strict conformance)
+    let documentPartPath = "word/document.xml";
+    const packageRelsXmlEarly = getText("_rels/.rels");
+    if (packageRelsXmlEarly) {
+        const pkgRelsEarly = (0, reader_context_1.parseRelationships)(packageRelsXmlEarly);
+        for (const rel of pkgRelsEarly) {
+            if (rel.type === constants_1.RelType.OfficeDocument) {
+                let target = rel.target;
+                if (target.startsWith("/")) {
+                    target = target.substring(1);
+                }
+                documentPartPath = target;
+                break;
+            }
+        }
+    }
+    const docRelsPath = (0, parse_utils_1.getPartRelsPath)(documentPartPath);
+    const docRelsXml = getText(docRelsPath);
+    const docRels = docRelsXml ? (0, reader_context_1.parseRelationships)(docRelsXml) : [];
+    const _relMap = new Map(docRels.map(r => [r.id, r]));
+    // Create reader context for this parse session (replaces module-level _session)
+    const ctx = (0, reader_context_1.createReaderContext)(policy);
+    ctx.relMap = _relMap;
+    // Parse document.xml (required)
+    const documentXml = getText(documentPartPath);
+    if (!documentXml) {
+        throw new errors_1.DocxMissingPartError(documentPartPath);
+    }
+    const { body, sectionProperties, background } = parseDocumentXml(documentXml, ctx);
+    // Parse styles (resolve path via relationship, fallback to hardcoded)
+    const stylesPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Styles, documentPartPath) ?? "word/styles.xml";
+    const stylesXml = getText(stylesPath);
+    const stylesResult = stylesXml ? tryParse(() => (0, styles_parser_1.parseStyles)(stylesXml)) : undefined;
+    // Parse numbering
+    const numberingPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Numbering, documentPartPath) ?? "word/numbering.xml";
+    const numberingXml = getText(numberingPath);
+    const numberingResult = numberingXml
+        ? tryParse(() => (0, numbering_parser_1.parseNumberingXml)(numberingXml))
+        : undefined;
+    // Parse footnotes/endnotes — swap ctx.relMap to the notes part's own
+    // .rels (footnotes.xml.rels / endnotes.xml.rels) so hyperlinks and images
+    // inside notes resolve against the correct relationship map. Without this,
+    // any rId used in a footnote silently resolves to undefined.
+    const footnotesPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Footnotes, documentPartPath) ?? "word/footnotes.xml";
+    const footnotesXml = getText(footnotesPath);
+    let footnotes;
+    if (footnotesXml) {
+        const footnotesRelsPath = (0, parse_utils_1.getPartRelsPath)(footnotesPath);
+        const footnotesRelsXml = getText(footnotesRelsPath);
+        const savedRelMap = ctx.relMap;
+        if (footnotesRelsXml) {
+            const footnotesRels = (0, reader_context_1.parseRelationships)(footnotesRelsXml);
+            ctx.relMap = new Map(footnotesRels.map(r => [r.id, r]));
+            consumedPaths.add(footnotesRelsPath);
+        }
+        else {
+            ctx.relMap = new Map();
+        }
+        footnotes = tryParse(() => parseNotesXml(footnotesXml, "footnote", ctx));
+        ctx.relMap = savedRelMap;
+    }
+    const endnotesPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Endnotes, documentPartPath) ?? "word/endnotes.xml";
+    const endnotesXml = getText(endnotesPath);
+    let endnotes;
+    if (endnotesXml) {
+        const endnotesRelsPath = (0, parse_utils_1.getPartRelsPath)(endnotesPath);
+        const endnotesRelsXml = getText(endnotesRelsPath);
+        const savedRelMap = ctx.relMap;
+        if (endnotesRelsXml) {
+            const endnotesRels = (0, reader_context_1.parseRelationships)(endnotesRelsXml);
+            ctx.relMap = new Map(endnotesRels.map(r => [r.id, r]));
+            consumedPaths.add(endnotesRelsPath);
+        }
+        else {
+            ctx.relMap = new Map();
+        }
+        endnotes = tryParse(() => parseNotesXml(endnotesXml, "endnote", ctx));
+        ctx.relMap = savedRelMap;
+    }
+    // Parse headers/footers + detect watermarks
+    const headers = new Map();
+    const footers = new Map();
+    let watermark;
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.Header) {
+            const headerPartPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            const xml = getText(headerPartPath);
+            if (xml) {
+                // Parse header's own rels and switch ctx.relMap so hyperlinks/images
+                // referenced inside the header resolve against its own relationship map.
+                const headerRelsPath = (0, parse_utils_1.getPartRelsPath)(headerPartPath);
+                const headerRelsXml = getText(headerRelsPath);
+                const savedRelMap = ctx.relMap;
+                if (headerRelsXml) {
+                    const headerRels = (0, reader_context_1.parseRelationships)(headerRelsXml);
+                    const headerRelMap = new Map(headerRels.map(r => [r.id, r]));
+                    ctx.relMap = headerRelMap;
+                    consumedPaths.add(headerRelsPath);
+                }
+                else {
+                    ctx.relMap = new Map();
+                }
+                try {
+                    // Parse XML once, re-use for both header content and watermark detection
+                    const headerRoot = (0, dom_1.parseXml)(xml).root;
+                    headers.set(rel.id, { content: parseHeaderFooterRoot(headerRoot, ctx), rId: rel.id });
+                    if (!watermark) {
+                        watermark = (0, watermark_parser_1.detectWatermarkFromRoot)(headerRoot);
+                    }
+                }
+                catch {
+                    // Skip a malformed header; preserve other headers and the document.
+                }
+                ctx.relMap = savedRelMap;
+            }
+        }
+        else if (rel.type === constants_1.RelType.Footer) {
+            const footerPartPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            const xml = getText(footerPartPath);
+            if (xml) {
+                // Parse footer's own rels and switch ctx.relMap so hyperlinks/images
+                // referenced inside the footer resolve against its own relationship map.
+                const footerRelsPath = (0, parse_utils_1.getPartRelsPath)(footerPartPath);
+                const footerRelsXml = getText(footerRelsPath);
+                const savedRelMap = ctx.relMap;
+                if (footerRelsXml) {
+                    const footerRels = (0, reader_context_1.parseRelationships)(footerRelsXml);
+                    const footerRelMap = new Map(footerRels.map(r => [r.id, r]));
+                    ctx.relMap = footerRelMap;
+                    consumedPaths.add(footerRelsPath);
+                }
+                else {
+                    ctx.relMap = new Map();
+                }
+                try {
+                    footers.set(rel.id, { content: parseHeaderFooterXml(xml, ctx), rId: rel.id });
+                }
+                catch {
+                    // Skip a malformed footer; preserve other footers and the document.
+                }
+                ctx.relMap = savedRelMap;
+            }
+        }
+    }
+    // Parse settings
+    const settingsPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Settings, documentPartPath) ?? "word/settings.xml";
+    const settingsXml = getText(settingsPath);
+    const settings = settingsXml ? tryParse(() => (0, metadata_parsers_1.parseSettingsXml)(settingsXml)) : undefined;
+    // Parse web settings
+    const webSettingsPath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.WebSettings, documentPartPath) ?? "word/webSettings.xml";
+    const webSettingsXml = getText(webSettingsPath);
+    const webSettings = webSettingsXml ? tryParse(() => (0, metadata_parsers_1.parseWebSettings)(webSettingsXml)) : undefined;
+    // Parse people
+    const peoplePath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.People, documentPartPath) ?? "word/people.xml";
+    const peopleXml = getText(peoplePath);
+    const people = peopleXml ? tryParse(() => (0, metadata_parsers_1.parsePeople)(peopleXml)) : undefined;
+    // Parse thumbnail (from package rels — reuse already-parsed rels)
+    let thumbnail;
+    if (packageRelsXmlEarly) {
+        const pkgRels = (0, reader_context_1.parseRelationships)(packageRelsXmlEarly);
+        for (const rel of pkgRels) {
+            if (rel.type.endsWith("/thumbnail")) {
+                // Target in package rels is relative to package root; may include or exclude leading slash
+                let target = rel.target;
+                if (target.startsWith("/")) {
+                    target = target.substring(1);
+                }
+                // If the target doesn't include docProps/ prefix, add it (some writers emit bare filenames)
+                const normalized = target.includes("/") ? target : `docProps/${target}`;
+                consumedPaths.add(normalized);
+                const thumbData = entries.get(normalized);
+                if (thumbData) {
+                    const ext = (0, parse_utils_1.getFileExt)(normalized);
+                    const ct = ext === "jpeg" || ext === "jpg"
+                        ? "image/jpeg"
+                        : ext === "png"
+                            ? "image/png"
+                            : "image/x-wmf";
+                    thumbnail = {
+                        contentType: ct,
+                        data: thumbData
+                    };
+                }
+                break;
+            }
+        }
+    }
+    // Parse font table
+    const fontTablePath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.FontTable, documentPartPath) ?? "word/fontTable.xml";
+    const fontTableXml = getText(fontTablePath);
+    const fonts = fontTableXml ? tryParse(() => (0, doc_props_parsers_1.parseFontTableXml)(fontTableXml)) : undefined;
+    // Parse embedded fonts
+    let embeddedFonts;
+    const fontTableRelsXml = getText("word/_rels/fontTable.xml.rels");
+    if (fontTableRelsXml && fonts) {
+        const fontRels = (0, reader_context_1.parseRelationships)(fontTableRelsXml);
+        const efs = [];
+        // Build rId → { key } map from font table
+        const rIdToKey = new Map();
+        for (const f of fonts) {
+            if (f.embedRegular && f.embedRegularKey) {
+                rIdToKey.set(f.embedRegular, f.embedRegularKey);
+            }
+            if (f.embedBold && f.embedBoldKey) {
+                rIdToKey.set(f.embedBold, f.embedBoldKey);
+            }
+            if (f.embedItalic && f.embedItalicKey) {
+                rIdToKey.set(f.embedItalic, f.embedItalicKey);
+            }
+            if (f.embedBoldItalic && f.embedBoldItalicKey) {
+                rIdToKey.set(f.embedBoldItalic, f.embedBoldItalicKey);
+            }
+        }
+        for (const rel of fontRels) {
+            if (rel.type === constants_1.RelType.Font) {
+                const fontPath = (0, parse_utils_1.resolvePartPath)("word/fontTable.xml", rel.target);
+                consumedPaths.add(fontPath);
+                const data = entries.get(fontPath);
+                if (data) {
+                    const fileName = (0, parse_utils_1.getFileName)(rel.target);
+                    const fontKey = rIdToKey.get(rel.id);
+                    const ef = {
+                        rId: rel.id,
+                        data,
+                        fileName
+                    };
+                    if (fontKey) {
+                        ef.fontKey = fontKey;
+                    }
+                    efs.push(ef);
+                }
+            }
+        }
+        if (efs.length > 0) {
+            embeddedFonts = efs;
+        }
+    }
+    // Parse Custom XML parts (for SDT data binding)
+    const customXmlParts = [];
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.CustomXml) {
+            const targetPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            consumedPaths.add(targetPath);
+            const xmlContent = getText(targetPath);
+            if (!xmlContent) {
+                continue;
+            }
+            // Parse itemProps*.xml to get storeItemID
+            const fileName = (0, parse_utils_1.getFileName)(targetPath);
+            // itemProps file is typically at the same directory
+            const dir = targetPath.substring(0, targetPath.lastIndexOf("/"));
+            // Extract item number from fileName (e.g. "item1.xml" → "1")
+            const match = fileName.match(/item(\d+)\.xml$/);
+            let itemId = "";
+            let schemaReferences;
+            if (match) {
+                const num = match[1];
+                const propsPath = `${dir}/itemProps${num}.xml`;
+                consumedPaths.add(propsPath);
+                const propsXml = getText(propsPath);
+                if (propsXml) {
+                    const propsDoc = (0, dom_1.parseXml)(propsXml);
+                    const dsItemEl = propsDoc.root;
+                    const id = dsItemEl.attributes["ds:itemID"];
+                    if (id) {
+                        itemId = id.replace(/[{}]/g, "");
+                    }
+                    // Schema references
+                    const refs = [];
+                    const schemaRefsEl = (0, dom_1.findChild)(dsItemEl, "ds:schemaRefs") ?? (0, dom_1.findChild)(dsItemEl, "schemaRefs");
+                    if (schemaRefsEl) {
+                        for (const srChild of schemaRefsEl.children) {
+                            if (srChild.type === "element") {
+                                const uri = srChild.attributes["ds:uri"] ?? srChild.attributes["uri"];
+                                if (uri) {
+                                    refs.push(uri);
+                                }
+                            }
+                        }
+                    }
+                    if (refs.length > 0) {
+                        schemaReferences = refs;
+                    }
+                }
+            }
+            customXmlParts.push({
+                itemId,
+                xmlContent,
+                fileName,
+                schemaReferences
+            });
+        }
+    }
+    // Parse core properties
+    const corePropsXml = getText("docProps/core.xml");
+    const coreProperties = corePropsXml ? tryParse(() => (0, doc_props_parsers_1.parseCoreProps)(corePropsXml)) : undefined;
+    // Parse app properties
+    const appPropsXml = getText("docProps/app.xml");
+    const appProperties = appPropsXml ? tryParse(() => (0, doc_props_parsers_1.parseAppProps)(appPropsXml)) : undefined;
+    // Parse comments — switch ctx.relMap to comments.xml.rels so any
+    // hyperlinks/images referenced from inside comment paragraphs resolve
+    // against the comment part's own relationships rather than document.xml.rels.
+    const commentsXml = getText("word/comments.xml");
+    let comments;
+    if (commentsXml) {
+        const commentsRelsPath = "word/_rels/comments.xml.rels";
+        const commentsRelsXml = getText(commentsRelsPath);
+        const savedRelMap = ctx.relMap;
+        if (commentsRelsXml) {
+            const commentsRels = (0, reader_context_1.parseRelationships)(commentsRelsXml);
+            ctx.relMap = new Map(commentsRels.map(r => [r.id, r]));
+            consumedPaths.add(commentsRelsPath);
+        }
+        else {
+            ctx.relMap = new Map();
+        }
+        comments = tryParse(() => parseCommentsXmlFromCtx(commentsXml, ctx));
+        ctx.relMap = savedRelMap;
+    }
+    // Merge in commentsExtended.xml data if present
+    const commentsExtXml = getText("word/commentsExtended.xml");
+    if (commentsExtXml && comments) {
+        const extMap = tryParse(() => (0, comments_parser_1.parseCommentsExtendedXml)(commentsExtXml));
+        if (extMap) {
+            comments = comments.map(c => {
+                const firstPara = c.content[0];
+                if (!firstPara?.paraId) {
+                    return c;
+                }
+                const ext = extMap.get(firstPara.paraId);
+                if (!ext) {
+                    return c;
+                }
+                return {
+                    ...c,
+                    ...(ext.done !== undefined ? { done: ext.done } : {}),
+                    ...(ext.parentId !== undefined ? { parentId: ext.parentId } : {})
+                };
+            });
+        }
+    }
+    // Parse custom properties
+    const customPropsXml = getText("docProps/custom.xml");
+    const customProperties = customPropsXml
+        ? tryParse(() => (0, doc_props_parsers_1.parseCustomPropsXml)(customPropsXml))
+        : undefined;
+    // Parse theme
+    const themePath = (0, parse_utils_1.resolveRelTarget)(docRels, constants_1.RelType.Theme, documentPartPath) ?? "word/theme/theme1.xml";
+    const themeXml = getText(themePath);
+    const theme = themeXml ? tryParse(() => (0, metadata_parsers_1.parseThemeXml)(themeXml)) : undefined;
+    // Collect images from main document relationships
+    const images = [];
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.Image) {
+            const imgPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            consumedPaths.add(imgPath);
+            const data = entries.get(imgPath);
+            if (data) {
+                const fileName = (0, parse_utils_1.getFileName)(rel.target);
+                const ext = (0, parse_utils_1.getFileExt)(fileName) || "png";
+                images.push({
+                    data,
+                    mediaType: ext,
+                    fileName,
+                    rId: rel.id
+                });
+            }
+        }
+    }
+    // Also collect images from header/footer relationships to ensure full round-trip.
+    // Headers and footers have their own .rels files which may reference unique
+    // images, OR they may share a media file with the main document. In the
+    // latter case we keep the existing ImageDef but record the local rId as an
+    // alias so the packager can rebuild header1.xml.rels with the original
+    // (header-local) id intact.
+    const collectedImagePaths = new Map();
+    for (const img of images) {
+        collectedImagePaths.set(img.fileName, img);
+    }
+    for (const rel of docRels) {
+        if (rel.type !== constants_1.RelType.Header && rel.type !== constants_1.RelType.Footer) {
+            continue;
+        }
+        const partPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+        const partRelsPath = (0, parse_utils_1.getPartRelsPath)(partPath);
+        const partRelsXml = entries.get(partRelsPath);
+        if (!partRelsXml) {
+            continue;
+        }
+        const partRels = (0, reader_context_1.parseRelationships)(decoder.decode(partRelsXml));
+        for (const pRel of partRels) {
+            if (pRel.type !== constants_1.RelType.Image) {
+                continue;
+            }
+            const imgPath = (0, parse_utils_1.resolvePartPath)(partPath, pRel.target);
+            consumedPaths.add(imgPath);
+            const data = entries.get(imgPath);
+            if (!data) {
+                continue;
+            }
+            const fileName = (0, parse_utils_1.getFileName)(pRel.target);
+            const existing = collectedImagePaths.get(fileName);
+            if (existing) {
+                // Same physical file as one we already know — keep one ImageDef and
+                // append this part-local rId to its aliases (if it differs from the
+                // primary rId and isn't already recorded).
+                if (pRel.id && pRel.id !== existing.rId) {
+                    const aliases = existing.aliasRIds ? [...existing.aliasRIds] : [];
+                    if (!aliases.includes(pRel.id)) {
+                        aliases.push(pRel.id);
+                        existing.aliasRIds = aliases;
+                    }
+                }
+                continue;
+            }
+            const ext = (0, parse_utils_1.getFileExt)(fileName) || "png";
+            const newImg = {
+                data,
+                mediaType: ext,
+                fileName,
+                rId: pRel.id
+            };
+            images.push(newImg);
+            collectedImagePaths.set(fileName, newImg);
+        }
+    }
+    // Parse chart parts and replace opaque drawings with typed ChartContent
+    const chartRIdToChart = new Map();
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.Chart) {
+            const chartPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            consumedPaths.add(chartPath);
+            const chartXml = getText(chartPath);
+            if (chartXml) {
+                const chart = tryParse(() => (0, chart_parser_1.parseChartXml)(chartXml));
+                if (chart) {
+                    chartRIdToChart.set(rel.id, chart);
+                }
+            }
+        }
+    }
+    // Replace OpaqueDrawing items that reference chart rIds with proper ChartContent
+    if (chartRIdToChart.size > 0) {
+        (0, chart_parser_1.replaceOpaqueCharts)(body, chartRIdToChart);
+    }
+    // Parse ChartEx parts and replace opaque drawings with typed ChartExContent
+    const chartExRIdToContent = new Map();
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.ChartEx) {
+            const chartExPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            consumedPaths.add(chartExPath);
+            const chartExXml = getText(chartExPath);
+            if (chartExXml) {
+                const data = tryParse(() => (0, chart_parser_1.parseChartExXml)(chartExXml));
+                const content = {
+                    type: "chartEx",
+                    chartExXml,
+                    ...(data !== undefined && { data })
+                };
+                chartExRIdToContent.set(rel.id, content);
+            }
+        }
+    }
+    // Replace OpaqueDrawing items that reference ChartEx rIds with proper ChartExContent
+    if (chartExRIdToContent.size > 0) {
+        (0, chart_parser_1.replaceOpaqueChartExDrawings)(body, chartExRIdToContent);
+    }
+    // Detect document type from main document part content type
+    let docType;
+    const mainDocCT = contentTypeOverrides.get(documentPartPath) ?? contentTypeOverrides.get(`/${documentPartPath}`);
+    if (mainDocCT) {
+        if (mainDocCT.includes("template.main") && mainDocCT.includes("macroEnabled")) {
+            docType = "macroEnabledTemplate";
+        }
+        else if (mainDocCT.includes("template.main")) {
+            docType = "template";
+        }
+        else if (mainDocCT.includes("macroEnabled")) {
+            docType = "macroEnabledDocument";
+        }
+        // "document" is the default — only set if non-standard
+    }
+    // Extract VBA project binary for .docm/.dotm round-trip.
+    // Honour `preserveVbaProject`: if disabled, mark the relationship's
+    // target consumed (so opaqueParts won't retain it either) but leave
+    // `vbaProject` undefined so the produced model does not surface macro
+    // payloads to downstream consumers.
+    let vbaProject;
+    for (const rel of docRels) {
+        if (rel.type === constants_1.RelType.VbaProject) {
+            const vbaPath = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+            consumedPaths.add(vbaPath);
+            if (policy.preserveVbaProject) {
+                vbaProject = entries.get(vbaPath);
+            }
+            break;
+        }
+    }
+    // Resolve altChunk data: body elements of type "altChunk" reference a rId.
+    // The target file is stored in docRels + entries. We populate the altChunk
+    // body item with its data here AND mark the target path as consumed so the
+    // opaqueParts collector below does not retain a duplicate copy that would
+    // later be written back to the ZIP twice.
+    //
+    // Honour `preserveAltChunks`: when disabled, we still consume the target
+    // path (so it doesn't leak into opaqueParts) but skip data attachment
+    // and remove altChunk entries from the body before the document is
+    // returned. Embedded HTML/RTF in altChunks is a common attack vector
+    // for downstream renderers, so strict mode strips them entirely.
+    for (const item of body) {
+        if (item.type === "altChunk" && item.rId) {
+            const rel = _relMap.get(item.rId);
+            if (rel) {
+                const target = (0, parse_utils_1.resolvePartPath)(documentPartPath, rel.target);
+                const targetData = entries.get(target);
+                if (targetData) {
+                    consumedPaths.add(target);
+                    if (policy.preserveAltChunks) {
+                        const fileName = (0, parse_utils_1.getFileName)(target);
+                        const mItem = item;
+                        mItem.data = targetData;
+                        mItem.fileName = fileName;
+                        // Infer content type from extension
+                        const ext = fileName ? (0, parse_utils_1.getFileExt)(fileName) : "";
+                        if (ext === "html" || ext === "htm") {
+                            mItem.contentType = "text/html";
+                        }
+                        else if (ext === "rtf") {
+                            mItem.contentType = "text/rtf";
+                        }
+                        else if (ext === "txt") {
+                            mItem.contentType = "text/plain";
+                        }
+                    }
+                }
+            }
+        }
+    }
+    // Remove altChunk body entries entirely when not preserving them.
+    if (!policy.preserveAltChunks) {
+        for (let i = body.length - 1; i >= 0; i--) {
+            if (body[i].type === "altChunk") {
+                body.splice(i, 1);
+            }
+        }
+    }
+    // Collect opaque (unrecognized) parts for round-trip preservation
+    const opaqueParts = [];
+    for (const [path, data] of entries) {
+        // Skip consumed paths and all .rels files (structural)
+        if (consumedPaths.has(path) || path.includes("_rels/")) {
+            continue;
+        }
+        // Honour `preserveOleObjects`: when disabled, drop OLE embedding
+        // binaries (word/embeddings/*.bin and similar) before they reach the
+        // returned model. The relationship targets remain in their parent
+        // part's .rels, so the caller is responsible for stripping or
+        // ignoring those if they need a fully-clean document.
+        if (!policy.preserveOleObjects &&
+            (path.startsWith("word/embeddings/") || (path.endsWith(".bin") && path.includes("embed")))) {
+            continue;
+        }
+        // Parse rels for this part if they exist
+        const partRelsPath = (0, parse_utils_1.getPartRelsPath)(path);
+        const partRelsData = entries.get(partRelsPath);
+        let relationships;
+        if (partRelsData) {
+            const rels = (0, reader_context_1.parseRelationships)(decoder.decode(partRelsData));
+            relationships = rels.map(r => ({
+                id: r.id,
+                type: r.type,
+                target: r.target,
+                // Preserve the source string verbatim ("External", "Internal", or
+                // any non-standard value) so opaque round-trip is byte-faithful.
+                targetMode: r.targetMode
+            }));
+        }
+        // Resolve content type from [Content_Types].xml (override > default by extension)
+        let contentType = contentTypeOverrides.get(path);
+        if (!contentType) {
+            const ext = (0, parse_utils_1.getFileExt)(path);
+            contentType = contentTypeDefaults.get(ext);
+        }
+        opaqueParts.push({ path, data, contentType, relationships });
+    }
+    return {
+        ...(docType ? { docType } : {}),
+        body,
+        sectionProperties,
+        styles: stylesResult?.styles,
+        docDefaults: stylesResult?.docDefaults,
+        abstractNumberings: numberingResult?.abstractNums,
+        numberingInstances: numberingResult?.instances,
+        numPicBullets: numberingResult?.numPicBullets && numberingResult.numPicBullets.length > 0
+            ? numberingResult.numPicBullets
+            : undefined,
+        headers: headers.size > 0 ? headers : undefined,
+        footers: footers.size > 0 ? footers : undefined,
+        footnotes: footnotes && footnotes.length > 0 ? footnotes : undefined,
+        endnotes: endnotes && endnotes.length > 0 ? endnotes : undefined,
+        images: images.length > 0 ? images : undefined,
+        fonts: fonts && fonts.length > 0 ? fonts : undefined,
+        embeddedFonts: embeddedFonts && embeddedFonts.length > 0 ? embeddedFonts : undefined,
+        customXmlParts: customXmlParts.length > 0 ? customXmlParts : undefined,
+        webSettings,
+        thumbnail,
+        people: people && people.length > 0 ? people : undefined,
+        settings,
+        coreProperties,
+        appProperties,
+        comments: comments && comments.length > 0 ? comments : undefined,
+        background,
+        customProperties: customProperties && customProperties.length > 0 ? customProperties : undefined,
+        theme,
+        watermark,
+        opaqueParts: opaqueParts.length > 0 ? opaqueParts : undefined,
+        vbaProject
+    };
+}