0ef00a5 · oficina/docs

diff --git a/docs/superpowers/specs/2026-09-08-docx-fidelity-corpus.md b/docs/superpowers/specs/2026-09-08-docx-fidelity-corpus.md

new file mode 100644

index 0000000..1fb271c

--- /dev/null

+++ b/docs/superpowers/specs/2026-09-08-docx-fidelity-corpus.md

@@ -0,0 +1,40 @@

+# DOCX fidelity corpus — internal review proposal

+

+Scope: synthetic fixtures and regression evidence only, based main87662f5b246004aca604a4b040cb6691cfd7f130. No importer/exporter implementation, new UI/copy, schema change, merge or deployment. Existing import branch worktree-import-documents e64fcc7 and discussion20 remain the owners. Frozen links293ccd0 and Estilo candidates remain untouched.

+

+## Actual capability boundary

+

+Main has no internal/importer, POST /documents/import, DOCX parser or DOCX export route. Its real exporters are Markdown and PDF. This absence is recorded, not encoded as a test that merely expects404. The existing import branch implements Rust canonical JSON→Y.XmlFragment task0 only; its planned store/HTTP/parser work is unimplemented and DOCX is plan2. Its recorded fragment_text/ABI-versus-snapshot-version rulings remain pending. This proposal neither changes those rulings nor bypasses Docs' spec-before-plan and whole-branch Fable gates.

+

+Main's editor inherits basic link/image nodes; its server/export models do not preserve their full attributes. Tables are not in the current editor schema. The actual Markdown/PDF projection probe on main recorded: link label survives but target disappears; table cell text is flattened to paragraphs; image disappears. PDF returns no substitution report for those losses. This was a temporary measurement probe, not a passing regression that would require loss forever. Evidence /tmp/docs-docx-loss-probe.log. Frozen link work owns that first correction. No DOCX input was parsed to obtain these projection results.

+

+Main spec§11.2 requires named, located import loss reporting; §11.3 promises clean DOCX and Markdown lossless within its subset. Import discussion20's revision3 specifies warning enums and structured locators with no document text/link targets in reports. It currently proposes warned degradation for links/tables/images. Future review must reconcile that subset with the newer link implementation and full-suite migration expectations; do not quietly turn warned degradation into claimed full fidelity or silently flatten a document.

+

+## Checked-in corpus

+

+All files in fixtures/docx are generated from synthetic literals; no external documents, network downloads or customer content. ZIP entries have fixed metadata. generate.py creates valid XML parts, content types and OPC relationships, including styles/numbering. generate-update.mjs creates a separate supported-subset Yjs projection via pinned upstream ProseMirror/Yjs packages; it is not a DOCX importer and does not claim to reuse the complete editor schema.

+

+- core.docx: heading1, paragraph, bold/italic runs, nested bullet list and decimal list. core.expected.json is the independent future-import structural oracle; core.expected.md specifies current Markdown projection; core.update.bin is the separate upstream-authored Yjs projection.

+- alternate-body.docx: identical body at word/document2.xml, selected by package officeDocument relationship. A fixed word/document.xml parser must fail this case.

+- numbered-heading.docx: Heading1 plus numPr on one paragraph. The expectation remains heading1, not a list item, per import revision3.

+- link.docx: Reference label and external relationship to https://example.test/reference. Preserve the target or produce the reviewed located warning/refusal; never fetch the URL during import.

+- table.docx: two rows and two columns with distinct cells. Preserve geometry/cell ordering, or explicitly report the reviewed degradation; success with flattened text and no report is failure.

+- image.docx: embedded synthetic one-pixel PNG, drawing relationship, extent and alt text. Preserve media/dimensions/alt and per-person custody, or explicitly report the reviewed unsupported construct. No external image fetching.

+

+features.expected.json captures the non-core semantic expectations without inventing future asset IDs, image upload routes or table node APIs. The unit validator checks ZIP safety/uniqueness, XML well-formedness, body relationship resolution, internal targets, the feature-bearing source shapes and PNG dimensions. This is structural OPC evidence, not certification by Word/LibreOffice or proof of DOCX round-trip pagination.

+

+## What the real tests certify

+

+internal/export/docx_corpus_test.go passes the supported expected tree through the actual Markdown/PDF exporters. Exact Markdown checks heading/list nesting and marks; PDF text extraction checks all source words through the generated PDF's text maps. internal/docs/docx_projection_test.go applies the upstream Yjs projection through real Store.ApplyUpdate, then calls authenticated export.md/export.pdf routes and checks owner isolation. The Markdown result matches the structural expectation. The route's PDF media/header check is not presented as a PDF semantic test; the exporter-level extraction test supplies that evidence.

+

+Mutation evidence: deleting heading markers in actual Markdown output fails both projection tests for changed content, not a route error. Deleting the PDF content fails copied-text preservation, and removing the link relationship mode fails the structural fixture test. No test claims the nonexistent DOCX importer round-tripped anything. Existing import owner's Rust→Yjs cross-implementation gate is a separate dependency and must eventually accept these canonical expectations, with the actual shared editor schema rather than a drifting clone.

+

+## Implementation proposal to review next

+

+1. Resolve or explicitly reaffirm discussion20's two recorded rulings, then finish the reviewed import mechanism on its existing branch. Preserve safe JS client IDs, the real Store.ApplyUpdate path, sealed content, report-write failure cleanup and per-person grants. A report must not leak document text or URL targets into plaintext metadata.

+2. Connect a bounded Go ZIP/XML DOCX parser to this corpus. Resolve officeDocument and its relationship parts rather than fixed filenames; validate duplicate/escaping entries, decompressed size/part/depth budgets, numbering/styles, heading-over-numPr precedence and explicit XML breaks. Do not follow external relationships. Reject malformed/ambiguous packages without creating a partial document.

+3. For each supported input, compare canonical semantic output to checked-in expectations, build through the reviewed Rust importer and read through Yjs/real editor schema, then store/reload/export through actual routes. Content-equivalence expectations must ignore irrelevant ZIP order, namespace prefixes and relationship IDs while retaining node ordering, list depth, marks and destinations.

+4. Add tables/images only behind independently settled schema/media-custody contracts. Reuse the frozen link contract once merged; do not create a second URL policy. Before those capabilities exist, enforce the reviewed visible report/refusal path and transactional cleanup. This evidence branch does not choose a new policy for loss acceptance.

+5. Implement DOCX export as a separate reviewed slice, then add import→stored Yjs→DOCX→reimport semantic comparisons using this same corpus. Until that exists, report export parity as missing. Formatting beyond these synthetic cases, comments, page setup, complex numbering, headers/footers and tracked changes need additional corpus cases and their own settled scope; no 'replaces Word' claim follows from this small corpus.

+

+No deployment is required for this evidence deliverable. Product implementation and its generated UI artifacts still require their normal independent review/copy/CI/canary gates.

diff --git a/fixtures/docx/alternate-body.docx b/fixtures/docx/alternate-body.docx

new file mode 100644

index 0000000..bcd962c

Binary files /dev/null and b/fixtures/docx/alternate-body.docx differ

diff --git a/fixtures/docx/core.docx b/fixtures/docx/core.docx

new file mode 100644

index 0000000..97b7395

Binary files /dev/null and b/fixtures/docx/core.docx differ

diff --git a/fixtures/docx/core.expected.json b/fixtures/docx/core.expected.json

new file mode 100644

index 0000000..b2d9067

--- /dev/null

+++ b/fixtures/docx/core.expected.json

@@ -0,0 +1 @@

+{"type":"doc","content":[{"type":"heading","attrs":{"level":1},"content":[{"type":"text","text":"Section"}]},{"type":"paragraph","content":[{"type":"text","text":"Plain "},{"type":"text","text":"bold","marks":[{"type":"strong"}]},{"type":"text","text":" "},{"type":"text","text":"italic","marks":[{"type":"em"}]}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"type":"text","text":"Outer"}]},{"type":"bullet_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"type":"text","text":"Inner"}]}]}]}]}]},{"type":"ordered_list","content":[{"type":"list_item","content":[{"type":"paragraph","content":[{"type":"text","text":"First"}]}]},{"type":"list_item","content":[{"type":"paragraph","content":[{"type":"text","text":"Second"}]}]}]}]}

diff --git a/fixtures/docx/core.expected.md b/fixtures/docx/core.expected.md

new file mode 100644

index 0000000..603723f

--- /dev/null

+++ b/fixtures/docx/core.expected.md

@@ -0,0 +1,10 @@

+# Section

+

+Plain **bold** *italic*

+

+- Outer

+

+ - Inner

+

+1. First

+2. Second

diff --git a/fixtures/docx/core.update.bin b/fixtures/docx/core.update.bin

new file mode 100644

index 0000000..e982d0e

Binary files /dev/null and b/fixtures/docx/core.update.bin differ

diff --git a/fixtures/docx/features.expected.json b/fixtures/docx/features.expected.json

new file mode 100644

index 0000000..16252ee

--- /dev/null

+++ b/fixtures/docx/features.expected.json

@@ -0,0 +1 @@

+{"alternate-body":{"same_semantics_as":"core"},"numbered-heading":{"block":"heading","level":1,"text":"Numbered heading","not":"list_item"},"link":{"label":"Reference","target":"https://example.test/reference","relationship_mode":"External"},"table":{"rows":[["Cell A","Cell B"],["Cell C","Cell D"]]},"image":{"media":"word/media/pixel.png","mime":"image/png","width_px":1,"height_px":1,"alt":"One green pixel"}}

diff --git a/fixtures/docx/generate-update.mjs b/fixtures/docx/generate-update.mjs

new file mode 100644

index 0000000..ba4cb07

--- /dev/null

+++ b/fixtures/docx/generate-update.mjs

@@ -0,0 +1,16 @@

+// Synthetic supported-subset projection, not a DOCX parser. Upstream's

+// ProseMirror/Yjs bridge authors bytes for the existing Store/export path.

+import { readFileSync, writeFileSync } from 'node:fs'

+import * as Y from 'yjs'

+import { Schema } from 'prosemirror-model'

+import { schema as basic } from 'prosemirror-schema-basic'

+import { addListNodes } from 'prosemirror-schema-list'

+import { prosemirrorToYXmlFragment } from 'y-prosemirror'

+const schema = new Schema({nodes: addListNodes(basic.spec.nodes, 'paragraph block*', 'block'), marks: basic.spec.marks})

+const expected = JSON.parse(readFileSync(new URL('./core.expected.json', import.meta.url)))

+const node = schema.nodeFromJSON(expected)

+node.check()

+const doc = new Y.Doc({guid: 'docx-core-projection'})

+doc.clientID = 8107

+prosemirrorToYXmlFragment(node, doc.getXmlFragment('prosemirror'))

+writeFileSync(new URL('./core.update.bin', import.meta.url), Y.encodeStateAsUpdate(doc))

diff --git a/fixtures/docx/generate.py b/fixtures/docx/generate.py

new file mode 100644

index 0000000..219efcd

--- /dev/null

+++ b/fixtures/docx/generate.py

@@ -0,0 +1,47 @@

+"""Deterministic synthetic OPC fixtures; no document conversion code."""

+import json

+from pathlib import Path

+import struct

+import zipfile

+import zlib

+

+ROOT = Path(__file__).parent

+W = 'http://schemas.openxmlformats.org/wordprocessingml/2006/main'

+R = 'http://schemas.openxmlformats.org/officeDocument/2006/relationships'

+P = 'http://schemas.openxmlformats.org/package/2006/relationships'

+

+def paragraph(text, properties=''):

+ return '<w:p>' + properties + '<w:r><w:t xml:space="preserve">' + text + '</w:t></w:r></w:p>'

+

+core = '<w:p><w:pPr><w:pStyle w:val="Heading1"/></w:pPr><w:r><w:t>Section</w:t></w:r></w:p>'

+core += '<w:p><w:r><w:t xml:space="preserve">Plain </w:t></w:r><w:r><w:rPr><w:b/></w:rPr><w:t>bold</w:t></w:r><w:r><w:t xml:space="preserve"> </w:t></w:r><w:r><w:rPr><w:i/></w:rPr><w:t>italic</w:t></w:r></w:p>'

+for text, num, level in [('Outer',1,0),('Inner',1,1),('First',2,0),('Second',2,0)]:

+ core += paragraph(text, f'<w:pPr><w:numPr><w:ilvl w:val="{level}"/><w:numId w:val="{num}"/></w:numPr></w:pPr>')

+numbered = paragraph('Numbered heading', '<w:pPr><w:pStyle w:val="Heading1"/><w:numPr><w:ilvl w:val="0"/><w:numId w:val="2"/></w:numPr></w:pPr>')

+link = '<w:p><w:hyperlink r:id="rLink"><w:r><w:t>Reference</w:t></w:r></w:hyperlink></w:p>'

+table = '<w:tbl><w:tblPr/><w:tblGrid><w:gridCol w:w="2400"/><w:gridCol w:w="2400"/></w:tblGrid>' + ''.join('<w:tr>'+''.join('<w:tc><w:tcPr/>'+paragraph(c)+'</w:tc>' for c in row)+'</w:tr>' for row in [('Cell A','Cell B'),('Cell C','Cell D')])+'</w:tbl>'

+image = '<w:p><w:r><w:drawing><wp:inline><wp:extent cx="9525" cy="9525"/><wp:docPr id="1" name="Synthetic pixel" descr="One green pixel"/><a:graphic><a:graphicData uri="http://schemas.openxmlformats.org/drawingml/2006/picture"><pic:pic><pic:nvPicPr><pic:cNvPr id="1" name="pixel.png"/><pic:cNvPicPr/></pic:nvPicPr><pic:blipFill><a:blip r:embed="rImage"/><a:stretch><a:fillRect/></a:stretch></pic:blipFill><pic:spPr><a:xfrm><a:off x="0" y="0"/><a:ext cx="9525" cy="9525"/></a:xfrm><a:prstGeom prst="rect"><a:avLst/></a:prstGeom></pic:spPr></pic:pic></a:graphicData></a:graphic></wp:inline></w:drawing></w:r></w:p>'

+numbering = f'<w:numbering xmlns:w="{W}">' + ''.join(f'<w:abstractNum w:abstractNumId="{n}">'+''.join(f'<w:lvl w:ilvl="{level}"><w:start w:val="1"/><w:numFmt w:val="{kind}"/><w:lvlText w:val="{mark}"/></w:lvl>' for level in range(2))+f'</w:abstractNum><w:num w:numId="{n}"><w:abstractNumId w:val="{n}"/></w:num>' for n,kind,mark in [(1,'bullet','•'),(2,'decimal','%1.')])+'</w:numbering>'

+styles=f'<w:styles xmlns:w="{W}"><w:style w:type="paragraph" w:styleId="Heading1"><w:name w:val="heading 1"/><w:pPr><w:outlineLvl w:val="0"/></w:pPr></w:style></w:styles>'

+def chunk(kind, data):

+ return struct.pack('>I',len(data))+kind+data+struct.pack('>I',zlib.crc32(kind+data)&0xffffffff)

+png=b'\x89PNG\r\n\x1a\n'+chunk(b'IHDR',struct.pack('>IIBBBBB',1,1,8,2,0,0,0))+chunk(b'IDAT',zlib.compress(b'\x00\x00\xff\x00'))+chunk(b'IEND',b'')

+manifest=[]

+for name,body,part in [('core',core,'word/document.xml'),('alternate-body',core,'word/document2.xml'),('numbered-heading',numbered,'word/document.xml'),('link',link,'word/document.xml'),('table',table,'word/document.xml'),('image',image,'word/document.xml')]:

+ rels=[('rStyles','styles','styles.xml',''),('rNumbering','numbering','numbering.xml','')]

+ if name=='link': rels.append(('rLink','hyperlink','https://example.test/reference',' TargetMode="External"'))

+ if name=='image': rels.append(('rImage','image','media/pixel.png',''))

+ files={

+ '[Content_Types].xml':'<Types xmlns="http://schemas.openxmlformats.org/package/2006/content-types"><Default Extension="rels" ContentType="application/vnd.openxmlformats-package.relationships+xml"/><Default Extension="xml" ContentType="application/xml"/><Default Extension="png" ContentType="image/png"/>'+f'<Override PartName="/{part}" ContentType="application/vnd.openxmlformats-officedocument.wordprocessingml.document.main+xml"/><Override PartName="/word/styles.xml" ContentType="application/vnd.openxmlformats-officedocument.wordprocessingml.styles+xml"/><Override PartName="/word/numbering.xml" ContentType="application/vnd.openxmlformats-officedocument.wordprocessingml.numbering+xml"/></Types>',

+ '_rels/.rels':f'<Relationships xmlns="{P}"><Relationship Id="root" Type="{R}/officeDocument" Target="{part}"/></Relationships>',

+ part:f'<w:document xmlns:w="{W}" xmlns:r="{R}" xmlns:wp="http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing" xmlns:a="http://schemas.openxmlformats.org/drawingml/2006/main" xmlns:pic="http://schemas.openxmlformats.org/drawingml/2006/picture"><w:body>{body}<w:sectPr/></w:body></w:document>',

+ 'word/_rels/'+Path(part).name+'.rels':f'<Relationships xmlns="{P}">'+''.join(f'<Relationship Id="{i}" Type="{R}/{kind}" Target="{target}"{mode}/>' for i,kind,target,mode in rels)+'</Relationships>',

+ 'word/styles.xml':styles,'word/numbering.xml':numbering,

+ }

+ if name=='image': files['word/media/pixel.png']=png

+ with zipfile.ZipFile(ROOT/(name+'.docx'),'w',compression=zipfile.ZIP_STORED) as archive:

+ for path,data in sorted(files.items()):

+ info=zipfile.ZipInfo(path,date_time=(2026,1,1,0,0,0)); info.external_attr=0o100644<<16

+ archive.writestr(info,data)

+ manifest.append({'file':name+'.docx','body_part':part})

+(ROOT/'manifest.json').write_text(json.dumps(manifest,indent=2)+'\n')

diff --git a/fixtures/docx/image.docx b/fixtures/docx/image.docx

new file mode 100644

index 0000000..7c42904

Binary files /dev/null and b/fixtures/docx/image.docx differ

diff --git a/fixtures/docx/link.docx b/fixtures/docx/link.docx

new file mode 100644

index 0000000..00dd7dc

Binary files /dev/null and b/fixtures/docx/link.docx differ

diff --git a/fixtures/docx/manifest.json b/fixtures/docx/manifest.json

new file mode 100644

index 0000000..53f4c42

--- /dev/null

+++ b/fixtures/docx/manifest.json

@@ -0,0 +1,26 @@

+[

+ {

+ "file": "core.docx",

+ "body_part": "word/document.xml"

+ },

+ {

+ "file": "alternate-body.docx",

+ "body_part": "word/document2.xml"

+ },

+ {

+ "file": "numbered-heading.docx",

+ "body_part": "word/document.xml"

+ },

+ {

+ "file": "link.docx",

+ "body_part": "word/document.xml"

+ },

+ {

+ "file": "table.docx",

+ "body_part": "word/document.xml"

+ },

+ {

+ "file": "image.docx",

+ "body_part": "word/document.xml"

+ }

+]

diff --git a/fixtures/docx/numbered-heading.docx b/fixtures/docx/numbered-heading.docx

new file mode 100644

index 0000000..fbd0893

Binary files /dev/null and b/fixtures/docx/numbered-heading.docx differ

diff --git a/fixtures/docx/table.docx b/fixtures/docx/table.docx

new file mode 100644

index 0000000..f05cc7f

Binary files /dev/null and b/fixtures/docx/table.docx differ

diff --git a/internal/docs/docx_projection_test.go b/internal/docs/docx_projection_test.go

new file mode 100644

index 0000000..21c2611

--- /dev/null

+++ b/internal/docs/docx_projection_test.go

@@ -0,0 +1,49 @@

+package docs

+

+import (

+ "bytes"

+ "context"

+ "net/http"

+ "os"

+ "strings"

+ "testing"

+)

+

+func TestDOCXCoreProjectionThroughStoredYjsAndExportRoutes(t *testing.T) {

+ app, err := NewForTest(t)

+ if err != nil {

+ t.Fatal(err)

+ }

+ owner := app.TestPerson(t, "docx-fixture@example.test")

+ doc, err := app.Store().Create(context.Background(), owner, "DOCX fixture")

+ if err != nil {

+ t.Fatal(err)

+ }

+ update := readFixture(t, "../../fixtures/docx/core.update.bin")

+ if _, err := app.Store().ApplyUpdate(context.Background(), doc, app.TestActorFor(t, doc), update); err != nil {

+ t.Fatal(err)

+ }

+ expected, err := os.ReadFile("../../fixtures/docx/core.expected.md")

+ if err != nil {

+ t.Fatal(err)

+ }

+ markdown := serve(app, app.TestRequest(t, "docx-fixture@example.test", http.MethodGet, "/d/"+doc.Sqid+"/export.md", nil))

+ if markdown.Code != 200 {

+ t.Fatalf("Markdown route status %d", markdown.Code)

+ }

+ want := "# DOCX fixture\n\n" + string(expected)

+ if markdown.Body.String() != want {

+ t.Fatalf("persisted Yjs projection changed structure\ngot %q\nwant %q", markdown.Body.String(), want)

+ }

+ pdf := serve(app, app.TestRequest(t, "docx-fixture@example.test", http.MethodGet, "/d/"+doc.Sqid+"/export.pdf", nil))

+ if pdf.Code != 200 || !bytes.HasPrefix(pdf.Body.Bytes(), []byte("%PDF-")) {

+ t.Fatal("PDF route did not emit a PDF")

+ }

+ if !strings.HasPrefix(pdf.Header().Get("Content-Type"), "application/pdf") {

+ t.Fatal("PDF content type missing")

+ }

+ denied := serve(app, app.TestRequest(t, "foreign@example.test", http.MethodGet, "/d/"+doc.Sqid+"/export.md", nil))

+ if denied.Code != 404 {

+ t.Fatal("corpus document escaped owner boundary")

+ }

+}

diff --git a/internal/export/docx_corpus_test.go b/internal/export/docx_corpus_test.go

new file mode 100644

index 0000000..47ff9ed

--- /dev/null

+++ b/internal/export/docx_corpus_test.go

@@ -0,0 +1,192 @@

+package export_test

+

+import (

+ "archive/zip"

+ "bytes"

+ "encoding/json"

+ "encoding/xml"

+ "image/png"

+ "io"

+ "os"

+ "path"

+ "strings"

+ "testing"

+

+ "github.com/carlosframework/docs/internal/export"

+)

+

+const corpusDirectory = "../../fixtures/docx/"

+

+func corpusFile(t *testing.T, name string) []byte {

+ t.Helper()

+ data, err := os.ReadFile(corpusDirectory + name)

+ if err != nil {

+ t.Fatal(err)

+ }

+ return data

+}

+

+func docxParts(t *testing.T, name string) map[string][]byte {

+ t.Helper()

+ raw := corpusFile(t, name)

+ archive, err := zip.NewReader(bytes.NewReader(raw), int64(len(raw)))

+ if err != nil {

+ t.Fatal(err)

+ }

+ parts := map[string][]byte{}

+ for _, file := range archive.File {

+ if path.Clean(file.Name) != file.Name || strings.HasPrefix(file.Name, "/") || strings.HasPrefix(file.Name, "../") {

+ t.Fatal("fixture contains unsafe OPC part")

+ }

+ if _, exists := parts[file.Name]; exists {

+ t.Fatal("duplicate OPC part")

+ }

+ reader, err := file.Open()

+ if err != nil {

+ t.Fatal(err)

+ }

+ data, err := io.ReadAll(reader)

+ reader.Close()

+ if err != nil {

+ t.Fatal(err)

+ }

+ parts[file.Name] = data

+ if strings.HasSuffix(file.Name, ".xml") || strings.HasSuffix(file.Name, ".rels") {

+ decoder := xml.NewDecoder(bytes.NewReader(data))

+ for {

+ _, err := decoder.Token()

+ if err == io.EOF {

+ break

+ }

+ if err != nil {

+ t.Fatalf("malformed fixture XML %s", file.Name)

+ }

+ return parts

+}

+

+func TestDOCXCorpusPackageRelationshipsAndFeatureExpectations(t *testing.T) {

+ var manifest []struct {

+ File string `json:"file"`

+ BodyPart string `json:"body_part"`

+ }

+ if err := json.Unmarshal(corpusFile(t, "manifest.json"), &manifest); err != nil {

+ t.Fatal(err)

+ }

+ if len(manifest) != 6 {

+ t.Fatal("fixture corpus changed without expectation review")

+ }

+ for _, item := range manifest {

+ t.Run(item.File, func(t *testing.T) {

+ parts := docxParts(t, item.File)

+ var roots struct {

+ Relationships []struct {

+ Type string `xml:"Type,attr"`

+ Target string `xml:"Target,attr"`

+ } `xml:"Relationship"`

+ }

+ if err := xml.Unmarshal(parts["_rels/.rels"], &roots); err != nil {

+ t.Fatal(err)

+ }

+ body := ""

+ for _, rel := range roots.Relationships {

+ if strings.HasSuffix(rel.Type, "/officeDocument") {

+ body = rel.Target

+ }

+ if body != item.BodyPart || len(parts[body]) == 0 {

+ t.Fatal("officeDocument relationship does not select expected body")

+ }

+ // XML attributes need explicit tags; these independent reads validate the

+ // archived wire shape, not the Python fixture generator's in-memory data.

+ var rels struct {

+ Relationships []struct {

+ ID string `xml:"Id,attr"`

+ Target string `xml:"Target,attr"`

+ Mode string `xml:"TargetMode,attr"`

+ } `xml:"Relationship"`

+ }

+ relPart := path.Join(path.Dir(body), "_rels", path.Base(body)+".rels")

+ if err := xml.Unmarshal(parts[relPart], &rels); err != nil {

+ t.Fatal(err)

+ }

+ for _, rel := range rels.Relationships {

+ if rel.Mode != "External" && len(parts[path.Join(path.Dir(body), rel.Target)]) == 0 {

+ t.Fatal("internal relationship target missing")

+ }

+ content := string(parts[body])

+ switch item.File {

+ case "core.docx", "alternate-body.docx":

+ for _, text := range []string{"Section", "Plain ", "bold", "italic", "Outer", "Inner", "First", "Second", `<w:ilvl w:val="1"/>`} {

+ if !strings.Contains(content, text) {

+ t.Fatalf("core source lacks %q", text)

+ }

+ if item.File == "alternate-body.docx" && parts["word/document.xml"] != nil {

+ t.Fatal("alternate-body fixture accidentally permits fixed-path parsing")

+ }

+ case "numbered-heading.docx":

+ if !strings.Contains(content, `w:pStyle w:val="Heading1"`) || !strings.Contains(content, "w:numPr") || !strings.Contains(content, "Numbered heading") {

+ t.Fatal("numbered heading no longer combines both properties")

+ }

+ case "link.docx":

+ found := false

+ for _, rel := range rels.Relationships {

+ if rel.ID == "rLink" && rel.Mode == "External" && rel.Target == "https://example.test/reference" {

+ found = true

+ }

+ if !found || !strings.Contains(content, `w:hyperlink r:id="rLink"`) || !strings.Contains(content, "Reference") {

+ t.Fatal("link source or relationship missing")

+ }

+ case "table.docx":

+ if strings.Count(content, "<w:tr>") != 2 || strings.Count(content, "<w:tc>") != 4 {

+ t.Fatal("table geometry changed")

+ }

+ for _, text := range []string{"Cell A", "Cell B", "Cell C", "Cell D"} {

+ if !strings.Contains(content, text) {

+ t.Fatal("table cell missing")

+ }

+ case "image.docx":

+ config, err := png.DecodeConfig(bytes.NewReader(parts["word/media/pixel.png"]))

+ if err != nil || config.Width != 1 || config.Height != 1 {

+ t.Fatal("synthetic embedded image invalid")

+ }

+ if !strings.Contains(content, `r:embed="rImage"`) || !strings.Contains(content, `descr="One green pixel"`) {

+ t.Fatal("image reference or alt text missing")

+ }

+ })

+ }

+}

+

+func TestDOCXCoreExpectedProjectionSurvivesExistingExports(t *testing.T) {

+ doc, err := export.Parse("", string(corpusFile(t, "core.expected.json")))

+ if err != nil {

+ t.Fatal(err)

+ }

+ if got, want := export.Markdown(doc), string(corpusFile(t, "core.expected.md")); got != want {

+ t.Fatalf("structural Markdown projection mismatch\ngot: %q\nwant: %q", got, want)

+ }

+ data, report, err := export.PDF(doc)

+ if err != nil {

+ t.Fatal(err)

+ }

+ if !report.Empty() {

+ t.Fatal("unexpected core projection substitution")

+ }

+ pdf := readPDF(t, data)

+ extracted := ""

+ for _, page := range pdf.pages(t) {

+ extracted += pdf.text(t, page)

+ }

+ for _, text := range []string{"Section", "Plain", "bold", "italic", "Outer", "Inner", "First", "Second"} {

+ if !strings.Contains(extracted, text) {

+ t.Fatalf("PDF projection lost %q", text)

+ }

+}

oficina / docs Public

Add synthetic DOCX fidelity corpus and stored export projections