#!/usr/bin/env python3 """Build and inspect a synthetic PDF accessibility example, not a PDF/UA template. Reproduce: python -m pip install reportlab==4.4.9 pypdf==6.10.0 python build-fixture.py No network requests, customer files, or translation API calls are made. """ from io import BytesIO import json from pathlib import Path import re import pypdf import reportlab from pypdf import PdfReader, PdfWriter from pypdf.generic import ( ArrayObject, BooleanObject, ContentStream, DictionaryObject, NameObject, NumberObject, TextStringObject, ) from reportlab.lib.colors import HexColor from reportlab.pdfgen import canvas OUTPUT = Path(__file__).resolve().parent BLOCKS = [ (0, "H1", 48, 744, 21, "Preparar un archivo para traducir"), (1, "P", 48, 713, 10, "Ejemplo sintético: contenido inventado para inspección técnica."), (2, "H2", 48, 655, 15, "1. Revisar el archivo"), (3, "P", 48, 619, 12, "A1. Confirme que el archivo\ncontiene todas las páginas."), (4, "P", 48, 559, 12, "A2. Guarde una copia del\noriginal antes de editarlo."), (5, "H2", 326, 655, 15, "2. Revisar la entrega"), (6, "P", 326, 619, 12, "B1. Compare los títulos\ncon el archivo original."), (7, "P", 326, 559, 12, "B2. Revise las notas antes\nde compartir el resultado."), (8, "P", 48, 476, 12, "Etiqueta conservada de la interfaz:"), (9, "Span", 48, 453, 12, "Read only"), (10, "Figure", 48, 380, 12, ""), (11, "P", 48, 255, 10, "Los dos archivos tienen el mismo aspecto visual."), (12, "P", 48, 235, 10, "Este ejemplo no demuestra conformidad con PDF/UA."), ] FIXED_ORDER = list(range(len(BLOCKS))) BROKEN_ORDER = [0, 1, 2, 5, 3, 6, 4, 7, 8, 9, 10, 11, 12] ALT = { "before": "A source folder points to a translated document.", "after": "Una carpeta de origen apunta a un documento traducido.", } def visible_page(): stream = BytesIO() pdf = canvas.Canvas(stream, pagesize=(612, 792), invariant=1, pageCompression=0) pdf.setTitle("Ejemplo de accesibilidad de PDF traducido") pdf.setAuthor("BookTranslator Team") pdf.addLiteral("/Artifact BMC") pdf.setFillColor(HexColor("#F6F3EC")) pdf.rect(0, 0, 612, 792, fill=1, stroke=0) pdf.setStrokeColor(HexColor("#B8B0A0")) pdf.line(302, 535, 302, 675) pdf.line(48, 692, 564, 692) pdf.addLiteral("EMC") for mcid, role, x, y, size, value in BLOCKS: pdf.addLiteral(f"/{role} <> BDC") pdf.setFillColor(HexColor("#232C31")) if role == "Figure": pdf.setStrokeColor(HexColor("#23665D")) pdf.setLineWidth(2) pdf.roundRect(55, 316, 99, 56, 4, fill=0, stroke=1) pdf.line(55, 372, 75, 390) pdf.line(75, 390, 112, 390) pdf.line(112, 390, 129, 372) pdf.line(174, 344, 250, 344) pdf.line(239, 352, 250, 344) pdf.line(239, 336, 250, 344) pdf.rect(278, 306, 53, 80, fill=0, stroke=1) for line_y in [369, 355, 341, 327]: pdf.line(288, line_y, 321, line_y) else: text = pdf.beginText(x, y) text.setFont("Helvetica-Bold" if role.startswith("H") else "Helvetica", size) text.setLeading(size * 1.45) for line in value.splitlines(): text.textLine(line) pdf.drawText(text) pdf.addLiteral("EMC") pdf.showPage() pdf.save() return stream.getvalue() def tagged_version(raw, variant): writer = PdfWriter() writer.clone_document_from_reader(PdfReader(BytesIO(raw))) root = writer.root_object page = writer.pages[0] root[NameObject("/Lang")] = TextStringObject("es-ES" if variant == "after" else "en-US") root[NameObject("/MarkInfo")] = DictionaryObject({NameObject("/Marked"): BooleanObject(True)}) root[NameObject("/ViewerPreferences")] = DictionaryObject({NameObject("/DisplayDocTitle"): BooleanObject(True)}) page[NameObject("/StructParents")] = NumberObject(0) tree = DictionaryObject({NameObject("/Type"): NameObject("/StructTreeRoot")}) tree_ref = writer._add_object(tree) document = DictionaryObject({ NameObject("/Type"): NameObject("/StructElem"), NameObject("/S"): NameObject("/Document"), NameObject("/P"): tree_ref, }) document_ref = writer._add_object(document) elements = [] for mcid, role, *_ in BLOCKS: element = DictionaryObject({ NameObject("/Type"): NameObject("/StructElem"), NameObject("/S"): NameObject("/" + role), NameObject("/P"): document_ref, NameObject("/Pg"): page.indirect_reference, NameObject("/K"): NumberObject(mcid), }) if role == "Figure": element[NameObject("/Alt")] = TextStringObject(ALT[variant]) if mcid == 9 and variant == "after": element[NameObject("/Lang")] = TextStringObject("en-US") elements.append(writer._add_object(element)) order = FIXED_ORDER if variant == "after" else BROKEN_ORDER document[NameObject("/K")] = ArrayObject([elements[index] for index in order]) tree[NameObject("/K")] = document_ref tree[NameObject("/ParentTree")] = writer._add_object(DictionaryObject({ NameObject("/Nums"): ArrayObject([NumberObject(0), ArrayObject(elements)]) })) tree[NameObject("/ParentTreeNextKey")] = NumberObject(1) root[NameObject("/StructTreeRoot")] = tree_ref destination = OUTPUT / f"translated-pdf-{variant}.pdf" with destination.open("wb") as handle: writer.write(handle) return destination def inspect(path): reader = PdfReader(path) page = reader.pages[0] root = reader.trailer["/Root"] text_by_id = {} stack = [] for operands, operator in ContentStream(page["/Contents"], reader).operations: if operator in (b"BMC", b"BDC"): properties = operands[1] if operator == b"BDC" else {} stack.append(properties.get("/MCID")) elif operator == b"EMC": stack.pop() elif operator == b"Tj" and stack and stack[-1] is not None: text_by_id.setdefault(int(stack[-1]), []).append(str(operands[0])) assert not stack, "Unbalanced marked content" tree = root["/StructTreeRoot"] elements = tree["/K"]["/K"] parent_map = tree["/ParentTree"]["/Nums"][1] rows = [] for reference in elements: element = reference.get_object() mcid = int(element["/K"]) assert parent_map[mcid] == reference, "MCID parent-tree mismatch" rows.append({ "mcid": mcid, "role": str(element["/S"]), "effective_language": str(element.get("/Lang", root["/Lang"])), "text": re.sub(r"\s+", " ", " ".join(text_by_id.get(mcid, []))), "alt": str(element.get("/Alt", "")), }) assert sorted(row["mcid"] for row in rows) == FIXED_ORDER return {"file": path.name, "document_language": str(root["/Lang"]), "tag_order": rows} def main(): raw = visible_page() reports = [inspect(tagged_version(raw, version)) for version in ("before", "after")] before, after = reports assert [row["mcid"] for row in before["tag_order"]] == BROKEN_ORDER assert [row["mcid"] for row in after["tag_order"]] == FIXED_ORDER assert before["document_language"] == "en-US" assert after["document_language"] == "es-ES" assert after["tag_order"][9]["effective_language"] == "en-US" assert before["tag_order"][10]["alt"] == ALT["before"] assert after["tag_order"][10]["alt"] == ALT["after"] before_reader = PdfReader(OUTPUT / before["file"]) after_reader = PdfReader(OUTPUT / after["file"]) assert before_reader.pages[0]["/Contents"].get_data() == after_reader.pages[0]["/Contents"].get_data() result = { "created": "2026-09-07", "provenance": "Synthetic Spanish example authored for this article; not translation service output.", "method": "Read PDF catalog, structure tree, MCID parent tree and marked-content text with pypdf.", "versions": {"reportlab": reportlab.Version, "pypdf": pypdf.__version__}, "limitations": ["Not a screen-reader execution or audio transcript.", "Not a PDF/UA or WCAG conformance audit.", "No interactive controls or tables in this fixture."], "checks": {"page_content_streams_identical": True, "mcid_parent_tree_mapping": True, "expected_tag_order": True, "expected_language_and_alt": True}, "files": reports, } (OUTPUT / "inspection.json").write_text(json.dumps(result, ensure_ascii=False, indent=2) + "\n", encoding="utf-8") print(json.dumps({"outputs": [report["file"] for report in reports], "checks": result["checks"]}, indent=2)) if __name__ == "__main__": main()