Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Python ai4sciencedocument-analysisdocxextract-datalayout-analysisocrparserpdfpdf-converterpdf-extractor-llmpdf-extractor-pretrainpdf-extractor-ragpdf-parserpptxpythonxlsx
Validate EMF+ Only images through Office and render consumers
myhloli2d329bfb76b0
diff --git a/tests/unittest/fixtures/emfplus/fallback-only.emf b/tests/unittest/fixtures/emfplus/fallback-only.emfnew file mode 100644index 00000000..f7259d73Binary files /dev/null and b/tests/unittest/fixtures/emfplus/fallback-only.emf differdiff --git a/tests/unittest/fixtures/emfplus/geometry-only.emf b/tests/unittest/fixtures/emfplus/geometry-only.emfnew file mode 100644index 00000000..80d66d68Binary files /dev/null and b/tests/unittest/fixtures/emfplus/geometry-only.emf differdiff --git a/tests/unittest/test_emfplus_integration.py b/tests/unittest/test_emfplus_integration.pynew file mode 100644index 00000000..97ac4a51--- /dev/null+++ b/tests/unittest/test_emfplus_integration.py@@ -0,0 +1,64 @@+"""验证 metafile-render 0.2 的 Only 图片经 Office 解析与四种消费者输出。"""++from __future__ import annotations++from importlib.metadata import version+from io import BytesIO+from pathlib import Path+from unittest.mock import Mock+from zipfile import ZipFile++import pytest+from _office_image_mtef_test_utils import build_image_docx+from packaging.version import Version+from PIL import Image+from pypdf import PdfReader++from mineru.backend.analyze import doc_analyze+from mineru.model.flash.office import image as office_image+from mineru.render import render_docx, render_epub, render_html, render_pdf++pytestmark = pytest.mark.skipif(+ Version(version("metafile-render")) < Version("0.2.0"),+ reason="EMF+ Only feature validation requires metafile-render 0.2.0; 0.1 remains a valid base dependency",+)+_FIXTURES = Path(__file__).parent / "fixtures" / "emfplus"++[email protected]("scene", ["geometry", "fallback"])+def test_only_docx_flows_into_html_docx_pdf_epub(scene: str, monkeypatch: pytest.MonkeyPatch) -> None:+ """真实 Only 图片经 DOCX 解析进入全部消费端,安全 SVG 仍可提取 PNG fallback。"""+ monkeypatch.setattr(office_image, "is_windows_environment", lambda: False)+ package = build_image_docx((_FIXTURES / f"{scene}-only.emf").read_bytes())+ middle, _ = doc_analyze(package, effort="flash", image_analysis=False, file_suffix="docx")+ html = render_html(middle)+ assert "data:image/svg+xml;base64," in html+ for payload in (render_docx(middle), render_epub(middle)):+ with ZipFile(BytesIO(payload)) as archive:+ images = [archive.read(name) for name in archive.namelist() if name.endswith(".png")]+ assert images+ with Image.open(BytesIO(images[0])) as image:+ assert image.width > 100 and image.height > 100+ assert image.convert("RGB").getextrema() != ((255, 255), (255, 255), (255, 255))+ pdf = PdfReader(BytesIO(render_pdf(middle)))+ assert pdf.pages+ assert any(page.images for page in pdf.pages)+++def test_only_partial_reports_diagnostics_and_keeps_image(monkeypatch: pytest.MonkeyPatch) -> None:+ """样式近似的 Only 仍输出 SVG,并把部分结果诊断交给 Office 日志。"""+ warning = Mock()+ monkeypatch.setattr(office_image.logger, "warning", warning)+ monkeypatch.setattr(office_image, "is_windows_environment", lambda: False)+ result = office_image.serialize_office_image((_FIXTURES / "fallback-only.emf").read_bytes())+ assert result is not None and result.startswith("data:image/svg+xml;base64,")+ assert any(+ "Rendered partial EMF" in str(call) and "emfplus_brush_approximation" in str(call) for call in warning.call_args_list+ )+++def test_broken_only_uses_existing_placeholder(monkeypatch: pytest.MonkeyPatch) -> None:+ """签名有效但损坏的 Only 文件沿用原有失败兜底。"""+ monkeypatch.setattr(office_image, "is_windows_environment", lambda: False)+ data = (_FIXTURES / "geometry-only.emf").read_bytes()[:48]+ assert office_image.serialize_office_image(data) == office_image.get_standard_vector_placeholder_data_uri()