Skip to content

Python tool for converting files and office documents to Markdown.

Fix Zip converter forwarding kwargs to nested conversions (#2409)

SpongeBob0504d7d50f37
diff --git a/packages/markitdown/src/markitdown/converters/_zip_converter.py b/packages/markitdown/src/markitdown/converters/_zip_converter.py
index 3d388a78..554b59c8 100644
--- a/packages/markitdown/src/markitdown/converters/_zip_converter.py
+++ b/packages/markitdown/src/markitdown/converters/_zip_converter.py
@@ -98,6 +98,13 @@ class ZipConverter(DocumentConverter):
)
md_content = f"Content from the zip file `{file_path}`:\n\n"
+ # Drop the metadata describing the outer *archive*
+ member_kwargs = {
+ key: value
+ for key, value in kwargs.items()
+ if key not in ("file_extension", "url")
+ }
+
with zipfile.ZipFile(file_stream, "r") as zipObj:
for name in zipObj.namelist():
try:
@@ -109,6 +116,7 @@ class ZipConverter(DocumentConverter):
result = self._markitdown.convert_stream(
stream=z_file_stream,
stream_info=z_file_stream_info,
+ **member_kwargs,
)
if result is not None:
md_content += f"## File: {name}\n\n"
diff --git a/packages/markitdown/tests/test_zip_kwargs.py b/packages/markitdown/tests/test_zip_kwargs.py
new file mode 100644
index 00000000..a1470b92
--- /dev/null
+++ b/packages/markitdown/tests/test_zip_kwargs.py
@@ -0,0 +1,136 @@
+import io
+import os
+import zipfile
+from typing import Any, BinaryIO, Optional
+
+import pytest
+
+from markitdown import (
+ DocumentConverter,
+ DocumentConverterResult,
+ MarkItDown,
+ StreamInfo,
+)
+from markitdown.converters._zip_converter import ZipConverter
+
+TEST_FILES_DIR = os.path.join(os.path.dirname(__file__), "test_files")
+
+
+class _RecordingConverter(DocumentConverter):
+ def __init__(self):
+ super().__init__()
+ self.seen_kwargs = []
+
+ def accepts(
+ self,
+ file_stream: BinaryIO,
+ stream_info: StreamInfo,
+ **kwargs: Any,
+ ) -> bool:
+ return (stream_info.extension or "").lower() == ".txt"
+
+ def convert(
+ self,
+ file_stream: BinaryIO,
+ stream_info: StreamInfo,
+ **kwargs: Any,
+ ) -> DocumentConverterResult:
+ self.seen_kwargs.append(dict(kwargs))
+ return DocumentConverterResult(markdown=file_stream.read().decode("utf-8"))
+
+
+def test_zip_forwards_kwargs_to_nested_converters() -> None:
+ markitdown = MarkItDown(enable_builtins=False)
+ recorder = _RecordingConverter()
+ markitdown.register_converter(recorder)
+ markitdown.register_converter(ZipConverter(markitdown=markitdown))
+
+ buf = io.BytesIO()
+ with zipfile.ZipFile(buf, "w") as zf:
+ zf.writestr("hello.txt", "Hello world")
+ buf.seek(0)
+
+ result = markitdown.convert_stream(
+ buf,
+ stream_info=StreamInfo(mimetype="application/zip", extension=".zip"),
+ keep_data_uris=True,
+ )
+
+ assert "Hello world" in result.markdown
+ assert len(recorder.seen_kwargs) == 1
+ assert recorder.seen_kwargs[0].get("keep_data_uris") is True
+
+
+def test_zip_does_not_forward_outer_file_metadata() -> None:
+ """The archive's own extension/url must not override a member's."""
+ markitdown = MarkItDown(enable_builtins=False)
+ recorder = _RecordingConverter()
+ markitdown.register_converter(recorder)
+ markitdown.register_converter(ZipConverter(markitdown=markitdown))
+
+ buf = io.BytesIO()
+ with zipfile.ZipFile(buf, "w") as zf:
+ zf.writestr("hello.txt", "Hello world")
+ buf.seek(0)
+
+ result = markitdown.convert_stream(
+ buf,
+ stream_info=StreamInfo(
+ mimetype="application/zip",
+ extension=".zip",
+ url="https://example.com/archive.zip",
+ ),
+ )
+
+ assert "Hello world" in result.markdown
+ assert len(recorder.seen_kwargs) == 1
+ # The member sees its own extension, not the archive's
+ assert recorder.seen_kwargs[0].get("file_extension") == ".txt"
+ assert "url" not in recorder.seen_kwargs[0]
+
+
+def test_zip_member_reaches_its_own_converter() -> None:
+ """A ZIP-based member (docx) must not be unpacked as a nested archive."""
+ markitdown = MarkItDown()
+ result = markitdown.convert(
+ os.path.join(TEST_FILES_DIR, "test_files.zip"),
+ )
+
+ assert "## File: test.docx" in result.markdown
+ # Raw OOXML parts would appear if the docx were treated as a plain zip
+ assert "[Content_Types].xml" not in result.markdown
+
+
+ "archive_url",
+ [None, "https://example.test/archive.zip"],
+ ids=["no_archive_url", "with_archive_url"],
+)
+def test_zip_member_docx_converts_as_docx(archive_url: Optional[str]) -> None:
+ """A DOCX member converts via the DOCX converter, archive URL or not.
+
+ Regression: the archive's own file_extension/url used to be forwarded into
+ the nested convert_stream call, where they take precedence over the
+ member's StreamInfo -- so the DOCX was re-selected as a ZIP and unpacked
+ into its raw OOXML parts instead of being converted.
+ """
+ markitdown = MarkItDown()
+ docx_path = os.path.join(TEST_FILES_DIR, "test.docx")
+ standalone = markitdown.convert(docx_path).markdown
+
+ buf = io.BytesIO()
+ with zipfile.ZipFile(buf, "w") as zf:
+ zf.write(docx_path, "test.docx")
+ buf.seek(0)
+
+ result = markitdown.convert_stream(
+ buf,
+ stream_info=StreamInfo(extension=".zip", url=archive_url),
+ ).markdown
+
+ assert "## File: test.docx" in result
+ # The member's real conversion output is present ...
+ assert standalone.strip() in result
+ # ... and it was not unpacked as a nested archive
+ assert "word/document.xml" not in result
+ assert "[Content_Types].xml" not in result