fix(docreader): detect legacy doc payloads

This commit is contained in:
wizardchen
2026-07-11 12:03:43 +08:00
committed by lyingbug
parent ac8c5c5540
commit 083e9738f6
2 changed files with 85 additions and 6 deletions
+27 -6
View File
@@ -8,6 +8,28 @@ from docreader.parser.web_parser import WebParser
logger = logging.getLogger(__name__)
# OLE Compound File magic used by legacy binary Microsoft Office files.
# Some WPS/Word documents keep this payload while being renamed to .docx.
_OLE_COMPOUND_FILE_MAGIC = b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1"
def detect_effective_file_type(file_type: str, content: bytes) -> str:
"""Return the parser file type after checking trustworthy file magic.
OOXML ``.docx`` files are ZIP containers, while legacy ``.doc`` files
use the OLE Compound File format. Word and WPS tolerate a legacy file
renamed to ``.docx``, so route that well-known mismatch through the DOC
parser instead of feeding binary OLE data to the DOCX parser.
"""
normalized = file_type.lower().lstrip(".")
if normalized == "docx" and content.startswith(_OLE_COMPOUND_FILE_MAGIC):
logger.warning(
"Detected legacy DOC content with a DOCX extension; using DOC parser"
)
return "doc"
return normalized
class Parser:
"""Document parser facade (lightweight version).
@@ -40,15 +62,16 @@ class Parser:
engine or "builtin",
)
cls = self.registry.get_parser_class(engine, file_type)
effective_file_type = detect_effective_file_type(file_type, content)
cls = self.registry.get_parser_class(engine, effective_file_type)
logger.info(
"Creating %s parser instance for %s file",
cls.__name__,
file_type,
effective_file_type,
)
parser = cls(
file_name=file_name,
file_type=file_type,
file_type=effective_file_type,
**overrides,
)
@@ -57,9 +80,7 @@ class Parser:
if not result.content:
logger.warning("Parser returned empty content for file: %s", file_name)
logger.info(
"Parsed file %s, content length=%d", file_name, len(result.content)
)
logger.info("Parsed file %s, content length=%d", file_name, len(result.content))
return result
def parse_url(
+58
View File
@@ -0,0 +1,58 @@
import unittest
from docreader.models.document import Document
from docreader.parser.base_parser import BaseParser
from docreader.parser.parser import Parser, detect_effective_file_type
OLE_MAGIC = b"\xd0\xcf\x11\xe0\xa1\xb1\x1a\xe1"
class _RecordingParser(BaseParser):
last_file_type = ""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
type(self).last_file_type = self.file_type
def parse_into_text(self, content: bytes) -> Document:
return Document(content="parsed")
class _RecordingRegistry:
def __init__(self):
self.requested_file_type = ""
def get_parser_class(self, engine: str, file_type: str):
self.requested_file_type = file_type
return _RecordingParser
class ParserRoutingTest(unittest.TestCase):
def test_legacy_doc_payload_renamed_to_docx_uses_doc_parser(self):
registry = _RecordingRegistry()
parser = Parser()
parser.registry = registry
result = parser.parse_file(
"legacy.docx",
"docx",
OLE_MAGIC + b"legacy-word-payload",
)
self.assertEqual("parsed", result.content)
self.assertEqual("doc", registry.requested_file_type)
self.assertEqual("doc", _RecordingParser.last_file_type)
def test_real_docx_payload_keeps_docx_parser_route(self):
self.assertEqual(
"docx",
detect_effective_file_type(".DOCX", b"PK\x03\x04ooxml-payload"),
)
def test_unrelated_ole_type_is_not_reclassified(self):
self.assertEqual("xls", detect_effective_file_type("xls", OLE_MAGIC))
if __name__ == "__main__":
unittest.main()