zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-06-29 15:31:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix: docx parsing raises ValueError on 'Heading' styles (#16284)

Browse Source
This commit is contained in:
minion1227
2026-06-23 22:16:16 -07:00
committed by GitHub
parent 0c19190daf
commit 14565b289a
2 changed files with 94 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
rag/nlp/__init__.py
View File

@@ -1201,7 +1201,12 @@ def naive_merge_with_images(texts, images, chunk_token_num=128, delimiter="\n。
def docx_question_level(p, bull=-1):
txt = re.sub(r"\u3000", " ", p.text).strip()
if hasattr(p.style, 'name') and p.style.name and p.style.name.startswith('Heading'):
return int(p.style.name.split(' ')[-1]), txt
# Heading styles are usually "Heading N", but the base "Heading" style,
# custom "Heading"-prefixed styles, or "HeadingN" (no space) have no
# space-separated trailing integer. Extract the level digits safely and
# fall back to the top heading level instead of raising ValueError (#16163).
m = re.search(r"\d+", p.style.name)
return (int(m.group()) if m else 1), txt
else:
if bull < 0:
return 0, txt