mirror of
https://github.com/Unstructured-IO/unstructured.git
synced 2025-10-03 04:14:15 +00:00
e6637592d1
**Summary.** The `python-docx` table API is designed for _uniform_
tables (no merged cells, no nested tables). Naive processing of DOCX
tables using this API produces duplicate text when the table has merged
cells. Add a more sophisticated parsing method that reads only "root"
cells (those with an actual `<tc>` element) and skip cells spanned by a
merge.
In the process, abandon use of the `tabulate` package for this job
(which is also designed for uniform tables) and remove the whitespace
padding it adds for visual alignment of columns. Separate the text for
each cell with a single newline ("\n").
Since it's little extra trouble, add support for nested tables such that
their text also contributes to the `Table.text` string.
The new `._iter_table_texts()` method will also be used for parsing
tables in headers and footers (where they are frequently used for layout
purposes) in a closely following PR.
Fixes #2106.
17 lines
325 B
Python
17 lines
325 B
Python
"""Table-related XML element-types."""
|
|
|
|
from __future__ import annotations
|
|
|
|
from typing import List
|
|
|
|
from docx.oxml.xmlchemy import BaseOxmlElement
|
|
|
|
class CT_Row(BaseOxmlElement):
|
|
tc_lst: List[CT_Tc]
|
|
|
|
class CT_Tc(BaseOxmlElement):
|
|
@property
|
|
def vMerge(self) -> str | None: ...
|
|
|
|
class CT_Tbl(BaseOxmlElement): ...
|