A pipe in a DOCX table cell becomes a column separator in the extracted Markdown

Start at WordExtractor._table_to_markdown, using the DOCX table reproduction in the issue to inspect how cell content is joined into Markdown rows. Compare the handling with core/rag/extractor/excel_extractor.py, then verify that the rendered table preserves the pipe as cell content and that the accompanying tests pass.

Self Checks

• I have read the Contributing Guide and Language Policy.
• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report, otherwise it will be closed.
• 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
• Please do not modify this template :) and fill in all the required fields.

Dify version

main (38f9d85d5a)

Cloud or Self Hosted

Self Hosted (Source)

Steps to reproduce

Add a .docx to a knowledge base that contains a table where a cell holds a | character. A pipe is ordinary content in a table cell: a regular expression (a|b), an "and/or" shorthand, a shell pipeline in a documentation table, or P(A|B) in a statistics table.

Two-column table:

WordExtractor._table_to_markdown renders it by joining the cells:

markdown.append("| " + " | ".join(headers) + " |")
markdown.append("| " + " | ".join(["---"] * total_cols) + " |")
for row in table.rows[1:]:
    row_cells = self._parse_row(row, image_map, total_cols)
    markdown.append("| " + " | ".join(row_cells) + " |")

Reproduced directly against the renderer:

>>> extractor = WordExtractor.__new__(WordExtractor)
>>> print(extractor._table_to_markdown(Document("t.docx").tables[0], {}))
| Name | Pattern |
| --- | --- |
| alternation | a|b |

✔️ Expected Behavior

The pipe inside the cell is escaped, so the row describes the same two columns as the header:

| Name | Pattern |
| --- | --- |
| alternation | a\|b |

❌ Actual Behavior

The cell's pipe is emitted as a column separator, so the last row claims three columns where the header declares two. From that row on, the table is misaligned: a renderer that trims to the header width drops the trailing cell, and a model reading the chunk sees a and b as two separate columns rather than one value.

Nothing fails and nothing is logged — the output is still a syntactically valid Markdown table, just not the table the document contains.

core/rag/extractor/excel_extractor.py already escapes the structural character of the format it emits — value.replace('"', '\\"'), in three places, so a quote inside a cell cannot corrupt the row — which is what makes the Markdown table renderer look like an oversight rather than a decision.

I have a fix and tests ready and will open a PR referencing this issue.

Transparency: this report was prepared with AI assistance. The reproduction above was run against the extractor at 38f9d85d5a and the output is copied verbatim from that run.