Deleting a column in a simple table breaks cell alignment

[Frolfast]

In python-docx, attempting to delete a column from a simple table (created via python-docx itself) causes cell misalignment or destruction of the table structure.

Environment:

• No merged cells
• No nested elements
• Uniform number of columns in all rows
• Tables created with docx.Document().add_table()

Steps to reproduce:

1. Create a 5-column table
2. Populate with plain text
3. Attempt to remove column 3 using:
   row._tr.remove(row.cells[2]._tc)

Result:

• Table becomes malformed
• Some rows have fewer cells
• In MS Word, rendering is broken

Expected:

• Clean removal of the column with remaining structure preserved

This seems to be a long-standing limitation, but it now affects even simple programmatically generated documents. Please consider addressing this bug or documenting a safe workaround.

[scanny]

@Frolfast tables in Word are a complex data structure and you're using lxml methods to modify that structure directly, so I'm afraid it's up to you to work out how to fix any problems you encounter with that.

We can only support use of published interface methods and there is no table.delete_column() method in the API.

I'm happy to give some guidance, but you'll have to provide enough for me to go on; I'm not going to try to reproduce the problem you're having and do my own analysis; you'll have to do that bit.

But from what you've mentioned, removing a single cell is not going to work. You'll at least need to remove all cells in the column, something like:

for row in table.rows:
    row._tr.remove(row.cells[2]._tc

You should also be looking for gridcol elements and attend to removing those as appropriate as well:
https://python-docx.readthedocs.io/en/latest/dev/analysis/features/table/index.html

If you're going to manipulate underlying XML you'll want to make friends with the underlying XML Schema and learn how to view the XML directly and see how Word does the operations you're trying to duplicate. The opc_diag tool can be handy for that:
https://opc-diag.readthedocs.io/en/latest/

I believe pip install opc-diag gets that installed for you.