Inconsistent behavior of "set_column" formatting in constant memory mode

We are using xlsxwriter (verison 3.0.3 on Python 3.8.14) to create and style excel files. When using constant memory mode, we found inconsistent behavior of how the set_column method applies cell formatting. Here is an example to demonstrate the issue:

import xlsxwriter

with open("test.xlsx", "wb") as fs:
    workbook = xlsxwriter.Workbook(fs, {"constant_memory": True})
    worksheet = workbook.add_worksheet("Test")

    worksheet.write_row(0, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])
    worksheet.write_row(1, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])
    worksheet.write_row(2, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])

    worksheet.set_column(0, 3, cell_format=workbook.add_format({"num_format": "#,##0.00"}))

    workbook.close()

The resulting excel file looks like this. You can see that the number format was only applied to the last row:

If I set {"constant_memory": False}, the file looks like expected:

The same (correct) result can also be achieved by moving the styling above writing the data:

import xlsxwriter

with open("test.xlsx", "wb") as fs:
    workbook = xlsxwriter.Workbook(fs, {"constant_memory": True})
    worksheet = workbook.add_worksheet("Test")

    worksheet.set_column(0, 3, cell_format=workbook.add_format({"num_format": "#,##0.00"}))

    worksheet.write_row(0, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])
    worksheet.write_row(1, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])
    worksheet.write_row(2, 0, data=[100000.123, 100000.123, 100000.123, 100000.123])

    workbook.close()

[Birne94]

Thank you @jmcnamara for the quick response.

Do you know why only the last column gets the correct style in this case?

I think this limitation should be worth mentioning in the documentation, especially since a lot of examples we found were following the pattern of writing data first and then applying the styling. Alternatively, maybe the library could also raise an exception instead, what do you think?

[jmcnamara]

Another interesting finding is that when looking at the generated excel files, we found the styles were correctly applied to the column definitions in the XML and were just missing in the cell definition:

That is for the reason explained above: the library writes the row to disk once a new row is initiated and if there isn’t a column or row format available at that time then the format isn’t applied.

This shouldn’t be an issue if you set the column formats first like in your first example.

The cell styling that works should look like this:

I am 100% aware of that. 😃