openpyxl: Workbook opened in readonly mode should return `ReadOnlyCell | EmptyCell` from `worksheet.rows`

[jamesbeith]

Currently the type hints for worksheet.rows returns Cell objects. However, opening a workbook in readonly mode results in ReadOnlyCell or EmptyCell objects being returned.

>>> workbook = openpyxl_excel_reader.load_workbook(file, read_only=True)
>>> worksheet = workbook.worksheets[0]  # openpyxl.worksheet._read_only.ReadOnlyWorksheet
>>> [row for row in worksheet.rows]
[
    (<ReadOnlyCell 'Sheet1'.A1>, <EmptyCell>, <ReadOnlyCell 'Sheet1'.C1>),
    (<ReadOnlyCell 'Sheet1'.A2>, <EmptyCell>, <ReadOnlyCell 'Sheet1'.C2>),
]

[sobolevn]

The problem is that .worksheets always return Worksheet and does not ever return _ReadOnlyWorksheet: https://github.com/python/typeshed/blob/main/stubs/openpyxl/openpyxl/workbook/workbook.pyi#L61

And load_workbook always returns Workbook

typeshed/stubs/openpyxl/openpyxl/reader/excel.pyi

Line 50 in a544b75

) -> Workbook: ...

So, we don't have enough typing information right now to make this work.

Ideas?

[jamesbeith]

Could we use @overload and Literal for the load_workbook() function, where if read_only is the literal True then we instead return some kind of new ReadOnlyWorkbook type which itself returns _ReadOnlyWorksheet from .worksheets (and then ReadOnlyCell | EmptyCell for it's rows). Is that a feasible idea?

https://mypy.readthedocs.io/en/stable/literal_types.html#literal-types

[sobolevn]

Yes, seems ok to me :)
Do you want to work on this?

[jamesbeith]

Do you want to work on this?

@sobolevn I'll aim to take a look 👍🏼

[Avasam]

Do note that this problem goes one step further: ReadOnlyWorksheet cell methods come from Worksheet, so you'd have to specialize ReadOnlyWorksheet's cell-related method types on top of making Workbook generic.