gspread
gspread copied to clipboard
burnash
`combine_merged_cells` cannot capture all merges
Merges are actually just one value in the top left cell, and some metadata. If the top-left cell is not in your requested range, you cannot faithfully recreate the merge. For example:
If you called worksheet.get_values("F3:H4", combine_merged_cells=True), what would you expect?
1
| "big merge" | 2 | 3 |
|---|---|---|
| 4 | 5 | 6 |
or 2
| 2 | 3 | |
|---|---|---|
| 4 | 5 | 6 |
We cannot recreate 1 without added complexity. We must return 2. This may be unexpected.
combine_merged_cells was added in #1215. Its addition has brought several problems (#1298, #1330).
What should we do? If we leave it in, it will always have this issue. If we remove it, it will be sad.
Users will expect the value Big Merge in F3. We should add the extra complexity. When we list the merges and see that one overlaps the requested range then we should get and fill what is necessary.
For sure this will take some time and either add a very large util function or introduce a big code change in Worksheet.py
We should start working on it after 6.0.0 is out that would be a great improvement to bring after the release.
Thinking about it I would solve it as:
- we get values from a range
- if user requests combined merged values
- get the list of merges from the API
- check if any merge range overlaps the requested range in all boundaries
- in order to solve this, we might find some already optimized algorithm online 🤔
- we could rely on some third party library too (check some classic packages like bumpy etc )
- for any overlapping merge set the values in the "empty" cells.
A different solution that we can try is: open an issue on Google side and request some API changes or some extra parameters that does it all for us.
I think this is a very large change with lots to think about. combine_merged_cells has already caused two minor releases due to bugs (#1298 #1330). I am worried that making it even more complex is the wrong direction to go in. I think it is a small feature in the context of the size of gspread and not worth the effort to fix this issue.
I would rather add a warning that says it will not pull in data from outside the requested range, or as you say submit a request to Google's API.