ENH: Open Document Format ODS support in read_excel() by davidovitch · Pull Request #9070 · pandas-dev/pandas (original) (raw)

@jreback thanks for your patience. Too bad I couldn't make it for the 0.17 release, but at least I tried and learned something in the process :-)

Before squashing everything into a single commit, what I think makes sense to do is the following:

• I create a new PR that contains the refactoring of the writer tests in test_excel.py. This PR would also contain the ods test files, and the renamed MS Excel files (xls, xlsx, xlsm). I've renamed them so it is more consistent, and corresponds to the ods files. This refactoring was necessary to ease the processes of supporting multiple reader mechanisms. Regardless of the outcome of the actual implementation of the new ods reader, I think this work got to an acceptable level, and can be merged independent from the actual new ODS reader. Not that is such a significant contribution, but it would otherwise get kind of lost (or create a more laborious merge later if other changes have appeared in the mean time). I would suggest to take this one up into the 0.17 release.
• I close this PR and create a new PR that only contains the changes to the reader in a single commit. In that way this messy PR and its history is kept for "the record". In the description of the new PR I summarize with what I struggled, and some of the key comments of the reviewers, and refer to the old PR.
• I think it would be great if someone could step in to fix this faster and better then I can, but until that happens, I hope it is ok if I try to find away to figure out how to do it more lean, clean and Pythonic (but I will not bother the reviewers again until I really think the results have improved significantly).

If you think this it too much whatever, please say so and I will just squash into one commit and forgot about the whole thing, no hard feelings :-)

@andresmrm, if you really want this badly, you can always give this fork a spin. The current implementation actually works, and passes all the tests (if you have a lot of ODS files hanging around you can test it more extensively and report back any bugs in this PR ;-) ). However, the implementation is not on par with the code quality everyone expects from a library as Pandas.