maartenbreddels commented on pull request #7701: URL: https://github.com/apache/arrow/pull/7701#issuecomment-656785094
I'm mainly looking at my own code, especially the unittests. They are littered with .tolist, similar to how this project has a lot of .to_pylist. For very basic operations, you'd expect similar outputs, so that would mean many tests (and code) can stay the same. One could argue the same for .item() and .to_py for scalars, but comparing scalars is less common, and some effort to support two libraries is ok. I'm happy to see that both libraries have a .take method, but only Arrow has .filter. I'm ok with that, again, it's not that common. Overloading pyarrow's getitem to act as .filter would make them compatible, but putting on my purist hat, I'd rather vote for numpy to adopt a .filter instead. I'd refrain from ambiguity, so no overloading, an alias like this is... not pretty, but I doubt it will affect readability or maintainability. I could change the docstring not to stimulate more of these ideas :). ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
