n3world commented on pull request #10794: URL: https://github.com/apache/arrow/pull/10794#issuecomment-886728438
> The problem with this approach is that yields column chunks of differing types. You can see this if you add the following test: Yes that will because the type is unknown, yet. This test seems artificial in that it doesn't follow how the column decoder is actually used. In use all empty record batches get discarded so their type don't actually matter. That is why it works for the csv streaming test I modified to have multiple empty blocks before a block with data. That test would not work without this change so the change does not make anything better or worse as far as that test goes. I would argue it makes things better because eventually you can get data from the converter. The only ways I can think to get that test to work would be to allow the API not to have to return a result until the type is known but that is dependent on the read ahead, for csv, to be high enough to actually find data. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
