corwinjoy commented on issue #39676: URL: https://github.com/apache/arrow/issues/39676#issuecomment-1899362878
To your other comments, yes the giant metadata header is the big bottleneck for large tables. Once there are a lot of rows and columns it can take 100x longer to parse that big header than to read the data (if we are just taking a sample). I think this is because the header gets huge, and, parsing the thrift data is actually quite slow since it needs to be decoded field by field and then recopied. It is probably hard to get the format changed, so my first thought was to only read the metadata for the first rowgroup and use this as a kind of prototype. But, there may be better approaches although I'm not sure what they would be. -- 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]
