jorgecarleitao commented on pull request #7751: URL: https://github.com/apache/arrow/pull/7751#issuecomment-660445062
I understood that many of the arrow data types are written to parquet with different types, and are casted to the correct type on read/write. IMO the reason we cannot test it is because the current C++ does not write date64 as a 64 bit int in parquet, and thus we have no way of constructing a 64int column in parquet with data type date64. However, if another implementation writes a date64 to parquet, we need a method to read from it and AFAI can tell, the previous code is wrong. IMO this PR contains the most correct code to handle date64 - it is just untestable atm, like the previous code. So, between a certainly wrong and untested code and a potentially correct but untested code, I would go for the latter :P We could also just throw an error, but we also can't test that because that branch is unreachable in tests. ^_^ I rest my case, and I will change the code to whatever version you prefer, @sunchao :-) ---------------------------------------------------------------- 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: us...@infra.apache.org