Look up parquet columns by name case-insensitively

[billonahill]

Parquet schemas are case sensitive but hive is not so camelCase parquet fields are not found when accessed as lowercase hive columns.

[billonahill]

No, nested fields with caps are still broken (IQ-129).

[dain]

@zhenxiao, does this look good to you? Also, any ideas on how we can test for this?

[zhenxiao]

@dain @billonahill the fix looks good to me.
Kind of difficult to reproduce it in our current test case. Let me think of a way.

[billonahill]

@dain @zhenxiao yeah, I also looked at the tests to see how to verify this in a test case but I came up blank. Let me know if you have any ideas.

Also, since submitting this patch I released that we need to further patch the twitter fork to handle hive reserved words. When a parquet field is a reserved Hive word (i.e., keyword, location), we append an '_' to the field name before adding to Hive. This works fine when looking up presto columns by position, but not by name. As a result we also need to check for a match after stripping the _ (see https://github.com/twitter-forks/presto/pull/32/files). This isn't something we want in the OSS master but I am curious how others deal with reserved Hive keywords.

[nezihyigitbasi]

@dain @billonahill @zhenxiao I created a PR for a simple unit test.

[zhenxiao]

@dain @billonahill
@nezihyigitbasi 's PR looks good to me. What do you think?

[billonahill]

@dain @zhenxiao unit tests look good to me. Thanks @nezihyigitbasi!

I just refactored this patch to be easier to further patch on a private branch, like we're doing in https://github.com/twitter-forks/presto/pull/32/files.

[dain]

Merged, thanks!