Avro reader has very limited test coverage

[vuule]

Avro reader currently has a single test, with a few more basic tests pending to be merged.

Some cases not covered:

• Large input files;
• Columns of all supported types;

[vuule]

Marking as P0 for 0.18 as it's preventing other Avro refactoring work.

[cwharris]

Maybe we can use Apache's python-based reader/writer to generate test data. Generating files via python from the c++ side seems backwards, though. Do we have a methodology to store/load pre-generated test data/files in tests?

[vuule]

Maybe we can use Apache's python-based reader/writer to generate test data. Generating files via python from the c++ side seems backwards, though. Do we have a methodology to store/load pre-generated test data/files in tests?

Maybe I'm misunderstanding the question, but you can generate tmp avro files using fastavro in Python tests. There are tests that already do this.

[cwharris]

It would be nice to have c++ tests, though, since the reader is implemented at that level.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[vuule]

missing large tests still relevant

[vuule]

Closing, since #12912 added larger tests