Sorry I didn't reply to this earlier. I think that this probably breaks what it was intended to test.
You can pass Parquet the same buffer with different values and Parquet should make a defensive copy of the buffer if it is clear that the buffer will be reused. By changing to a new buffer for every value, we are no longer testing the case where Parquet has to make that defensive copy. We need to test this path for cases where input buffers are reused. For example, when you read an Avro file with a fixed field, the same GenericFixed instance is used and its buffer is passed as a reused buffer when writing to Parquet. If this is misused for testing in `FileEncodingsIT`, then we should fix that class. The random data generator works like a common real-world case, so it shouldn't change. [ Full content available at: https://github.com/apache/parquet-mr/pull/504 ] This message was relayed via gitbox.apache.org for [email protected]
