Github user sachouche commented on the issue:
https://github.com/apache/drill/pull/1060
I have updated this pull request with the following changes:
- Excluded the implicit column optimizations from this pull request (will
be included as part of another Drill Jira)
- Tuned a threshold (16bytes vs 1024) on when to use "putByte() vs
copyMemory()" based on @vrozov findings when benchmarking using JMH
- There were lingering questions on the use of byte arrays for intermediary
processing; after a discussion with @vrozov we agreed to tackle such an
analysis within a new JIRA
([DRILL-6301](https://issues.apache.org/jira/browse/DRILL-6301?filter=-1)) so
that the Drill customer can immediately rip the benefits of the current
enhancements
- Moved the newly added MemoryUtils class to the Parquet flat reader
project (made it package private); this is temporary till DRILL-6301 is
resolved.
---
