[core] [revert] Revert abstract manifest entry #1114
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
[core] revert abstract manifest entry code
Purpose
Abstract manifest entry is created as a loader for scan by projection push down. We can define a SimpleManifestEntry class when we trigger a manifest file scan with projection push down. But the main format we use for manifest file is "avro", which is a column oriented file format. We test it with projection push down, the result is not what we expected.
The promotion for avro is extraordinarily limited. About 10-20 percent performance improved. It is not worth making the project complex by defining a AbstractManifestEntry for such a little performance. So we decided to revert the related code, but we still remain the code about the memory limit.
In addition, we test orc file format as well. Found it has a big performance improvement about 100%. But it's unrealistic to shift to this in production environment. I will list the result of the test below:
Entries in one manifest file: 30000
Columns in one table: 30 (20 String Type)
For orc manifest file format:
Benchmark Mode Cnt Score Error Units
ReadWrite.testListPartition thrpt 4 66.479 ± 38.925 ops/s
ReadWrite.testListPartitionFull thrpt 4 34.702 ± 23.081 ops/s
For avro manifest file format:
Benchmark Mode Cnt Score Error Units
ReadWrite.testListPartition thrpt 4 14.758 ± 5.765 ops/s
ReadWrite.testListPartitionFull thrpt 4 12.903 ± 1.489 ops/s
Tests
API and Format
Documentation