Quanlong Huang created IMPALA-13553:
---------------------------------------
Summary: Always evaluate small runtime filters on parquet
dictionary
Key: IMPALA-13553
URL: https://issues.apache.org/jira/browse/IMPALA-13553
Project: IMPALA
Issue Type: Improvement
Reporter: Quanlong Huang
IMPALA-5509 supports evaluating runtime filters on Parquet dictionaries to
filter out RowGroups. It adds a query option,
PARQUET_DICTIONARY_RUNTIME_FILTER_ENTRY_LIMIT (defaults to 1024), to control
how large a dictionary should disable this optimization. The motivation is to
avoid false positives.
However, for runtime filters that generated by small build sides, e.g. just
containing one value, it still worths to evaluate them on the dictionaries. And
this won't take long comparing to evaluating the small runtime filters on all
rows.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
