Martin Junghanns created SPARK-26088:
----------------------------------------
Summary: DataSourceV2 should expose row count and attribute
statistics
Key: SPARK-26088
URL: https://issues.apache.org/jira/browse/SPARK-26088
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 2.3.0
Reporter: Martin Junghanns
During investigation of CBO and DataSourceV2 we found, that
{code}
org.apache.spark.sql.sources.v2.reader.Statistics
{code}
misses attribute/column statistics and that
{code}
DataSourceV2Relation#computeStats
{code}
wraps
{code}
org.apache.spark.sql.sources.v2.reader.Statistics
{code}
into
{code}
org.apache.spark.sql.catalyst.plans.logical.Statistics
{code}
without forwarding the optional {{rowCount}} if present.
However {{rowCount}} and {{attributeStats}} are used during CBO e.g. in
{{JoinEstimation}} and {{AggregateEstimation}}.
We propose that:
* {{org.apache.spark.sql.sources.v2.reader.Statistics}} mirrors
{{org.apache.spark.sql.catalyst.plans.logical.Statistics}}
* {{DataSourceV2Relation}} forwards all the information to be available during
CBO
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
