adriangb opened a new issue, #19387: URL: https://github.com/apache/datafusion/issues/19387
The projection pushdown optimizer rule / implementations generally only push down a projection if it "narrows" a schema (i.e. has less output expressions than input expressions) and the output expressions are all columns or literals: https://github.com/apache/datafusion/blob/d68b629dc610972295d8f310b09cd854cf250dd3/datafusion/physical-plan/src/filter.rs#L470-L471 https://github.com/apache/datafusion/blob/d68b629dc610972295d8f310b09cd854cf250dd3/datafusion/physical-plan/src/repartition/mod.rs#L1045-L1055 https://github.com/apache/datafusion/blob/d68b629dc610972295d8f310b09cd854cf250dd3/datafusion/physical-plan/src/projection.rs#L255-L268 This is problematic with a plan like: ``` copy ( select 1 as id, named_struct('large_string_field', 'big text!', 'small_int_field', 2) as large_struct ) TO 'struct.parquet'; create external table t stored as parquet location 'struct.parquet'; explain format indent select large_struct['small_int_field'] * 2 from t where id = 1; ``` ``` +---------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | plan_type | plan | +---------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ | logical_plan | Projection: get_field(t.large_struct, Utf8("small_int_field")) * Int64(2) | | | Filter: t.id = Int64(1) | | | TableScan: t projection=[id, large_struct], partial_filters=[t.id = Int64(1)] | | physical_plan | ProjectionExec: expr=[get_field(large_struct@0, small_int_field) * 2 as t.large_struct[small_int_field] * Int64(2)] | | | CoalesceBatchesExec: target_batch_size=8192 | | | FilterExec: id@0 = 1, projection=[large_struct@1] | | | RepartitionExec: partitioning=RoundRobinBatch(12), input_partitions=1 | | | DataSourceExec: file_groups={1 group: [[Users/adrian/GitHub/datafusion/struct.parquet]]}, projection=[id, large_struct], file_type=parquet, predicate=id@0 = 1 | | | | +---------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
