Skip to content

Reading parquet with (pre-release) arrow fails with "out of order projection is not supported" #2543

@alamb

Description

@alamb

Describe the bug

When testing using a pre-release version of arrow-rs with apache/arrow-rs#1682 from @tustvold which added stricter error checking two datafusion tests fail with "out of order projection is not supported"

---- physical_plan::file_format::parquet::tests::evolved_schema_filter stdout ----
thread 'physical_plan::file_format::parquet::tests::evolved_schema_filter' panicked at 'called `Result::unwrap()` on an `Err` value: ArrowError(ExternalError(ParquetError(General("out of order projection is not supported"))))', datafusion/core/src/physical_plan/file_format/parquet.rs:968:14

---- physical_plan::file_format::parquet::tests::evolved_schema_inconsistent_order stdout ----
thread 'physical_plan::file_format::parquet::tests::evolved_schema_inconsistent_order' panicked at 'called `Result::unwrap()` on an `Err` value: ArrowError(ExternalError(ParquetError(General("out of order projection is not supported"))))', datafusion/core/src/physical_plan/file_format/parquet.rs:819:14

To Reproduce
See #2530

Expected behavior
Tests should succeed

Additional context
@tustvold suggests there is some interplay of the parquet schema adapter logic that was silently masking a bug in arrow-rs. Now that arrow-rs has stricter error checking, the query is erroring #2530 (comment)

Metadata

Metadata

Assignees

Labels

bugSomething isn't working

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions