Skip to content

Correctness issues in S3 Select Pushdown #17775

Closed
@alexjo2144

Description

@alexjo2144

An initial set of corrections is here: #17563

  • This fixed issues around null comparison and added tests

While testing that PR there were a few other issues I ran into:

  • Comparison of decimal types produces incorrect results on AWS. (works on Minio)
  • TEXTFILE tables (using CSV in S3 Select) fail to parse or produce incorrect results if these three properties are not explicitly set
    textfile_field_separator, textfile_field_separator_escape, null_format
    • The reason for this is that AWS uses default values which assume CSV encoding, but they are not the same defaults as LazySimpleSerde which is what it's mapped to in Trino.
  • The S3SelectCsvRecordReader checks for a quote character in the Hive schema, but that property can't be set when creating tables in Trino using the TEXTFILE format

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions