Skip to content

Row lineage #4080

Open
Open

Description

@rambleraptor

Feature Request / Improvement

We want to be able to read the row-lineage values on a table. This is a blocker for writing v3 tables.

table.scan(selected_fields=("id", "_row_id", "_last_updated_sequence_number"))
  • Data files with a null first_row_id inherit it from the manifest's first_row_id plus a running record count: Add row-lineage support with Spark tests #3972
  • Define the reserved columns _row_id (2147483540) and _last_updated_sequence_number (2147483539)
  • Let scans project them
  • Use the values stored in the file if there are any. Fill nulls with first_row_id + pos and the file's data sequence number.
  • Compute pos before applying deletes
  • Show first_row_id in inspect.files()
  • REST scan planning drops first_row_id
  • Integration test: each file's first_row_id matches Spark's MIN(_row_id): Add row-lineage support with Spark tests #3972
  • Integration test: Spark MERGE / UPDATE / rewrite_data_files, then compare with Spark's SELECT _row_id, _last_updated_sequence_number

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions