Skip to content

Latest commit

 

History

History
35 lines (24 loc) · 1.24 KB

File metadata and controls

35 lines (24 loc) · 1.24 KB

Debugging log for Stack v3 metadata schemas

Keep Stack v3 repository metadata in Parquet without schema drift between writer batches.

Initial status

The CoreWeave repository transform failed after its download completed. PyArrow inferred github_metadata.forked_from as null in one batch and string in a later batch, then rejected the later table because the Parquet file schema was already fixed. After the transform fix, normalization failed for the same reason because its final writers inferred the schema again.

Nullable metadata schema

The transformed documents preserve nested repository and file metadata. Several source fields are nullable, so inference from an arbitrary first batch cannot define a stable output schema.

Changes to make

Pass explicit PyArrow schemas to the Stack v3 transform and normalization writers. Keep nullable values as nulls while fixing their logical types across every batch.

Results

A local transform and normalization containing a null github_metadata.forked_from now write that field as nullable string and read the null value back unchanged.

Production validation

Relaunch the GCP and CoreWeave transforms and confirm the affected shards complete before considering the incident resolved.