You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Match manifest_entry and data_file fields by field-id in Manifest::parse_avro #3650
Is your feature request related to a problem or challenge?
Manifest::parse_avro matches manifest_entry and data_file fields by name through serde. A manifest must use valid Iceberg column projection, and projection selects columns by field ID. The spec fixes these names, so a mismatch is less likely than for partition fields (#3566). A writer that uses other names would have its optional fields read as null and its required fields rejected. iceberg::arrow::read_manifest_entries (#3388) already matches by field ID.
Describe the solution you'd like
Match these fields by field-id before name, as #3566 does for partition fields. This came out of the #3354 work and is listed in #3262.
Activity