Icebird WRITER

[platypii]

Read path

• Iceberg v3 reads: row lineage (_row_id, _last_updated_sequence_number), v3 types (variant, geometry, geography), Puffin deletion vectors, equality delete matching, delete partition scope
• Hidden + identity partition support, manifest-list partition pruning, parallelized data-file reads
• Gzip metadata JSON (.gz.metadata.json, metadata.json.gz)

Write path (experimental, v2 + v3 deletion vectors)

• icebergCreateTable, icebergDropTable, icebergAppend, icebergDelete, icebergSetRef, icebergExpireSnapshots
• Generic Catalog abstraction — same calls work against fileCatalog or REST
• Position deletes as Parquet or Puffin deletion vectors; equality delete manifests
• Append/staged-commit flow with per-column stats, sort orders, partition spec evolution, fixed/decimal + geospatial stats, lower/upper bound truncation
• Honors format-version, write.format.default, write.metadata.previous-versions-max, write-default, compressor options
• Schema/spec updates: add-schema, set-current-schema, add-spec/set-default-spec, set-properties/remove-properties, full assert-* requirement checks, V3-only-type validation
• Async hyparquet writer, in-memory end-to-end test harness

REST catalog

• Connect, list/load tables, create/drop/rename table, create/delete namespace, register table, update table, load credentials

I/O

• New Resolver/Lister abstraction (pluggable storage); urlResolver ships writer/deleter for HTTP PUT/DELETE
• Auth now flows through requestInit on the resolver/lister rather than per-call (breaking)