fix: FusedCodecPipeline falls back to async path for sharded arrays with async-only inner codecs

[d-v-b]

🤖 AI text below 🤖

Problem

Under the opt-in FusedCodecPipeline (zarr.config.set({"codec_pipeline.path": "zarr.core.codec_pipeline.FusedCodecPipeline"})), a sharded array whose inner codec chain contains a third-party codec implementing only the async codec interface (_encode_single/_decode_single, no SupportsSyncCodec) raised on both write and read:

TypeError: All codecs must implement SupportsSyncCodec. The following do not: <name>

The pipeline's existing guard handles async-only codecs at the top level of the chain (evolve_from_array_spec catches the TypeError and falls back to the async path), but ShardingCodec structurally satisfies SupportsSyncCodec, so the pipeline committed to the sync fast path and crashed inside the sharding codec's inner ChunkTransform. The default BatchedCodecPipeline handles the same configuration correctly. Found in the 3.3.0 pre-release audit.

Fix

Sync capability is now a dynamic query, _codec_supports_sync (src/zarr/abc/codec.py): structural protocol membership plus an optional _sync_capable opt-out.

• ShardingCodec._sync_capable reports False when its inner or index codec chain is not fully sync-capable (recursive, so nested sharding propagates).
• ChunkTransform.__post_init__ consults the query, so construction raises for such chains and the pipeline's existing top-level guard sets sync_transform=None — declining the sync fast path and routing reads through the async partial shard decode (byte-range coalesced, matching BatchedCodecPipeline) and writes through the async fallback.

All-sync chains still build the sync transform and keep the fast path, byte-identical. BatchedCodecPipeline behavior is unchanged. No towncrier fragment: FusedCodecPipeline is new in unreleased 3.3.0.

Tests

• New regression test: sharded array with an async-only inner codec round-trips (full write, partial write, full read, partial read) under FusedCodecPipeline, asserts the codec actually ran (invocation counters), and re-reads the store under the default pipeline to prove the bytes are valid cross-pipeline.
• tests/test_fused_pipeline.py tests/test_pipeline_parity.py tests/test_fastpath_equivalence.py tests/test_codecs/test_sharding.py tests/test_codec_pipeline.py: 504 passed, 6 skipped (pre-existing skips).
• Instrumented check: async-only inner chain → partial reads go through ShardingCodec._decode_partial_single (no whole-shard over-read); all-sync chain → read_sync fast path still fires.
• mypy and ruff clean on all touched files.

🤖 Generated with Claude Code