Platform Support Status¶

Tags intermediate reference

BenchBox platform metadata now carries a support_status value from the public support taxonomy: stable, beta, experimental, repo_only, deprecated, or document_only.

This status is intentionally separate from local dependency availability. For example, Snowflake can be beta product surface while unavailable on a machine that has not installed snowflake-connector-python.

Registry Snapshot¶

The authoritative source is benchbox/core/platform_registry.py.

Platform registry: 50 metadata entries; 45 SQL-capable; 19 DataFrame-capable; 14 dual-mode; support status counts: stable=5, beta=27, experimental=17, deprecated=1.

Status	Entries	SQL-capable	DataFrame-capable	Dependency group	CLI exposure	MCP exposure	Docs exposure
`stable`	5	3	3	Core or single local extra	Listed by registry; runnable when dependency is installed	Discovery/control-plane metadata via registry	Full platform docs
`beta`	27	26	3	Optional extras, credentials, or local services	Listed by registry; may need SDK/credentials	Discovery/control-plane metadata via registry	Platform docs and setup notes
`experimental`	17	15	13	Optional, GPU/native, or cloud Spark extras	Listed but labeled experimental in metadata/docs	Discovery metadata only unless tool support is documented	Experimental caveats required
`repo_only`	0	0	0	Source checkout only	Hidden from normal runtime	Not exposed	Developer docs only
`deprecated`	1	1	0	Compatibility selector	Retained for migration	Existing-client exposure only	Migration docs required
`document_only`	0	0	0	None	Not runnable	Not exposed	Docs must say not executable

Optional Import Diagnostics¶

Normal platform discovery remains fail-open for optional dependencies. Missing optional SDKs, unavailable native libraries, and broken adapter imports are recorded only when diagnostics are requested:

uv run -- python - <<'PY'
from benchbox.core.platform_registry import PlatformRegistry

for platform, diagnostic in PlatformRegistry.diagnose_optional_adapter_imports(["snowflake", "datafusion"]).items():
    print(platform, diagnostic["status"], diagnostic["support_status"])
PY

Diagnostic statuses:

Status	Meaning
`available`	The adapter module imported and exposed the expected adapter class.
`missing_optional_dependency`	A package needed by the adapter is not installed.
`native_library_load_failure`	A native library or ABI-dependent module failed to load.
`broken_adapter_import`	The BenchBox adapter module imported incorrectly or did not expose the expected class.
`deprecated_platform`	The platform selector is retained only for migration.
`intentionally_disabled`	The support status indicates a non-runtime surface.
`not_configured`	The requested platform is not in optional adapter registration.

Inventory Notes¶

Top-level platform files that are not direct optional adapter registration modules were classified during this migration:

Path	Classification	Rationale
`benchbox/platforms/fabric_spark.py`	supported wrapper	Stable import path that re-exports `benchbox.platforms.azure.fabric_spark_adapter.FabricSparkAdapter`.
`benchbox/platforms/datafusion_query_transformer.py`	supported helper	Lazy helper imported by the DataFusion adapter for SQL compatibility.
`benchbox/platforms/datafusion_write_transformer.py`	supported helper	Write-path SQL compatibility helper for DataFusion.
`benchbox/platforms/presto_trino_utils.py`	supported helper	Shared helper module used by Presto-family adapters.
`benchbox/platforms/questdb_rewriter.py`	supported helper	QuestDB SQL compatibility helper.
`benchbox/platforms/cudf.py`	removal candidate	Legacy top-level cuDF adapter imports `benchbox.experimental.gpu`; current registry metadata marks `cudf` as experimental DataFrame support through the DataFrame platform path.

The cuDF removal/integration decision is tracked as a follow-up TODO so this migration can avoid deleting platform files outside its scope.