Skip to content

ZarrTable

zarr_datafusion_search.ZarrTable

A DataFusion table provider that exposes a Zarr metadata store as a SQL-queryable table.

ZarrTable implements the DataFusion TableProvider protocol via the FFI boundary, so it can be registered directly with a SessionContext using register_table_provider.

Use one of the async class methods to construct an instance:

Example 
import asyncio
from datafusion import SessionContext
from zarr_datafusion_search import ZarrTable

async def main():
    table = await ZarrTable.from_obstore(store, "/meta")
    ctx = SessionContext()
    ctx.register_table_provider("items", table)
    df = ctx.sql("SELECT date, collection FROM items LIMIT 10")
    df.show()

asyncio.run(main())

__datafusion_table_provider__ 

__datafusion_table_provider__(session: Any) -> Any

Return the FFI TableProvider capsule for DataFusion registration.

This is called automatically by SessionContext.register_table_provider. You do not need to call it directly.

from_icechunk classmethod 

from_icechunk(session: Any, group_path: str) -> Awaitable[ZarrTable]

Create a ZarrTable from an Icechunk session.

Parameters:

  • session (Any) –

    An open icechunk.Session pointing to the store.

  • group_path (str) –

    Absolute path to the Zarr group containing the metadata arrays (e.g. "/meta").

Returns:

  • Awaitable[ZarrTable] –

    An awaitable that resolves to a ZarrTable instance.

Example 
table = await ZarrTable.from_icechunk(session, "/meta")

from_obstore classmethod 

from_obstore(store: Any, group_path: str) -> Awaitable[ZarrTable]

Create a ZarrTable from an obstore object store.

Parameters:

  • store (Any) –

    Any obstore-compatible object store (e.g. obstore.store.S3Store, obstore.store.LocalStore).

  • group_path (str) –

    Absolute path to the Zarr group containing the metadata arrays (e.g. "/meta").

Returns:

  • Awaitable[ZarrTable] –

    An awaitable that resolves to a ZarrTable instance.

Example 
import obstore.store
store = obstore.store.LocalStore("/path/to/data.zarr")
table = await ZarrTable.from_obstore(store, "/meta")