Add a custom table¶
A custom table is a table type you define yourself, for data that does not fit any of the predefined types.
No public extension API yet
The registry that maps a type string to a table class is internal, so there is no
supported way to register your own table type today. Until there is,
generic tables are the way to store arbitrary tabular data: they
accept a pandas DataFrame, a polars DataFrame or LazyFrame, or an AnnData
object.
One caveat when choosing a backend: only the AnnData backend can
write an AnnData payload. If your generic table holds one, saving it through the
Parquet, CSV or JSON backend raises NotImplementedError — convert it to a
DataFrame first, or keep it on the AnnData backend.
If you need a genuinely new table type — with its own validation and metadata — open an issue on GitHub describing your use case.
See also¶
- Tables overview — the three-component table architecture.
- Generic tables — the untyped fallback.
- Tables API reference —
TablesContainerand the table classes.