1.5 KiB
Polars connector
PolarsConnector (lawsynth_connectors.polars) ingests a Polars DataFrame or
LazyFrame into the common batch pipeline. Polars is optional and ships under the
dataframes extra; using the connector without it raises
DependencyUnavailableError.
Sources and resolution
The frame is resolved from the request options data, then the connector's
configured data, then a datasets mapping keyed by resource name. A missing
frame raises ConfigurationError. Both eager DataFrame and lazy LazyFrame
inputs are accepted; a DataFrame is converted to lazy internally.
Reading
from lawsynth_connectors import ConnectorConfig, ReadRequest
from lawsynth_connectors.polars import PolarsConnector
connector = PolarsConnector(ConnectorConfig(name="polars", options={"data": frame}))
with connector:
rows = connector.read_all(ReadRequest("frame", columns=["t", "x"]))
records_from_polars collects lazily: it applies the projection with select,
slices by offset/limit, and collects with the streaming engine by default
(streaming=True), then yields named rows via iter_rows(named=True). Requested
columns that do not exist are rejected with DataValidationError. The
streaming option must be a boolean.
Capabilities
read and projections. The connector reads Polars frames into plain records
flowing through the standard bounded DataBatch pipeline with fingerprints; it
does not write frames or perform domain inference.