compute_series#

nominal.experimental.compute_as_code.compute_series(client: NominalClient, expr: nominal.compute.NumericSeries | nominal.compute.CategoricalSeries, inputs: Mapping[str, Channel], start: _FlexibleTimestamp | None = None, end: _FlexibleTimestamp | None = None, *, tags_by_reference: Mapping[str, Mapping[str, str]] | None = None, name: str = 'value', enable_gzip: bool = True) → pd.Series[Any]#

Compute a derived series by applying a nominal.compute expression to a set of channels.

Parameters:

  • client – The NominalClient to compute with.

  • expr – The compute-as-code expression to evaluate (numeric or categorical), typically authored in nominal.compute and imported from your own repository.

  • inputs – Maps each reference name used in expr to the Channel it should resolve to. Look these up from the run/dataset the caller wants to compute against.

  • start – Start of the time range to compute over. Defaults to the earliest supported timestamp.

  • end – End of the time range to compute over. Defaults to the latest supported timestamp.

  • tags_by_reference – Optional tag filters, keyed by reference name (the same keys as inputs) and then by tag key. Use this when the channel bound to a reference carries more than one tagged series. For example, tags_by_reference={"velocity": {"vehicle": "car_1"}} narrows the channel bound to reference "velocity" down to the series tagged vehicle=car_1. References omitted from tags_by_reference are bound without a tag filter.

  • name – Name for the returned series (also the CSV column name requested from the export service).

  • enable_gzip – If true, gzip the export from Nominal.

Returns:

A pandas.Series of the computed values, indexed by timestamp. The index name is “timestamp” and the series name is name.