FAQ

Is PlanFrame a dataframe library?

No. PlanFrame is a typed planning layer that delegates execution to a backend via an adapter.

Does chaining execute backend work?

No. PlanFrame is always lazy; execution happens at explicit boundaries like collect().

Is there an async API?

Yes. Frame exposes acollect, ato_dicts, and ato_dict, which await the adapter’s async hooks (BaseAdapter.acollect, etc.). Discoverable aliases collect_async, to_dicts_async, and to_dict_async are the same operations. For awaiting plan evaluation without blocking the event loop, core also provides execute_plan_async (it runs the sync execute_plan in asyncio.to_thread). Defaults run synchronous adapter methods in a worker thread so existing backends keep working without code changes. See Migrating since v1.1.0.

Can I detect PlanFrame column selectors at runtime?

Yes. planframe.selector.ColumnSelector is a @runtime_checkable protocol, so isinstance(obj, ColumnSelector) works for built-in selector types and for structural matches (objects with a compatible select(self, schema) method).

Is there a PySpark-like or pandas-like API?

Yes. The core package ships optional, typed skins—planframe.spark (PySpark-style SparkFrame, Column, functions) and planframe.pandas (pandas-style PandasLikeFrame, Series). They are mixins on Frame and do not add Spark or extra runtime deps. The planframe-pandas adapter’s PandasFrame uses PandasLikeFrame by default. See the PySpark-like API and pandas-like API guides.