fp-frame

DataFrame and Series with pandas-API parity, AACE index alignment, and GroupBy / Rolling / Resample integration.

Part of the frankenpandas workspace. The central data-structure crate; most downstream code imports from here.

When to depend on fp-frame directly

Most users should depend on the umbrella frankenpandas crate instead (it re-exports DataFrame, Series, and the prelude). Direct dependency makes sense when:

• Building a library that doesn't need the IO layer (fp-io) but does need DataFrame + Series.
• Embedding DataFrame primitives without the expression parser (fp-expr) or conformance harness (fp-conformance).

Key types

• DataFrame — columnar frame with flat or multi-level row index, optional column MultiIndex, per-column validity masks. Supports all pandas IO/arithmetic/reshape/groupby surfaces at API-parity.
• Series — single-column equivalent with pandas-compatible string accessor (.str), datetime accessor (.dt), categorical accessor (.cat).
• DataFrameGroupBy / SeriesGroupBy — three automatic execution paths (dense Int64 keys, arena-backed, HashMap) chosen per group cardinality. 87% speedup on dense paths over the naive reference.
• Rolling / Expanding / Ewm / Resample — window + resample engines, re-used by DataFrameRolling / etc.

Status

Stable surface at API-parity for covered operations (328+ ops per fp-conformance packet suite). All public error types are #[non_exhaustive] per br-frankenpandas-tne4.

Links

• Workspace README — full API tour.
• Conformance packets — which operations have differential conformance coverage.
• CHANGELOG