Interface parity: PlanFrame pandas skin vs pandas `DataFrame`

This page compares PlanFrame’s pandas-flavored interface (planframe.pandas.PandasLikeFrame) to pandas.DataFrame.

PlanFrame is always lazy and (in the typed surface) index-free, so parity is primarily about API shape and common call patterns, not exact eager/index semantics.

Legend

Exact parity: same name + same argument shape + same behavior (rare for pandas).
Typed-parity: same shape, but restricted to typed-safe subsets.
Divergence: present but intentionally differs (documented).
Unsupported: not implemented; should fail loudly.

High-value mapping

pandas	PlanFrame pandas skin	Status	Notes
`df["col"]`	`__getitem__(str)`	Typed-parity	Returns a typed `Series` expression wrapper, not data.
`df[["a","b"]]`	`__getitem__(Sequence[str])`	Typed-parity	Lowered to `Frame.select`.
boolean indexing	`__getitem__(Series[bool] \\| Expr[bool])`	Typed-parity	Lowered to `query(...)` (typed predicate).
`assign(...)`	`assign(...)`	Typed-parity	Lowered to repeated `with_columns(...)`.
`sort_values(...)`	`sort_values(...)`	Typed-parity	Lowered to core `sort`.
`drop(columns=...)`	`drop(...)`	Divergence	Columns-only; `axis="index"` is unsupported.
`rename(columns=..., errors=...)`	`rename_pandas(...)`	Typed-parity	Mirrors pandas argument shape; lowered to core `rename`.
`dropna(...)`	`dropna(...)`	Divergence	Row-wise only; lowered to core `drop_nulls(subset=...)`.
`fillna(...)`	`fillna(...)`	Divergence	Lowered to typed `fill_null`/`fill_null_many`.
`query("...")`	`query(Series[bool] \\| Expr[bool])`	Divergence	No string expression parser in typed surface.
`filter(...)`	`filter(predicate)` / `filter(items/like/regex)`	Divergence	Dual semantics (row filter vs column selection), chosen by arguments.
`merge(...)`	`merge(...)`	Divergence	Lowered to core `join`; index semantics differ.

Major gaps (expected)

Index semantics: .loc, .iloc, alignment, axis=0 drop, etc.
Mutation: assignment (df["x"]=...) and inplace ops.
String expression systems: pandas query("...") and eval("...").
Arbitrary UDFs: apply, map, Python lambdas over rows/columns.

Generated inventories (method-by-method)

For an exhaustive, generated inventory:

PandasLikeFrame surface: planframe/design/_generated/interface-inventory-pandas.md
pandas methods missing from PlanFrame pandas skin: planframe/design/_generated/pandas-missing.md

Missing / divergent areas (categorized)

Index / axis semantics (unsupported): .loc/.iloc, alignment, drop(axis=0), index-aware merges, etc.
Mutation / inplace (unsupported): inplace=True, assignment APIs.
String DSL (tiny subset): we can support only very small query("col op literal") patterns; the full pandas expression engine is out of scope for strict typing.
UDF-heavy APIs (unsupported): apply, map, applymap.

Interface parity: PlanFrame pandas skin vs pandas DataFrame