This idea came up during a conversation with @Fidget-Spinner.

dct = frozendict(a=1)
def foo():
    dct['a'] # fold this to 1

When a frozendict is a promoted constant in the JIT, subscript lookups like dct['a'] should be foldable to the constant value at compile time. This would make frozendict lookups and match/case on frozendict significantly faster. If implemented, this optimization could create strong motivation for users to adopt frozendict in many more places, anywhere a dict is used as a read only lookup table, configuration mapping, or dispatch table.

Plan to start from REPLACE_OPCODE_IF_EVALUATES_PURE in optimizer_bytecodes.c and extend it to handle _BINARY_OP_SUBSCR_DICT for frozendict constant

cc @vstinner

Linked PRs

• gh-146381: Fold frozendict(key=const, ...) calls to LOAD_CONST #146423
• gh-146381: Constant-fold frozendict subscript lookups via REPLACE_OPCODE_IF_EVALUATES_PURE #146490
• gh-146381: Constant fold local frozendict subscript lookups via recording #146564