Skip to content
Docs Pydantic Docs

List / set / dict comprehensions

Monty inlines list, set, and dict comprehensions into the surrounding code object. The user-visible behaviour follows PEP 709: inlined comprehensions, no synthetic frame in tracebacks, comprehension targets do not leak into the enclosing scope.

Divergences from CPython

  • A comprehension target must be a name. CPython accepts [i for obj.x in xs] and [i for d[k] in xs]; Monty raises SyntaxError: comprehension target must be a name, not an attribute (or not a subscript). A comprehension’s targets live in operand-stack slots, which a store to an object cannot reach. Attribute and subscript targets work in every other unpacking position.
  • locals() while a comprehension is running. CPython exposes the comprehension’s active targets in locals() during the comprehension body. Monty does not implement locals() introspection.
  • Generator expressions. (x for x in iterable) parses but currently materialises to a list rather than a lazy iterator.
  • Maximum number of for clauses. Monty caps a single comprehension at 255 for clauses; exceeding this raises SyntaxError: comprehension has too many nested clauses (N); maximum is 255. Per-clause operand-stack growth means real comprehensions hit a tighter SyntaxError: comprehension target + iterator count exceeds u8 depth operand well before that point. CPython has no equivalent compile-time limit. The cap bounds compiler recursion depth on attacker-controlled source.