java-topology/defects/cpython/patch/cpython-0001-codegen-pattern-stores-list-contains.patch

55 lines
2.3 KiB
Diff

# UNDF: UNDF-2026-000000039
# UNDF: (leave blank)
# cpython-0001: codegen pattern-match stores duplicate check O(S^2)
#
# In Python/codegen.c, codegen_pattern_helper_store_name() and the
# mapping-pattern loop both call PySequence_Contains(pc->stores, name)
# where pc->stores is a PyList. Each call is O(S) where S is the number
# of store names accumulated so far. Since this is called once per store
# name, the total cost is O(S^2).
#
# Fix: Use a PySet alongside pc->stores for O(1) membership checks.
# pc->stores remains a list (order matters for stack rotation), but
# a parallel set provides O(1) duplicate detection.
#
# Severity: MEDIUM — match statements with many capture variables
# (e.g., structural pattern matching on large data classes) hit O(S^2).
# At S=100 capture variables, ~5000 comparisons vs 100 set lookups.
#
--- a/Python/codegen.c
+++ b/Python/codegen.c
@@ -5908,7 +5908,7 @@ codegen_pattern_helper_store_name(compiler *c, location loc,
ADDOP(c, loc, POP_TOP);
return SUCCESS;
}
- int duplicate = PySequence_Contains(pc->stores, n);
+ int duplicate = PySet_Contains(pc->stores_set, n);
RETURN_IF_ERROR(duplicate);
if (duplicate) {
return codegen_error_duplicate_store(c, loc, n);
@@ -5916,6 +5916,7 @@ codegen_pattern_helper_store_name(compiler *c, location loc,
Py_ssize_t rotations = pc->on_top + PyList_GET_SIZE(pc->stores) + 1;
RETURN_IF_ERROR(codegen_pattern_helper_rotate(c, loc, rotations));
RETURN_IF_ERROR(PyList_Append(pc->stores, n));
+ RETURN_IF_ERROR(PySet_Add(pc->stores_set, n));
return SUCCESS;
}
@@ -6395,7 +6396,7 @@ codegen_pattern_helper_store_name(compiler *c, location loc,
// Update the list of previous stores with this new name, checking for
// duplicates:
PyObject *name = PyList_GET_ITEM(control, i);
- int dupe = PySequence_Contains(pc->stores, name);
+ int dupe = PySet_Contains(pc->stores_set, name);
if (dupe < 0) {
goto error;
}
@@ -6405,6 +6406,9 @@ codegen_pattern_helper_store_name(compiler *c, location loc,
}
if (PyList_Append(pc->stores, name)) {
goto error;
+ }
+ if (PySet_Add(pc->stores_set, name)) {
+ goto error;
}
}