Skip to content

gh-150737: Optimize bytecode for empty unpack cases such as {*()}. #150812

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
JelleZijlstra merged 7 commits into from
Sep 21, 2026
+190 −7
Merged

gh-150737: Optimize bytecode for empty unpack cases such as {*()}. #150812

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
67742d9
Optimize leading empty tuple unpack idioms in codegen
gesslerpd May 30, 2026
0baf564
add blurb
gesslerpd Jun 2, 2026
f7e8150
rerun ci
gesslerpd Jun 2, 2026
a0849b0
omit all null unpack idioms
gesslerpd Sep 18, 2026
b05ab89
add empty list/dict
gesslerpd Sep 21, 2026
e841f0c
update blurb
gesslerpd Sep 21, 2026
8425743
Update 2026-06-02-16-21-02.gh-issue-150737.LoYUFY.rst
gesslerpd Sep 21, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
154 changes: 154 additions & 0 deletions Lib/test/test_compile.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1119,6 +1119,160 @@ def or_false(x):
self.assertIn('LOAD_', opcodes[-2].opname)
self.assertEqual('RETURN_VALUE', opcodes[-1].opname)

def test_empty_set_unpack_literal_bytecode_optimization(self):
cases = {
# optimized cases
'{*()}': [
('RESUME', 0),
('BUILD_SET', 0),
('RETURN_VALUE', None),
],
'{*(), *()}': [
('RESUME', 0),
('BUILD_SET', 0),
('RETURN_VALUE', None),
],
'{*(), 1}': [
('RESUME', 0),
('LOAD_SMALL_INT', 1),
('BUILD_SET', 1),
('RETURN_VALUE', None),
],
'{*(), 1, 2, 3}': [
('RESUME', 0),
('BUILD_SET', 0),
('LOAD_CONST', frozenset({1, 2, 3})),
('SET_UPDATE', 1),
('RETURN_VALUE', None),
],
'{1, *()}': [
('RESUME', 0),
('LOAD_SMALL_INT', 1),
('BUILD_SET', 1),
('RETURN_VALUE', None),
],
'{1, 2, 3, *()}': [
('RESUME', 0),
('BUILD_SET', 0),
('LOAD_CONST', frozenset({1, 2, 3})),
('SET_UPDATE', 1),
('RETURN_VALUE', None),
],
'{1, 2, *(), 3}': [
('RESUME', 0),
('BUILD_SET', 0),
('LOAD_CONST', frozenset({1, 2, 3})),
('SET_UPDATE', 1),
('RETURN_VALUE', None),
],
# unoptimized cases
'{*(1,)}': [
('RESUME', 0),
('BUILD_SET', 0),
('LOAD_CONST', (1,)),
('SET_UPDATE', 1),
('RETURN_VALUE', None),
],
'{*(x,)}': [
('RESUME', 0),
('BUILD_SET', 0),
('LOAD_NAME', 'x'),
('BUILD_TUPLE', 1),
('SET_UPDATE', 1),
('RETURN_VALUE', None),
],
}

for source, expected in cases.items():
with self.subTest(source=source):
code = compile(source, '<test>', 'eval')
instructions = [
(instruction.opname, instruction.argval)
for instruction in dis.get_instructions(code)
]
self.assertEqual(instructions, expected)

def test_empty_leading_tuple_unpack_list_and_tuple_bytecode_optimization(self):
cases = {
# optimized cases
'[*()]': [
('RESUME', 0),
('BUILD_LIST', 0),
('RETURN_VALUE', None),
],
'[*(), *()]': [
('RESUME', 0),
('BUILD_LIST', 0),
('RETURN_VALUE', None),
],
'[*(), 1]': [
('RESUME', 0),
('LOAD_SMALL_INT', 1),
('BUILD_LIST', 1),
('RETURN_VALUE', None),
],
'(*(),)': [
('RESUME', 0),
('LOAD_COMMON_CONSTANT', ()),
('RETURN_VALUE', None),
],
'(*(), *())': [
('RESUME', 0),
('LOAD_COMMON_CONSTANT', ()),
('RETURN_VALUE', None),
],
'(*(), 1)': [
('RESUME', 0),
('LOAD_CONST', (1,)),
('RETURN_VALUE', None),
],
'[1, *()]': [
('RESUME', 0),
('LOAD_SMALL_INT', 1),
('BUILD_LIST', 1),
('RETURN_VALUE', None),
],
'[1, 2, 3, *()]': [
('RESUME', 0),
('BUILD_LIST', 0),
('LOAD_CONST', (1, 2, 3)),
('LIST_EXTEND', 1),
('RETURN_VALUE', None),
],
'[1, 2, *(), 3]': [
('RESUME', 0),
('BUILD_LIST', 0),
('LOAD_CONST', (1, 2, 3)),
('LIST_EXTEND', 1),
('RETURN_VALUE', None),
],
# unoptimized cases
'[*(1,)]': [
('RESUME', 0),
('BUILD_LIST', 0),
('LOAD_CONST', (1,)),
('LIST_EXTEND', 1),
('RETURN_VALUE', None),
],
'[*(x,)]': [
('RESUME', 0),
('BUILD_LIST', 0),
('LOAD_NAME', 'x'),
('BUILD_TUPLE', 1),
('LIST_EXTEND', 1),
('RETURN_VALUE', None),
],
}

for source, expected in cases.items():
with self.subTest(source=source):
code = compile(source, '<test>', 'eval')
instructions = [
(instruction.opname, instruction.argval)
for instruction in dis.get_instructions(code)
]
self.assertEqual(instructions, expected)

def test_imported_load_method(self):
sources = [
"""\
Expand Down
1 change: 1 addition & 0 deletions Misc/NEWS.d/next/Core_and_Builtins/2026-06-02-16-21-02.gh-issue-150737.LoYUFY.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
Optimize bytecode for empty unpack cases such as ``{*()}``.
42 changes: 35 additions & 7 deletions Python/codegen.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3412,24 +3412,46 @@ codegen_boolop(compiler *c, expr_ty e)
return SUCCESS;
}

static bool
is_empty_starred_literal(expr_ty elt)
{
if (elt->kind != Starred_kind) {
return false;
}
expr_ty value = elt->v.Starred.value;
return (value->kind == Tuple_kind &&
asdl_seq_LEN(value->v.Tuple.elts) == 0) ||
(value->kind == List_kind &&
asdl_seq_LEN(value->v.List.elts) == 0) ||
(value->kind == Dict_kind &&
asdl_seq_LEN(value->v.Dict.keys) == 0);
}

static int
starunpack_helper_impl(compiler *c, location loc,
asdl_expr_seq *elts, PyObject *injected_arg, int pushed,
int build, int add, int extend, int tuple)
{
Py_ssize_t n = asdl_seq_LEN(elts);
int big = n + pushed + (injected_arg ? 1 : 0) > _PY_STACK_USE_GUIDELINE;
Py_ssize_t end = asdl_seq_LEN(elts);
Py_ssize_t n = 0;
int seen_star = 0;
for (Py_ssize_t i = 0; i < n; i++) {
for (Py_ssize_t i = 0; i < end; i++) {
expr_ty elt = asdl_seq_GET(elts, i);
if (elt->kind == Starred_kind) {
if (is_empty_starred_literal(elt)) {
continue;
}
seen_star = 1;
break;
}
n++;
}
int big = n + pushed + (injected_arg ? 1 : 0) > _PY_STACK_USE_GUIDELINE;
if (!seen_star && !big) {
for (Py_ssize_t i = 0; i < n; i++) {
for (Py_ssize_t i = 0; i < end; i++) {
expr_ty elt = asdl_seq_GET(elts, i);
if (is_empty_starred_literal(elt)) {
continue;
}
VISIT(c, expr, elt);
}
if (injected_arg) {
Expand All @@ -3444,15 +3466,20 @@ starunpack_helper_impl(compiler *c, location loc,
return SUCCESS;
}
int sequence_built = 0;
Py_ssize_t nitems = 0;
if (big) {
ADDOP_I(c, loc, build, pushed);
sequence_built = 1;
}
for (Py_ssize_t i = 0; i < n; i++) {
for (Py_ssize_t i = 0; i < end; i++) {
expr_ty elt = asdl_seq_GET(elts, i);

if (elt->kind == Starred_kind) {
if (is_empty_starred_literal(elt)) {
continue;
}
if (sequence_built == 0) {
ADDOP_I(c, loc, build, i+pushed);
ADDOP_I(c, loc, build, nitems+pushed);
sequence_built = 1;
}
VISIT(c, expr, elt->v.Starred.value);
Expand All @@ -3464,6 +3491,7 @@ starunpack_helper_impl(compiler *c, location loc,
ADDOP_I(c, loc, add, 1);
}
}
nitems++;
}
assert(sequence_built);
if (injected_arg) {
Expand Down
Loading