⚡️ Speed up function fignum_exists by 7%

[codeflash-ai]

📄 7% (0.07x) speedup for fignum_exists in lib/matplotlib/pyplot.py

⏱️ Runtime : 29.1 microseconds → 27.2 microseconds (best of 28 runs)

📝 Explanation and details

The optimization replaces a costly list-based membership check with an efficient set-based lookup for string figure identifiers in fignum_exists().

Key Change:

• Added _figlabels_set() helper function that returns a set of figure labels instead of a sorted list
• Modified fignum_exists() to use _figlabels_set() for string lookups instead of get_figlabels()

Why This Is Faster:
The original code called get_figlabels() for string membership checks, which performed three expensive operations:

1. Retrieved all figure managers
2. Sorted them by number (unnecessary for membership testing)
3. Created a list and performed O(n) linear search with num in get_figlabels()

The optimized version uses _figlabels_set() which:

1. Retrieves figure managers (same cost)
2. Skips sorting entirely
3. Creates a set for O(1) average-case membership lookup

Performance Impact:

• Set creation from managers is O(n), but membership check is O(1) vs O(n) for lists
• Eliminates sorting overhead (O(n log n)) when only checking existence
• Test results show 16.9% improvement for non-existent string lookups, demonstrating the benefit for string-based figure identification

Behavior Preservation:

• Integer lookups unchanged (still use fast dict lookup in has_fignum)
• get_figlabels() maintains its sorted list contract for existing callers
• All return values and error handling remain identical

This optimization is particularly beneficial when fignum_exists() is called frequently with string identifiers, as it eliminates unnecessary sorting and improves lookup complexity from O(n) to O(1).

✅ Correctness verification report:

Test	Status
⚙️ Existing Unit Tests	✅ 62305 Passed
🌀 Generated Regression Tests	✅ 6 Passed
⏪ Replay Tests	🔘 None Found
🔎 Concolic Coverage Tests	🔘 None Found
📊 Tests Coverage	100.0%

⚙️ Existing Unit Tests and Runtime

Test File::Test Function	Original ⏱️	Optimized ⏱️	Speedup
test_backend_bases.py::test_canvas_change	1.38μs	1.21μs	14.5%✅
test_figure.py::test_fignum_exists	13.5μs	12.9μs	4.38%✅

🌀 Generated Regression Tests and Runtime

# imports
from matplotlib.pyplot import fignum_exists

# --- Function and helpers under test (copied/adapted from matplotlib source) ---


# Minimal fake Figure/Manager/Canvas classes to simulate behavior
class FakeFigure:
    def __init__(self, label):
        self._label = label

    def get_label(self):
        return self._label


class FakeCanvas:
    def __init__(self, fig):
        self.figure = fig


class FakeManager:
    def __init__(self, num, label):
        self.num = num
        self.canvas = FakeCanvas(FakeFigure(label))


# --- Basic Test Cases ---


def test_fignum_exists_with_nonexisting_int():
    # No figures at all
    codeflash_output = fignum_exists(42)  # 2.06μs -> 2.21μs (6.92% slower)


def test_fignum_exists_with_non_int_non_str():
    # Should not raise, but always False for unsupported types
    codeflash_output = fignum_exists(3.14)  # 4.69μs -> 4.00μs (17.2% faster)
    codeflash_output = fignum_exists(None)  # 1.19μs -> 1.00μs (18.3% faster)
    codeflash_output = fignum_exists((1, 2))  # 933ns -> 920ns (1.41% faster)

# imports
from matplotlib.pyplot import fignum_exists

# --- Minimal stubs and helpers to allow testing fignum_exists without matplotlib ---


# Minimal Figure and Manager classes to simulate the required interface
class DummyFigure:
    def __init__(self, label):
        self._label = label

    def get_label(self):
        return self._label


class DummyCanvas:
    def __init__(self, figure):
        self.figure = figure


class DummyManager:
    def __init__(self, num, label):
        self.num = num
        self.canvas = DummyCanvas(DummyFigure(label))


# -------------------- Basic Test Cases --------------------


def test_fignum_exists_with_non_existent_label_and_number():
    # No figures at all
    codeflash_output = fignum_exists(123)  # 2.07μs -> 2.17μs (4.48% slower)
    codeflash_output = fignum_exists("doesnotexist")  # 3.28μs -> 2.81μs (16.9% faster)

To edit these changes git checkout codeflash/optimize-fignum_exists-mjah4kid and push.