Skip to content

fix : skip memoizing 0,1,True,False in sanitization function #62226

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

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

fix : skip memoizing 0,1,True,False in sanitization function #62226

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
6047584
fix : remove memo usage
vignesh14052002 Aug 31, 2025
6eb738a
fix linting
vignesh14052002 Aug 31, 2025
3078807
include type in memo key to handle 0,1,True and False conflict
vignesh14052002 Sep 3, 2025
2aa4a01
skip memoization for 0,1,True,False
vignesh14052002 Sep 11, 2025
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
7 changes: 4 additions & 3 deletions pandas/_libs/parsers.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2129,12 +2129,13 @@ def sanitize_objects(ndarray[object] values, set na_values) -> int:

for i in range(n):
val = values[i]
memo_key = (val, type(val))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Perf impact? I suspect this hashing is slower

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

measured performance by sanitizing 5 million values

import time
import numpy as np
import contextlib

@contextlib.contextmanager
def log_duration(title):
    start = time.perf_counter()
    try:
        yield
    finally:
        end = time.perf_counter()
        print(f"{title}: {end - start:.4f} seconds")

million = 10**6

def get_synthetic_data():
    int_values = np.array([i for i in range(million)], dtype=object)
    float_values = np.array([float(i) for i in range(million)], dtype=object)
    str_values = np.array([str(i) for i in range(million)], dtype=object)
    bool_values = np.array([i % 2 == 0 for i in range(million)], dtype=object)
    none_values = np.array(["NA" for _ in range(million)], dtype=object)

    mixed_values = np.empty(million * 5, dtype=object)
    mixed_values[0::5] = int_values
    mixed_values[1::5] = float_values
    mixed_values[2::5] = str_values
    mixed_values[3::5] = bool_values
    mixed_values[4::5] = none_values
    np.random.seed(42)
    np.random.shuffle(mixed_values)
    return mixed_values

values = get_synthetic_data()

with log_duration("sanitize_objects_old"):
    sanitize_objects_old(values,na_values={"NA"})

values = get_synthetic_data()

with log_duration("sanitize_objects_include_type_in_memo_key"):
    sanitize_objects(values,na_values={"NA"})

values = get_synthetic_data()

with log_duration("sanitize_objects_skip_bool"):
    sanitize_objects_skip_bool(values,na_values={"NA"})

Output

sanitize_objects_old: 1.5880 seconds
sanitize_objects_include_type_in_memo_key: 1.9344 seconds
sanitize_objects_skip_bool: 1.6926 seconds

yes you are right, using type in memo key is 20% slower, skipping those 4 values seems a better option, python automatically uses references for same int and bool values, so skipping will not increase memory footprint even if the data is mixture of only those 4 values

if val in na_values:
values[i] = onan
na_count += 1
elif val in memo:
values[i] = memo[val]
elif memo_key in memo:
values[i] = memo[memo_key]
else:
memo[val] = val
memo[memo_key] = val

return na_count
11 changes: 11 additions & 0 deletions pandas/tests/io/parser/common/test_common_basic.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@

from pandas._config import using_string_dtype

from pandas._libs import parsers as libparsers
from pandas.compat import HAS_PYARROW
from pandas.errors import (
EmptyDataError,
Expand Down Expand Up @@ -830,3 +831,13 @@ def test_read_seek(all_parsers):
actual = parser.read_csv(file)
expected = parser.read_csv(StringIO(content))
tm.assert_frame_equal(actual, expected)


def test_dtype_conversion_in_sanitization():
# GH60088
values = np.array([1, True], dtype=object)
expected = np.array([1, True], dtype=object)
libparsers.sanitize_objects(values, na_values=set())
for v, e in zip(values, expected):
assert v == e
assert type(v) == type(e)
Loading