PERF: bottleneck in `where()`

[auderson]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this issue exists on the latest version of pandas.

• I have confirmed this issue exists on the main branch of pandas.

Reproducible Example

import numpy as np
import pandas as pd

df = pd.DataFrame(np.random.randn(1, 1_000_000))

mask = df > 0.5

%%timeit
_ = df.where(mask)
# 693 ms ± 3.49 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

perf result taken from pyinstrument:

This issue seems to be related to this:

pandas/pandas/core/generic.py

Lines 9735 to 9737 in d1ec1a4

	for _dt in cond.dtypes:
	if not is_bool_dtype(_dt):
	raise TypeError(msg.format(dtype=_dt))

When dataframe is large, this overhead of is_bool_dtype accumulates. Would it be better to use cond.dtypes.unique() instead?

Installed Versions

INSTALLED VERSIONS

commit : 0691c5c
python : 3.10.14
python-bits : 64
OS : Linux
OS-release : 5.15.0-122-generic
Version : #132-Ubuntu SMP Thu Aug 29 13:45:52 UTC 2024
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.2.3
numpy : 1.26.4
pytz : 2024.1
dateutil : 2.9.0
pip : 24.0
Cython : 3.0.7
sphinx : 7.3.7
IPython : 8.25.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.12.3
blosc : None
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : 2024.6.0
html5lib : None
hypothesis : None
gcsfs : None
jinja2 : 3.1.4
lxml.etree : None
matplotlib : 3.9.2
numba : 0.60.0
numexpr : 2.10.0
odfpy : None
openpyxl : 3.1.5
pandas_gbq : None
psycopg2 : 2.9.9
pymysql : 1.4.6
pyarrow : 16.1.0
pyreadstat : None
pytest : 8.2.2
python-calamine : None
pyxlsb : None
s3fs : None
scipy : 1.14.0
sqlalchemy : 2.0.31
tables : 3.9.2
tabulate : None
xarray : None
xlrd : None
xlsxwriter : None
zstandard : 0.22.0
tzdata : 2024.1
qtpy : 2.4.1
pyqt5 : None

Prior Performance

No response

[samukweku]

Hi @auderson do U have a benchmark U r comparing against?

[auderson]

@samukweku No, but I do find the % of time spent by is_bool_dtype increases with the num of columns.

[samukweku]

a crude benchmark, comparing against numpy:

np.random.seed(3)
df = pd.DataFrame(np.random.randn(1, 1_000_000))
mask = df > 0.5

%timeit pd.DataFrame(np.where(mask, df, np.nan), columns= df.columns)
573 μs ± 102 μs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

 %timeit df.where(mask)
324 ms ± 1.28 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

pd.DataFrame(np.where(mask, df, np.nan), columns= df.columns).equals(df.where(mask))
Out[14]: True

[samukweku]

@auderson i think there is an expense associated with .unique though, so i'm not sure it would offer better performance. maybe you could test it locally?

[auderson]

@samukweku

From my local machine the performance increases by 10x if using .dtypes.unique()

But a better solution is to get the dtypes of each block, instead of dtypes of each column, then .dtypes.unique() will not be needed. e.g. [blk.dtype for blk in df._mgr.blocks]

I'm not familiar with pandas internals so not sure if the above usage is OK