PERF: DataFrame.query slow when it contains PeriodIndex/Array

[semio]

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

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

• (optional) I have confirmed this bug exists on the master branch of pandas.

Code Sample, a copy-pastable example

In [1]: import numpy as np

In [2]: import pandas as pd

In [23]: country = list(range(200))

In [24]: year = list(range(1600, 2100))

In [25]: idx = pd.MultiIndex.from_product([country, year], names=['country', 'year'])

In [27]: values = np.random.rand(idx.shape[0], 1)

In [30]: df = pd.DataFrame(values, index=idx, columns=['values']).reset_index()

In [41]: df2 = df.copy()

In [42]: df2['year'] = df2['year'].astype(str).apply(lambda x: pd.Period(x, 'Y'))


# performance when year column is int dtype
In [47]: %timeit df.query("year > 2020")
3.48 ms ± 80.2 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

In [48]: %timeit df[df.year > 2020]
1.07 ms ± 16.7 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

# performance when year column is Peroid dtype
In [49]: %timeit df2.query("year > '2020'")
1.03 s ± 18.3 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

In [50]: %timeit df2[df2.year > '2020']
1.46 ms ± 21.8 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

Problem description

In above example, the year column of df2 is a Series of Period objects. Running query() on the year column is a lot slower than running indexing (1.03s vs 1.46ms). Changing engine and parser for query() doesn't help.

Expected Output

I was expecting performance difference between querying and boolean indexing is similar to that when the year column's dtype is int (~3x)

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None
python : 3.8.1.final.0
python-bits : 64
OS : Darwin
OS-release : 18.7.0
machine : x86_64
processor : i386
byteorder : little
LC_ALL : en_US.UTF-8
LANG : zh_CN.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.0.5
numpy : 1.18.4
pytz : 2020.1
dateutil : 2.8.1
pip : 20.1.1
setuptools : 41.2.0
Cython : None
pytest : 5.4.2
hypothesis : 5.16.0
sphinx : 3.0.4
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.5.1
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.14.0
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : 4.5.1
matplotlib : None
numexpr : 2.7.1
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pytest : 5.4.2
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : 0.8.7
xarray : None
xlrd : 1.2.0
xlwt : None
xlsxwriter : None
numba : None

[TomAugspurger]

Do you have a solution in mind? I think that query sends ndarrays to numexper. But a PeriodArray can't be represented natively as an ndarray, since NumPy doesn't have a period dtype.

[semio]

Rigth, according to the doc numexpr is the default. I have tried engine = 'python' but it didn't help. I don't have enough knowledge about pandas internals so don't have a solution yet. But I'd like to help tackling this, will check source code when I have time.