PERF: RangeIndex cache is written when calling df.loc[array]

[GianlucaFicarelli]

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

The cache used by RangeIndex is written when accessing the _data or _values properties.

This doesn't happen when calling df.loc[k] and k is a scalar.
However, this happens when calling df.loc[a] and a is an array.

This isn't usually a big issue, but if the RangeIndex is big enough, it's possible to see that the first access is slower than expected.
I think that in general the creation of the RangeIndex cache should be avoided unless needed.

Example:

import numpy as np
import pandas as pd

df = pd.DataFrame(index=pd.RangeIndex(400_000_000))

%time x = df.loc[200_000_000]  # fast
print("_data" in df.index._cache)  # False
%time x = df.loc[[200_000_000]]  # slow on the first execution
print("_data" in df.index._cache)  # True
%time x = df.loc[[200_000_000]]  # fast after the first execution

Result:

CPU times: user 44 µs, sys: 73 µs, total: 117 µs
Wall time: 122 µs
False
CPU times: user 221 ms, sys: 595 ms, total: 815 ms
Wall time: 815 ms
True
CPU times: user 356 µs, sys: 165 µs, total: 521 µs
Wall time: 530 µs

I can open a PR with a possible approach to calculate the result without loading the cache, that should have better performance on the first access, and similar performance after that.

Installed Versions

INSTALLED VERSIONS

commit : 37ea63d
python : 3.10.8.final.0
python-bits : 64
OS : Linux
OS-release : 3.10.0-1160.42.2.el7.x86_64
Version : #1 SMP Tue Aug 31 20:15:00 UTC 2021
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : en_US.UTF-8
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.0.1
numpy : 1.24.3
pytz : 2023.3
dateutil : 2.8.2
setuptools : 67.7.2
pip : 23.1.2
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 8.13.2
pandas_datareader: None
bs4 : None
bottleneck : None
brotli : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.7.1
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.10.1
snappy : None
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.3
qtpy : None
pyqt5 : None

Prior Performance

No response

[jbrockmendel]

I can open a PR with a possible approach to calculate the result without loading the cache, that should have better performance on the first access, and similar performance after that.

I'd be interested to see this.

[GianlucaFicarelli]

take