BUG: index_col in read_csv ignores dtype argument

[seisman]

Pandas version checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of pandas.
• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import io
import numpy as np
import pandas as pd

data = io.StringIO("345.5 519.5 0\n519.5 726.5 1\n")


df = pd.read_csv(
    data,
    sep=r"\s+",
    header=None,
    names=["start", "stop", "bin_id"],
    dtype={"start": np.float32, "stop": np.float32, "bin_id": np.uint32},
    index_col="bin_id"
)
print(df.index.dtype)

Issue Description

df.index.dtype is int64 with pandas 3.0.0.dev0+1132.ga5e812d86d, although the dtype parameter already sets it to np.uint32. The issue is similar to an old issue #9435.

Expected Behavior

Pandas 2.x correctly returns uint32, which is the expected behavior.

Installed Versions

``` INSTALLED VERSIONS ------------------ commit : a5e812d python : 3.12.3.final.0 python-bits : 64 OS : Linux OS-release : 6.8.11-300.fc40.x86_64 Version : #1 SMP PREEMPT_DYNAMIC Mon May 27 14:53:33 UTC 2024 machine : x86_64 processor : byteorder : little LC_ALL : None LANG : en_US.UTF-8 LOCALE : en_US.UTF-8

pandas : 3.0.0.dev0+1132.ga5e812d86d
numpy : 1.26.4
pytz : 2024.1
dateutil : 2.9.0.post0
setuptools : 70.0.0
pip : 24.0
Cython : None
pytest : 8.2.1
hypothesis : None
sphinx : 7.3.7
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.4
IPython : 8.25.0.dev
pandas_datareader : None
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : None
bottleneck : None
fastparquet : None
fsspec : 2024.5.0
gcsfs : None
matplotlib : 3.10.0.dev202+gd901275d7c
numba : None
numexpr : None
odfpy : None
openpyxl : None
pyarrow : 16.1.0
pyreadstat : None
python-calamine : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : 0.9.0
xarray : 2024.5.1.dev6+g12123be8
xlrd : None
zstandard : None
tzdata : 2024.1
qtpy : None
pyqt5 : None

</details>

[xouyang1]

take

[xouyang1]

Seems to be a behavior change for the default c engine only.
python engine is also incorrect (returning int64) but was also doing that in v2.2.2

[MarcoGorelli]

thanks for the report - we should probably first do a git bisect to see which commit introduced the bug

running it now: https://www.kaggle.com/code/marcogorelli/pandas-regression-example?scriptVersionId=187955574

[xouyang1]

@MarcoGorelli Hi Marco, I tried rerunning the regression example with just data = io.StringIO("345.5 519.5 0) and it always lands at the latest commit, which is incorrect. Can you help take a look?

[MarcoGorelli]

thanks @xouyang1 - I was missing an escape character in the script, and so it wasn't running properly

having tried again, I'm getting #57943 as the commit that introduced it, but I don't know if that seems reasonable, haven't looked closer yet

[xouyang1]

thanks @xouyang1 - I was missing an escape character in the script, and so it wasn't running properly

having tried again, I'm getting #57943 as the commit that introduced it, but I don't know if that seems reasonable, haven't looked closer yet

It looks like the right commit, specifically this change to pandas/core/indexes/base.py ded256d#diff-c34a28314fc8cb12f0d2aa710f1c15f06cdfe3e48f03e658f01f99a43d4f5d09

• Before: does not convert to range (np_sequence.dtype.kind is “u” for uint so returns the sequence as is)

  pandas/pandas/core/indexes/base.py

  Line 7164 in cf40e56

  return sequence

• After: converts to range, which drops the dtype association

  pandas/pandas/core/indexes/base.py

  Line 7170 in ded256d

  return range(sequence[0], sequence[-1] + diff, diff)

[MarcoGorelli]

It looks like the right commit, specifically this change to pandas/core/indexes/base.py ded256d#diff-c34a28314fc8cb12f0d2aa710f1c15f06cdfe3e48f03e658f01f99a43d4f5d09

cool, thanks for checking (cc @mroeschke just fyi, no blame 🤗 )