BUG: PythonParser does not use decimal separator when usecols and parse_date are specified

[cmulders]

• 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.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

Repl: https://repl.it/repls/TiredLightcyanNetworking#main.py

import pandas as pd
import io

data = io.StringIO('"dump","-9,1",20101010')

read_options = {
  'names': ['col', 'col1', 'col2'],
  'usecols': ['col1', 'col2'],
  'parse_dates': ["col2"],
  'decimal': ",",
}

data.seek(0)
c_engine = pd.read_csv(
  data,
  engine="c",
  **read_options,
)

print(c_engine['col1'].dtype) # float64

data.seek(0)
python_engine = pd.read_csv(
  data,
  engine="python",
  **read_options,
)

print(python_engine['col1'].dtype) # object instead of float64

pd.testing.assert_frame_equal(c_engine, python_engine)
# Raises
# Attribute "dtype" are different
# [left]:  float64
# [right]: object

Problem description

The PythonParser produces an invalid result when using a combination of options. It does not convert the comma correctly when the options 'usecols' and 'decimal' are used. The C engine does produce a correct result.

It seems that this check is wrong, the columns are shifted:
https://github.com/pandas-dev/pandas/blob/v1.1.1/pandas/io/parsers.py#L3089

The columns specified in that list are adjusted for the 'usecols' option (so shifted by 1 in the example), while the read line still has all columns.

Expected Output

Expect that decimals are correctly converted for all columns. However, the python engine does not when there are columns specified for date parsing.

Output of pd.show_versions()

INSTALLED VERSIONS

commit : f2ca0a2
python : 3.8.3.final.0
python-bits : 64
OS : Linux
OS-release : 5.4.0-1019-gcp
Version : #19-Ubuntu SMP Tue Jun 23 15:46:40 UTC 2020
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 : 1.1.1
numpy : 1.19.1
pytz : 2020.1
dateutil : 2.8.1
pip : 20.1.1
setuptools : 47.3.1
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : 1.1
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : None
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : None
fastparquet : None
gcsfs : None
matplotlib : 3.2.2
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : None
scipy : 1.5.0
sqlalchemy : 1.3.17
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[bashtage]

The issue appears that decimal is ignored. For example,

data = io.StringIO('"dump","-9.1",20101010')

python_engine = pd.read_csv(
  data,
  engine="python",
  **read_options,
)

print(python_engine['col1'].dtype) # float64

[cmulders]

Decimal is not ignored, otherwise the example below should not work.

Either 'engine', 'usecols', 'parse_dates' can be commented out, and both Series dtypes will be float64.

Very strange behaviour..

import pandas as pd
import io

data = io.StringIO('"dump","-9,1","-9,1",20101010')

python_engine = pd.read_csv(
  data,
  engine="python", # Comment out and both are float64
  names= ['col', 'col1', 'col2', 'col3'],
  usecols= ['col1', 'col2', 'col3'], # Comment out and both are float64
  parse_dates= ["col3"], # Comment out and both are float64
  decimal= ",",
)

print(python_engine['col1'].dtype) # float64
print(python_engine['col2'].dtype) # object

[bashtage]

When decimal="," then "9.1" should be a string (object) while "9,1" should be a float. In the testing I did above, it seemed like "9.1" was being converted to float even when decimal=",".