pd.to_numeric(..., errors="coerce") failing silently when strings contain "uint64"

[buhrmann]

Problem description

When trying to coerce strings to numeric values using to_numeric(), the occurrence of the substring "uint64" (but not any other dtype-like substring it seems) leads to silent failure to coerce.

strs = ["32", "64", "uint32", "float64", "sdnfonsdf uint32 knsdf", "sdnfonsdf uint64 knsdf", "uint64"]
print([pd.to_numeric(s, errors="coerce") for s in strs])
pd.to_numeric(pd.Series(["32", "64", "uint64"]), errors="coerce")

[32, 64, nan, nan, nan, 'sdnfonsdf uint64 knsdf', 'uint64']

0        32
1        64
2    uint64
dtype: object

Expected Output

[32, 64, nan, nan, nan, nan, nan]

0        32.0
1        64.0
2        NaN
dtype: float64

Seems to fail equally in 0.25.3 and 1.0...

Output of pd.show_versions()

INSTALLED VERSIONS

commit : None

pandas : 0.25.3
numpy : 1.17.3
pytz : 2019.3
dateutil : 2.8.1
pip : 20.0.2
setuptools : 45.1.0.post20200119
Cython : None
pytest : 5.3.4
hypothesis : None
sphinx : 2.3.1
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.4.1
html5lib : None
pymysql : 0.9.3
psycopg2 : 2.8.4 (dt dec pq3 ext lo64)
jinja2 : 2.10.3
IPython : 7.11.1
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : 4.4.1
matplotlib : 3.1.2
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 0.14.1
pytables : None
s3fs : None
scipy : 1.4.1
sqlalchemy : 1.3.12
tables : None
xarray : None
xlrd : 1.2.0
xlwt : None
xlsxwriter : None

[TomAugspurger]

A bit simpler reproducer:

In [9]: b = np.array(['uint64'], dtype=object)

In [10]: pd._libs.lib.maybe_convert_numeric(b, set(), coerce_numeric=True)

We explicitly check for uint64 in the error message at

pandas/pandas/_libs/lib.pyx

Line 2029 in f25ed6f

elif "uint64" in str(err): # Exception from check functions.

, which goes back a ways (at least before #22219, didn't look at the blame past that).

It'd be great if you could do more investigation here.

[roberthdevries]

take

[roberthdevries]

This seems to be a relic of really old type inference mechanics which were present in pandas/src/inference.pyx (commit 17d7ddb by gfyoung).
That code used some check functions for uint64 types (now removed) that would raise exceptions with the string "uint64". As this code is now no longer present, the check for the presence of the string in the exception text can be removed as well.
I did not spend time to figure out in which revision the checks were removed without removing the string check.

[jreback]

reopening as going to revert this -
needs more tests