groupby.agg (first, last, min, etc...) returns incorrect results for uint64 columns

[toliwaga]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np

df = pd.DataFrame({'x': 6903052872240755750, 'y': [1,2]})
print(df.groupby('y').agg({'x': 'first'}))
df.x = df.x.astype(np.uint64)
print(df.groupby('y').agg({'x': 'first'}))

Problem description

groupby.agg (first, last, min, etc...) returns incorrect results for uint64 columns

Expected Output

expect same results in both cases below:
x
y
1 6903052872240755750
2 6903052872240755750
x
y
1 6903052872240755712
2 6903052872240755712

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]
INSTALLED VERSIONS

commit: None
python: 3.6.8.final.0
python-bits: 64
OS: Darwin
OS-release: 18.5.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.24.2
pytest: None
pip: 19.1
setuptools: 41.0.1
Cython: None
numpy: 1.16.3
scipy: 1.2.1
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: None
dateutil: 2.8.0
pytz: 2019.1
blosc: None
bottleneck: None
tables: 3.5.1
numexpr: 2.6.9
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml.etree: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None
gcsfs: None

[WillAyd]

Looks like there is coercion behind the scenes to float which is causing the precision loss.

Investigation and PRs would certainly be welcome

[jreback]

this is a duplicate issue / have to search for it

[mahepe]

@WillAyd is correct: It seems to happen on line 489 here:

pandas/pandas/core/groupby/ops.py

Lines 483 to 489 in 6d7ba05

	elif is_integer_dtype(values):
	# we use iNaT for the missing value on ints
	# so pre-convert to guard this condition
	if (values == iNaT).any():
	values = ensure_float64(values)
	else:
	values = ensure_int64_or_float64(values)

uint64 won't convert to int64 so it's coerced to float64.