Unexpected behaviour of uint dtype

[nvm1]

Code Sample, a copy-pastable example if possible

import pandas as pd
# generate some data and create dataframe
col = {'x': []}
for i in range(10):
    col['x'].append(18446744073709500000 - i)
df = pd.DataFrame.from_dict(col, dtype='uint64')
# see the dtype:
print(df.dtypes, end='\n\n')
# see correct and expected behaviour:
print(df.x * -1, end='\n\n')
# see possibly inconsistent behaviour:
print(-df.x, end='\n\n')

Problem description

please see the example: in two cases which are expected (sorry if the expectation is wrong) to generate the same results we see huge difference. In our case we didn't check the type of a column and assigned it to itself but with a minus value.

Expected Output

Probably the expected behaviour should match the logic here: #15832

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.7.3.final.0
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.24.2
pytest: 4.3.1
pip: 19.0.3
setuptools: 40.8.0
Cython: 0.29.6
numpy: 1.16.2
scipy: 1.2.1
pyarrow: None
xarray: None
IPython: 7.4.0
sphinx: 1.8.5
patsy: 0.5.1
dateutil: 2.8.0
pytz: 2018.9
blosc: None
bottleneck: 1.2.1
tables: 3.5.1
numexpr: 2.6.9
feather: None
matplotlib: 3.0.3
openpyxl: 2.6.1
xlrd: 1.2.0
xlwt: 1.3.0
xlsxwriter: 1.1.5
lxml.etree: 4.3.2
bs4: 4.7.1
html5lib: 1.0.1
sqlalchemy: 1.3.1
pymysql: None
psycopg2: 2.7.6.1 (dt dec pq3 ext lo64)
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None
gcsfs: None
None

[jorisvandenbossche]

This seems to be an issue in numpy as well (so I suppose we inherit this behaviour):

In [21]: np.array([1, 2, 3], dtype='uint64') * -1 
Out[21]: array([-1., -2., -3.])

In [22]: -np.array([1, 2, 3], dtype='uint64') 
Out[22]: 
array([18446744073709551615, 18446744073709551614, 18446744073709551613],
      dtype=uint64)

Would you like to open an issue with numpy?

[R1j1t]

The reason for the dtype change I think is because uint cant take negative values, so numpy library handles multiply by negitive (x*-1) request by changing the dtype to float but does not change dtype for (-x). To convince myself I tried the following:

For uint dtype

>>> a=np.array([1, 2, 3], dtype='uint64')
>>> a.dtype
dtype('uint64')
>>> -a
array([18446744073709551615, 18446744073709551614, 18446744073709551613],
      dtype=uint64)
>>> a=a*-1
>>> a.dtype
dtype('float64')
>>> a
array([-1., -2., -3.])

for int dtype

>>> b=np.array([1, 2, 3], dtype='int64')
>>> b.dtype
dtype('int64')
>>> -b
array([-1, -2, -3])
>>> b=b*-1
>>> b.dtype
dtype('int64')
>>> b
array([-1, -2, -3])

[mroeschke]

Appears to be a numpy issue as discussed. Closing as an upstream issue.