BUG: diff with multiple columns of type np.int8 fails

[simonm3]

df = pd.DataFrame(np.random.randint(9, size=(4,5)), columns=list("abcde"))
df = df.astype(np.int8)
print(df.astype(int).diff())
print(df.a.diff())
df.diff()

diff works fine on a dataframe with int columns; or on a single column of type np.int8. However if there are multiple columns of np.int8 then it fails

     a    b    c    d    e
0  NaN  NaN  NaN  NaN  NaN
1  6.0 -5.0 -7.0 -7.0 -1.0
2 -5.0  7.0 -1.0  5.0 -7.0
3  2.0 -6.0  8.0  0.0  3.0
0    NaN
1    6.0
2   -5.0
3    2.0
Name: a, dtype: float64
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-145-2724544318c3> in <module>()
      3 print(df.astype(int).diff())
      4 print(df.a.diff())
----> 5 df.diff()

C:\Users\s\Anaconda3\lib\site-packages\pandas\core\frame.py in diff(self, periods, axis)
   4068         """
   4069         bm_axis = self._get_block_manager_axis(axis)
-> 4070         new_data = self._data.diff(n=periods, axis=bm_axis)
   4071         return self._constructor(new_data)
   4072 

C:\Users\s\Anaconda3\lib\site-packages\pandas\core\internals.py in diff(self, **kwargs)
   3172 
   3173     def diff(self, **kwargs):
-> 3174         return self.apply('diff', **kwargs)
   3175 
   3176     def interpolate(self, **kwargs):

C:\Users\s\Anaconda3\lib\site-packages\pandas\core\internals.py in apply(self, f, axes, filter, do_integrity_check, consolidate, **kwargs)
   3054 
   3055             kwargs['mgr'] = self
-> 3056             applied = getattr(b, f)(**kwargs)
   3057             result_blocks = _extend_blocks(applied, result_blocks)
   3058 

C:\Users\s\Anaconda3\lib\site-packages\pandas\core\internals.py in diff(self, n, axis, mgr)
   1040     def diff(self, n, axis=1, mgr=None):
   1041         """ return block for the diff of the values """
-> 1042         new_values = algos.diff(self.values, n, axis=axis)
   1043         return [self.make_block(values=new_values, fastpath=True)]
   1044 

C:\Users\s\Anaconda3\lib\site-packages\pandas\core\algorithms.py in diff(arr, n, axis)
   1196     if arr.ndim == 2 and arr.dtype.name in _diff_special:
   1197         f = _diff_special[arr.dtype.name]
-> 1198         f(arr, out_arr, n, axis)
   1199     else:
   1200         res_indexer = [slice(None)] * arr.ndim

pandas\src\algos_common_helper.pxi in pandas.algos.diff_2d_int8 (pandas\algos.c:62940)()

ValueError: Buffer dtype mismatch, expected 'float32_t' but got 'double'

######################################################################
INSTALLED VERSIONS

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 78 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.19.1
nose: 1.3.7
pip: 9.0.1
setuptools: 28.7.1
Cython: 0.24
numpy: 1.11.2
scipy: 0.17.1
statsmodels: 0.8.0rc1
xarray: None
IPython: 4.2.0
sphinx: 1.3.1
patsy: 0.4.1
dateutil: 2.5.3
pytz: 2016.7
blosc: None
bottleneck: 1.1.0
tables: 3.2.2
numexpr: 2.6.0
matplotlib: 1.5.3
openpyxl: 2.3.2
xlrd: 1.0.0
xlwt: 1.1.2
xlsxwriter: 0.9.2
lxml: 3.6.0
bs4: 4.4.1
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: 1.0.13
pymysql: None
psycopg2: None
jinja2: 2.8
boto: 2.40.0
pandas_datareader: None
time: 5.55 s

[chris-b1]

Yep, looks buggy, PR to fix welcome.

[chris-b1]

xref #4899

[kidpixo]

Just found this bug in my code.

This happens for np.int8, np.int16 but not for np.uint8,np.uint16 and np.int32.

pd.Series(data=[1,0,1], dtype=np.int8).diff()
pd.Series(data=[1,0,1], dtype=np.int16).diff()
# same output 
*** NotImplementedError
Traceback (most recent call last):
  File "[...]/lib/python3.8/site-packages/pandas/core/series.py", line 2438, in diff
    result = algorithms.diff(self.array, periods)
  File "[...]/lib/python3.8/site-packages/pandas/core/algorithms.py", line 2002, in diff
    algos.diff_2d(arr, out_arr, n, axis, datetimelike=is_timedelta)
  File "pandas/_libs/algos.pyx", line 1211, in pandas._libs.algos.diff_2d

[mroeschke]

This looks to work on master now. Could use a test

In [32]: df = pd.DataFrame(np.random.randint(9, size=(4,5)), columns=list("abcde"))
    ...: df = df.astype(np.int8)

In [33]: df.diff()
Out[33]:
     a    b    c    d    e
0  NaN  NaN  NaN  NaN  NaN
1  3.0 -2.0  3.0  4.0 -6.0
2  3.0  4.0 -2.0 -4.0  2.0
3  0.0 -6.0  0.0 -1.0  6.0

In [34]: pd.__version__
Out[34]: '1.3.0.dev0+1485.g6abb567cb1'