How to diff int64's with pandas.diff

Question

>>> pd.DataFrame([3,4,5], dtype='int64').diff()
     0
0  NaN
1  1.0
2  1.0

I would have hoped to not convert datatypes two times, is it possible?

Answer 1

Yes, if you need to, you can use np.diff -

numpy.diff(a, n=1, axis=-1)

Calculate the n-th discrete difference along given axis.

The first difference is given by out[n] = a[n+1] - a[n] along the given axis, higher differences are calculated by using diff recursively.

>>> df
   A  B
0  1  4
1  2  5
2  3  6

>>> np.diff(df.values, axis=0)
array([[1, 1],
       [1, 1]])

# or as in your example:
>>> np.diff(pd.DataFrame([3,4,5], dtype='int64')[0])
array([1, 1], dtype=int64)

Note that you cannot assign this back to the original dataframe without introducing NaNs, and NaNs are of float type, meaning your result will automatically be coerced to floats.