Possible bug in np.min, np.max, np.ndarray.min, np.ndarray.max

When using np.min or np.max by setting the value of initial to np.inf and -np.inf, respectively, the output is not what was expected. Perhaps I’m missing something, sorry if this is not actually a bug.

Reproducing code example:

In[44]: np.array([-10,20,30,-25]).min(initial=np.inf)
Out[44]: -9223372036854775808
In[45]: np.array([-10,20,30,-25]).min(initial=-np.inf)
Out[45]: -9223372036854775808
In[46]: np.array([-10,20,30,-25]).max(initial=np.inf)
Out[46]: 30
In[47]: np.array([-10,20,30,-25]).max(initial=-np.inf)
Out[47]: 30
In[48]: np.min([-10,20,30,-25], initial=np.inf)
Out[48]: -9223372036854775808
In[49]: np.min([-10,20,30,-25], initial=-np.inf)
Out[49]: -9223372036854775808
In[50]: np.max([-10,20,30,-25], initial=np.inf)
Out[50]: 30
In[51]: np.max([-10,20,30,-25], initial=-np.inf)
Out[51]: 30

Numpy/Python version information:

1.18.1 3.7.6 | packaged by conda-forge | (default, Mar 23 2020, 23:03:20) [GCC 7.3.0]

[mattip]

I think the initial should error if the value is not representable as the dtype used for computation.

[seberg]

I changed this and now it raises an error. If you were to pass np.float64(np.inf) you would at the very least get an “invalid value” warning (not quite sure, didn’t try). So closing.