[Numpy-discussion] Deprecate np.max/np.min ?

josef.pktd at gmail.com josef.pktd at gmail.com
Fri Nov 6 14:01:02 EST 2009

On Fri, Nov 6, 2009 at 1:22 PM, Pauli Virtanen <pav at iki.fi> wrote:
> pe, 2009-11-06 kello 12:57 -0500, josef.pktd at gmail.com kirjoitti:
> [clip]
>> It follows the same pattern for missing brackets as many of the other
>> functions, where I often have typos.
>> >>> np.min(3,2)
>> 3
>> >>> np.sum(3,2)
>> 3
>> >>> np.ones(3,2)
> [clip]
> I think you are thinking about Matlab now...

I'm using (almost) as much matlab as python, so some confusion and
typos show up very regularly.

> The issue here is that Numpy's `min` and `max`, if used like their
> Python counterparts, give surprising results. `sum`, `ones`, `round`,
> etc. don't have this problem. The prepended `a` in the beginning is IMO
> not a big price to pay for reduced overloading.
> OTOH, one can ask, why is
>        np.min(3, 2)
> allowed when
>        np.min([3], 2)
> gives "ValueError: axis(=2) out of bounds". It seems to me that
> 0-dimensional objects should accept only None as the axis? (Fixing this
> would also make misuse of np.min and np.max more difficult.)

That was also my first reaction, I think this axis check would be a
good improvement. I wanted to say we could improve the documentation
for numpy.min and add a warning,  but I don't find the function min in
the docs (only amin and the method)

In general, I think shadowing python's built-in functions is a bad
idea and the missing namespaces make some code difficult to read and
makes it difficult to find problems.

I could always do    import numpy.amin as npmin
which would be only a change by one dot.

But still breaking a lot of code, just for some cosmetic changes ?
What about np.ma.min, np.recarray.min (raises exception)? (matrix.min
is only available as method)


> --
> Pauli Virtanen
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion at scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion

More information about the NumPy-Discussion mailing list