Calling searchsorted on an int haystack with str needle converts haystack to str

[levitsky]

With numpy 1.9.2 I see quite counter-intuitive behavior of searchsorted on integer arrays with a str needle:

In [1]: import numpy as np

In [2]: a = np.arange(1000, dtype=int)

In [3]: a.searchsorted('15')
Out[3]: 150

In [4]: a.searchsorted('150')
Out[4]: 150

In [5]: a.searchsorted('1500')
Out[5]: 151

In [6]: a.searchsorted('foo')
Out[6]: 1000

The main reason why I believe this behavior is broken is because when one converts an integer array to strings, it's not sorted anymore.
I believe that the least surprising result would be an exception, especially because this is what happens if the array contains floats and not integers.

P.S. I initially asked about this on StackOverflow, and @jaimefrio suggested creating a bug report in the comments.

[jaimefrio]

I think fixing this boils down to whether it is a searchsorted bug, and we need to put a special check there, or it really is a can_cast bug, and we should disallow safe casting of ints to strings, as we already do for floats.

[njsmith]

+1 to making int->str considered not-safe. Generally "safe" for us means
"should be implicitly coerceable", and python in general does not allow
implicit coercion of int->str, so this seems pretty correct to me.
On Jul 10, 2015 12:33 PM, "Jaime" notifications@github.com wrote:

I think fixing this boils down to whether it is a searchsorted bug, and
we need to put a special check there, or it really is a can_cast bug, and
we should disallow safe casting of ints to strings, as we already do for
floats.

—
Reply to this email directly or view it on GitHub
#6061 (comment).

[OmerJog]

Is there a fix for this?