[Numpy-discussion] Inconsistent error messages.

[Eric Firing]

Charles R Harris wrote:
> 
> 
> On Sat, May 23, 2009 at 6:03 PM, Robert Kern <robert.kern@gmail.com 
> <mailto:robert.kern@gmail.com>> wrote:
> 
>     On Sat, May 23, 2009 at 18:57, Charles R Harris
>     <charlesr.harris@gmail.com <mailto:charlesr.harris@gmail.com>> wrote:
>      > You were supposed to be able to change the default behaviour, but
>     it didn't
>      > used to work. I think if you are going to use a warning as a flag
>     then it
>      > has to always be raised when a failure occurs, not just the first
>     time.
> 
>     A brief test suggest that in Python 2.5.4, at least, as long as you
>     set the action to be 'always' before the warning is first issued, it
>     works. We can do this just after the IOWarning (or whatever) gets
>     defined.
> 
> 
> OK, that would work. Although I think a named argument might be a more 
> transparent way to specify behaviour than setting the warnings.

I agree; using a warning strikes me as an abuse of the warnings 
mechanism.  Instead of a "strict" flag, which I find not particularly 
expressive--what is it being "strict" about?--how about a "min_count" 
kwarg to go with the existing "count" kwarg?

min_count=None # default; raise ValueError instead of the present 
warning if fewer than count are found.

min_count=0 # Accept whatever you get; no warning, no error.
min_count=N # raise ValueError if fewer than N are found.

This is more flexible than using a "strict" flag, and the kwarg is more 
descriptive.

Eric


> 
> Chuck
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Numpy-discussion mailing list
> Numpy-discussion@scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion