Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #105527 > unrolled thread

0 equals False, was Re: (unknown)

Back to article view | Back to comp.lang.python

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  0 equals False, was Re: (unknown) Peter Otten <__peter__@web.de> - 2016-03-23 09:51 +0100
    Re: 0 equals False, was Re: (unknown) gst <g.starck@gmail.com> - 2016-03-23 08:03 -0700

#105527 — 0 equals False, was Re: (unknown)

Nick Eubank wrote:

> Hello All,
> 
> 
> Found an odd behavior I'd never known about today, not sure if it's a bug
> or known. Python 3.4.4 (anaconda).

This is a feature. Old versions of Python did not have True and False, so 
they were added in a compatible way.

> True, False, 0, 1 can all be used as dictionary keys.
> 
> But Apparently True and 1 hash to the same item and False and 0 hash to
> the same item, so they can easily overwrite (which I spent a while banging
> my head over today).
> 
> In other words:
> 
>  In[1]:
>      d = {True: 'a', False: 'b'}
>      d[0] = 'z'
>      d[False]
> 
> Out[1]:
>      'z'
> 
> I understand that True and False are sub-types of ints, but it's not clear
> to me why (i.e. certainly didn't feel intuitive) that they would be
> treated the same as keys.
> 
> Relatedly, if this is a desired behavior, any advice one how best to work
> with dictionaries when one wants "True" and 1 to be different? I'm working
> on a function that accepts arguments that may be "True" or 1 (meaning very
> different things) and am seeking a pythonic solution...

The pythonic solution is "don't do this". The == operator cannot 
discriminate between 0, 0.0, and False, or 1, 1.0, and True. 

True and False are singletons, so you can check identity with

x is True or x is False

A type check will also work: 

type(x) == bool
isinstance(x, bool) # bool cannot be subclassed

If you provide some context we may be able to come up with an alternative 
approach that fits your use case.

[toc] | [next] | [standalone]

#105554

Le mercredi 23 mars 2016 04:52:02 UTC-4, Peter Otten a écrit :
> Nick Eubank wrote:
> 
> > Hello All,
> > 
> > 
> > Found an odd behavior I'd never known about today, not sure if it's a bug
> > or known. Python 3.4.4 (anaconda). 
> > True, False, 0, 1 can all be used as dictionary keys.
> > 
> > But Apparently True and 1 hash to the same item and False and 0 hash to
> > the same item, so they can easily overwrite (which I spent a while banging
> > my head over today).
> > 
> > In other words:
> > 
> >  In[1]:
> >      d = {True: 'a', False: 'b'}
> >      d[0] = 'z'
> >      d[False]
> > 
> > Out[1]:
> >      'z'
> > 
> > I understand that True and False are sub-types of ints, but it's not clear
> > to me why (i.e. certainly didn't feel intuitive) that they would be
> > treated the same as keys.
> > 
> > Relatedly, if this is a desired behavior, any advice one how best to work
> > with dictionaries when one wants "True" and 1 to be different? I'm working
> > on a function that accepts arguments that may be "True" or 1 (meaning very
> > different things) and am seeking a pythonic solution...


I would include the type in the dictionary key:

d = {}

x = True

d[(type(x), x)] = 42

x = 1

d[(type(x), x)] = "foo"

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web