«

I love doctest -- the way it combines documentation with verification
seems elegant and useful, and most of the time it's simple and easy to
use.

But I've run into a bit of a snag trying to test a method that returns
a dictionary, because (of course) the order in which the dictionary
pairs are printed may not match what I wrote as the expected result.
For example, my doctest string is:

"""
>>> t = Template("The $object in $location falls mainly on the
$subloc.")
>>> t.match( "The rain in Spain falls mainly on the train." )
{'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
"""

But when I run it, I get:

> Failed example:
> t.match( "The rain in Spain falls mainly on the train." )
> Expected:
> {'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
> Got:
> {'subloc': 'train', 'object': 'rain', 'location': 'Spain'}

Now, you and I can see that the obtained results really do match the
expected results, considered as a dictionary rather than as a string.
But doctest doesn't see it that way.

What's the standard solution for this? Should I iterate over the
sorted keys and print those out instead? Is there some built-in
method somewhere that will print a dictionary in a reliable order?
Does doctest have some special way to tell it to consider the result
as a dictionary rather than a string? Or something else?

Thanks,
- Joe

Joe Strout:
> What's the standard solution for this?

I don't know of any standard solution, I generally sort the items in
some ways, or add the result, that has to be equal:
>>> r = foo()
>>> r == {'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
True


> Does doctest have some special way to tell it to consider the result *
> as a dictionary rather than a string? *Or something else?

A more general solution like this for sets/dics may become useful...
do you have any idea regarding how to design it?

Bye,
bearophile

On Oct 31, 2:43*pm, Joe Strout <j...@strout.net> wrote:
> I love doctest -- the way it combines documentation with verification *
> seems elegant and useful, and most of the time it's simple and easy to *
> use.
>
> But I've run into a bit of a snag trying to test a method that returns *
> a dictionary, because (of course) the order in which the dictionary *
> pairs are printed may not match what I wrote as the expected result. *
> For example, my doctest string is:
>
> * * * * """
> * * * * >>> t = Template("The $object in $location falls mainly on the *
> $subloc.")
> * * * * >>> t.match( "The rain in Spain falls mainly on the train.." )
> * * * * {'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
> * * * * """
>
> But when I run it, I get:
>
> > Failed example:
> > * * t.match( "The rain in Spain falls mainly on the train." )
> > Expected:
> > * * {'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
> > Got:
> > * * {'subloc': 'train', 'object': 'rain', 'location': 'Spain'}
>
> Now, you and I can see that the obtained results really do match the *
> expected results, considered as a dictionary rather than as a string. *
> But doctest doesn't see it that way.
>
> What's the standard solution for this? *Should I iterate over the *
> sorted keys and print those out instead? *Is there some built-in *
> method somewhere that will print a dictionary in a reliable order? *
> Does doctest have some special way to tell it to consider the result *
> as a dictionary rather than a string? *Or something else?
>
How about:

"""
>>> t = Template("The $object in $location falls mainly on the $subloc.")
>>> sorted(t.match( "The rain in Spain falls mainly on the train." ).items())
[('location', 'Spain'), ('object', 'rain'), ('subloc', 'train')]
"""

Joe Strout <> writes:
> >>> t.match( "The rain in Spain falls mainly on the train." )
> {'object': 'rain', 'location': 'Spain', 'subloc': 'train'}

You can compare dictionaries for equality:

>>> t.match( "The rain in Spain falls mainly on the train." ) == \
{'object': 'rain', 'location': 'Spain', 'subloc': 'train'}
True