smart_split behaviour is surprising

[Stephen Kelly]

I attempted to fix #15092 with a smart_split and noticed that it does not split on unescaped " not surrounded by whitespace.

If this is intentional it should be documented. I don't think any other unit test for templates tries to use a 'corrupt' string like that.
It seems that FilterExpression doesn't accept strings like that anyway, so it should either be documented that they don't work in the template system, or smart_split should be changed.

In [1]: from django.utils.text import smart_split

In [2]: smart_split(' now "j "n" Y"'
   ...: )
Out[2]: <generator object smart_split at 0xa0f5d74>

In [3]: for a in smart_split(' now "j "n" Y"'): print a
   ...: 
now
"j "n" Y"

In [4]: for a in smart_split(' now "j " n " Y"'): print a
   ...: 
now
"j "
n
" Y"

[Russell Keith-Magee]

This looks like it's an oversight in the parsing algorithm; I can't imagine it would be intentional.

[ersame]

I am working on it.

[ersame]

I think it is not a bug, in the comment of smart_split definition it appears:
"Generator that splits a string by spaces".

If we look to the test case in django.utils.text we can see:

self.assertEquals(list(smart_split(u'url search_page words="something else"')),
            [u'url', u'search_page', u'words="something else"'])

This test won't pass if you patch smart_split in the way you report.

We would need to decide how we want smart_split works. In my opinion
smart_split should split by spaces just what we have.

Anyway if someone wants the other behaviour could use my patch to solve
the problem.

What do you think?

Regards,

Juan Antonio Infantes.

[Carl Meyer]

In [16108]:

Refs #15093 -- Fixed another get_models call missed in r16053. Thanks Luke for catching it.

[Carl Meyer]

Oops, fat-fingered that commit message - should have referred to #15903.

[Aymeric Augustin]

Change UI/UX from NULL to False.

[Aymeric Augustin]

The behavior is intentional and it's described in the docstring: "Generator that splits a string by spaces, leaving quoted phrases together."

This is a private API, and such APIs are only documented in their docstrings.