Inconsistency in TestCase.assertInHTML

[Andrew Plummer]

assertInHTML(needle, haystack) has the following behaviour

assertInHTML('<p>a</p>', '<div><p>a</p><p>b</p></div>') passes: clearly correct

assertInHTML('<p>a</p><p>b</p>', '<p>a</p><p>b</p>') passes: possibly correct

assertInHTML('<p>a</p><p>b</p>', '<div><p>a</p><p>b</p></div>') fails with an assertion error:

  File ".../django/test/testcases.py", line 673, in assertInHTML
    msg_prefix + "Couldn't find '%s' in response" % needle)
AssertionError: Couldn't find '<p>
a
</p><p>
b
</p>' in response

Which seems wrong. It doesn't handle the case correctly if the needle doesn't have a single top-level element (unless the two bits of html are equivalent). This is down to the _count method of django.test.html.Element.

This means that the test will fail (with the error above) rather than erroring.

I think this should throw a ValueError if the needle has multiple top level elements. However, this does change existing behaviour for the case where the html fragments are equivalent. An alternative would be to try to correctly handle the count in this case, which is possibly a bit fiddly, and the current implementation appears to make no attempt to do this.

I have a patch for the first case: ​https://github.com/plumdog/django/commit/bfdfda315ad74d067be52888a236ab7a4aadcf96

[Tim Graham]

Please make a pull request with your proposed fix and we'll go from there.

[Andrew Plummer]

Pull Request created:

​https://github.com/django/django/pull/4041

[Andrew Plummer]

I've updated my PR to include tests, and changed my proposed fix to preserve existing behaviour where I believe the existing behaviour to be correct, namely where the two HTML fragments are equal.

[Tim Graham]

Comments for improvement on PR.

[Adam Zapletal]

I'm willing to clean up the patch and open a new PR, or do you want to finish it yourself, plumdog?

[Claude Paroz]

After 18 months, any taker can freely go ahead!

[Adam Zapletal]

PR is here: ​https://github.com/django/django/pull/7079

[Adam Zapletal]

PR is updated with an attempt at an actual fix, not just a cleanup of plumdog's original PR

​https://github.com/django/django/pull/7079

[Tim Graham <timograham@…>]

In ca2ccf54:

Fixed #24112 -- Fixed assertInHTML()'s counting if needle has no root element.