Choose filters to apply

[davidlesieur]

Following up on getpelican/pelican#3436, here's a take at making the typogrify() function more flexible by allowing one to select which filters to apply.

Some remarks about the proposed changes:

• To keep the new keyword arguments to applyfilters() and typogrify() short while avoiding name clashes with their corresponding filter functions, they are accessed through kwargs.pop(). That makes the function signatures less self-explanatory, but I am mentioning the keyword arguments in the docstrings.
• I have moved a doctest from applyfilters() to typogrify(), because it was specifically testing typogrify().
• I have added doctests for the new options.
• I have noticed that widont ignores the ignore_tags list, and then discovered that this issue had already been reported in Move widont filter to applyfilters function mintchaos/typogrify#30. I have added a comment and a test about this behavior. I understand that widont only applies to a hardcoded list of tags, but it probably makes the API less predictable to one who wouldn't fully read the documentation or code. However, this would perhaps be better addressed separately from this PR.

[davidlesieur]

Regarding the widont issue, just noticed the related PR mintchaos#31. It moves the call to widont() to applyfilters(), subjecting it to the list of ignored tags. If you'd like, I could integrate that change here.

[justinmayer]

By all means — integrating that change here seems like a sensible approach. Thank you for suggesting that and for jumping on these changes! 😁

[davidlesieur]

Moved the widont filter to applyfilters(). However, this breaks a test in Pelican, which expects widont to ignore the ignore_tags.

Part of the Pelican tests output:

FAILED pelican/tests/test_readers.py::RstReaderTest::test_typogrify_ignore_tags - AssertionError: '<p>T[34 chars]ff to &quot;typogrify&quot;...</p>\n<p>Now wit[68 chars]p>\n' != '<p>T[34 chars]ff to&nbsp;&quot;typogrify&quot;...</p>\n<p>No[73 chars]p>\n'

Results (2.78s):
     268 passed
       1 failed
         - pelican/tests/test_readers.py:412 RstReaderTest.test_typogrify_ignore_tags
      27 skipped

[justinmayer]

If the following relevant comment is to be believed, the widont filter should be able to handle tags and thus not need to ignore them:

# apply widont at the end, as its already smart about tags. Hopefully.

That being the case, perhaps the Pelican test should be updated to no longer require widont to ignore the ignore_tags? What do you think?

[davidlesieur]

Yes, I think the Pelican test should be changed. My understanding of the comment on widont being "smart" is that it ever only applies to tags h1-h6, p, li, dd, dt, a, em, strong, span, b, i. By moving widont to applyfilters(), we make it more consistent with other filters in that it will now ignore any of those tags if they are specified in ignore_tags. This idea was also in mintchaos#30, which gave the example of someone not wanting widont on h1 tags.

[justinmayer]

Many thanks for these enhancements, David! 🏅

[justinmayer]

@davidlesieur: Would you be willing to submit a pull request to Pelican to modify the aforementioned test? That way, as soon as I publish the next Typogrify release, we can merge that Pelican PR and ensure tests continue to pass. What do you think?