only strip newlines when parsing strings

[johnowhitaker]

Related issue: #668

Issue: inline spaces weren't preserved:

e.g. "Howdy <a href="https://answer.ai">answer</a> how are you?" -> "Howdy" (no trailing space).

We changed it to only strip newlines from strings in _parse, and added a test. The original test failed, the HTML was functionally identical but different by one newline - independant of the trailing space bug. We now compare two soups of the HTML before vs after the round trip

[gitnotebooks]

Found 1 changed notebook. Review the changes at https://app.gitnotebooks.com/AnswerDotAI/fasthtml/pull/797

[johnowhitaker]

@jph00 can you review when conveneint (I can explain when we next chat :) )

[johnowhitaker]

To expand, this was a bug we picked up during an impromptu office hours.
We first looked at the example from the issue and explored it in a dialog here.

Then, once we were convinced it was a good idea, we moved to the fasthtml.components nb and made the tweak. We added an inline, informal test, but then noticed the 'tests' section right below and added a test for this case instead. It failed though, an errant newline unrelated to our change (I think). @erikgaas had the bright idea to check if bs4 could check equivalence of two soups, and we changed the test to use that instead of comparing the xml strings directly.

@jph00 I am not 100% sure this is right - can you think if there are cases where we definitely do want to strip() strings that are parts of elements? Or other issues with our fix? Do you think it makes sense to still strip newlines, or should weget rid of the strip()s entirely?

[jph00]

lgtm! :) tbh I don't recall why the strip is there at all...