Improve performances using lightningcss

[gjeusel]

I wonder if using lightningcss under the hood and distribute pre-compiled binaries would improve performances and ergonomie of the project ?

It would have a great impact on our side, as we generate a huge css for using tailwindcss classes in our templates.

[liZe]

Hi!

lightningcss is another CSS parser, so using it would mean replacing the whole project code. 😄

Why don’t you use lightningcss instead of tinycss2 in your project?

[gjeusel]

Ho ! Okay, i thought it was a subset, or that there was some logic on top of css parsing applied.
I'm using it in the context of weasyprint pdf generation, hence I don't have control over which css parser to use.
Do you think I should open this discussion on weasyprint directly ?

[liZe]

I'm using it in the context of weasyprint pdf generation, hence I don't have control over which css parser to use.
Do you think I should open this discussion on weasyprint directly ?

There’s a chapter in the documentation explaining why we use Python and don’t consider speed as one of WeasyPrint’s main goal, I think that we won’t change the library!

That being said, could you please share a stylesheet (here, or by mail if it’s confidential) that takes too much time to parse for you? We can definitely find if there’s a bottleneck we can fix.

(And if you have a complete HTML+CSS sample, we can even profile the whole WeasyPrint + cssselect2 + tinycss2 + pydyf performance.)