20 December 2017 0 comments Web development, JavaScript
I'm working on a project where it needs to evaluate CSS as a string. Basically, it compares CSS selectors against a DOM to see if the CSS selector is used in the DOM.
But CSS has pseudo classes. A common one a lot of people are familiar with is: a:hover { text-decoration: crazy }. So that :hover part is not relevant when evaluating the CSS selector against the DOM. So you chop off the :hover bit and is left with a which you can then look for in the DOM.
But there are some tricks and make this less trivial. Consider, this from Bootstrap 3
a[href^="#"]:after,
a[href^="javascript:"]:after {
content: "";
}
In this case we can't simply split on a : character.
Another non-trivial example comes from Semantic UI:
.ui[class*="4:3"].embed {
padding-bottom: 75%;
}
.ui[class*="16:9"].embed {
padding-bottom: 56.25%;
}
.ui[class*="21:9"].embed {
padding-bottom: 42.85714286%;
}
Basically, if you just split the selectors (e.g. a:hover) on the first : and keep everything to the left (e.g. a), with these non-trivial CSS selectors you'd get this:
a[href^="javascript
and
.ui[class*="4
etc. These CSS selectors will fail. Both Firefox and Chrome seem to swallow any errors but cheerio will raise a SyntaxError and not just that but the problem is that the CSS selector is just the wrong one to look for.
The solution has to be to split by the : character when it's not between two quotation marks.
This Stackoverflow post helped me with the regex. It was trivial to extend now my final solution looks like this:
/**
* Reduce a CSS selector to be without any pseudo class parts.
* For example, from 'a:hover' return 'a'. And from 'input::-moz-focus-inner'
* to 'input'.
* Also, more advanced ones like 'a[href^="javascript:"]:after' to
* 'a[href^="javascript:"]'.
* The last example works too if the input was 'a[href^='javascript:']:after'
* instead (using ' instead of ").
*
* @param {string} selector
* @return {string}
*/
const reduceCSSSelector = selector => {
return selector.split(
/:(?=([^"'\\]*(\\.|["']([^"'\\]*\\.)*[^"'\\]*['"]))*[^"']*$)/g
)[0]
}
I've been coding for about 20 years and would like to think I know my way around writing regular expressions in various languages. However, I'm also eager to admit that I often fumble and rely on googling/stackoverflow more than actually understanding what the heck I'm doing. That's why I found this comment so amusing:
Thank you! Didn't think it was possible. I understand 100% of the theory, about 60% of the regex, and I'm down to 0% when it comes to writing it on my own. Oh, well, maybe one of these days. – Azmisov