Comment by AdamTReineke | Hacker News Reader

Comment by AdamTReineke | Hacker News Reader

Parent Full thread

Fascinating report!

Are there any indicators that can reliably indicate usage of a CSS preprocessor? If so, what preprocessors are used and which ones? (Could property ordering, presence of hacks/comments, misspellings, prefixes, or source mapping comment be used to identify?)

What are the common values in ::before and ::after pseudo-element content properties?

If you leave out the identifiable CSS libraries, how does that affect the numbers? Do they scale by 10% or do the libraries (Bootstrap primarily) use rules/properties disproportionately enough to scale the other results?

What percent of files are minimized?

What does the histogram of selectors per site look like?

Edit: Also, how frequently do dataURIs show up (probably as background images)?

I thought long and hard about trying to identify preprocessor usage reliably because that is a super interesting piece of data, but nothing stuck out as a smoking gun. Sequentially increasing selectors

  tr

  tr td

  tr td a

could be LESS unrolling a nested bit:

  tr {
    td {
      a {

OR... could just be how people write CSS.

I mostly ignored by-domain grouping in my queries, but there is a host of analysis you could do that would get into the modeling portion of data science, teasing our correlation and trying to find 'groups' or styles' of CSS that is written on a site. I had only so much time ;-)

I also did a web perf report that is not fully edited or ready for consumption, but does not that 54% of sites I ran Google Page Speed Insights against didn't compress all of their CSS assets. (Same sample of domains)

Could you use the existence of CSS maps to identify the presence of certain preprocessors, even if not to distinguish them?

Reply on news.ycombinator.com