Finer-grained Normalization

[SachaG]

How should we handle a freeform input such as:

A selector for when a sticky element (position:sticky) is in its stuck state, like :stuck

Matching position:sticky is misleading, because in the case of the "what features are missing from CSS" question that's not the feature that the respondent thinks is missing.

So I'm thinking that the stuck entity could define an "override", in other words whenever stuck is matched in a string, then position: sticky cannot also be matched, and that token would then simply not be included in the dataset.

Is this a good idea? Are there other cases where it would be applicable? Or is it better to report counts for both position: sticky and stuck, and let readers figure it out?

[SachaG]

Or, a more generic approach could be to define a regexp for :stuck that also requires matching against sticky in the same freeform textfield. So the word sticky would "belong" to the :stuck feature, and not be matched by the position: sticky feature.

In fact that might already be possible without any further code changes, I have to check…

[SachaG]

So this seems to work. I added this regex to the :stuck feature:

position(:| |: )?sticky(.*?)stuck|stuck (.*?)position(:| |: )?sticky

This matches anything between position:sticky and stuck; or stuck and position:sticky, including both bounds. So even if position: sticky is then matched it will be discarded because it's being matched inside a longer matched string (same logic behind how we match "React Native" for React Native and not for React).