I'm working on a Stack Exchange extension to help tag wiki editors. It uses the API to find out which tags referred to on every page have or lack a wiki or excerpt.
I'm using the API as efficiently as possible and I've asked questions on StackApps about not overstepping the line, but it still can break the throttling rules.
To fix this I want to do some aggressive smart caching and only call the API when necessary.
One way to do this is to scrape the pages for where this info is already given "for free".
One place I'm scraping is the tag popup windows when you hover over a tag. I'm using the new HTML5 mutation observers for this. They pass a packet of DOM references for each change you've registered. You then have to walk these DOM fragments. It's not yet clear if jQuery selectors can be used to save all the tricky DOM walking.
The current HTML structure of the tag popup contains a
<div class="tm-description">, which contains text (I'm not sure if it's the excerpt, the wiki, or uses heuristics to be one or the other.
Now if there is no wiki the text is hard-coded to "help us edit this wiki". It would be much more deterministic if the
div had one or more additional
classeses. Checking for this will break if the text changes in the future or if localization is added to the site.
Easiest would be including or omitting
blank. But it may not be any extra effort to include classes such as
no-excerpt etc. Perhaps mimicking the field names the tags APIs use.