Skip to content

More holistic approach to "tables"

Right now, we're trying to identify all the table types and make them into individual instances. There are many types however and this isn't clearly a strategy that will achieve complete coverage, which is important for deciding what to remove from plaintext. I propose to switch to using the role attribute alongside a few core classes. I think this would mean:

  • class Wikitable: core content tables and nothing else (<table class="wikitable...).
  • class Infobox: infoboxes and nothing else (<table class="infobox...).
  • class Note: hatnotes, article stub boxes, disambiguation boxes, and anything else that uses role=note. These can various types of tags (div, table) and I don't specify a list of classes that they must match to allow for edge cases.
  • class Navigation: navboxes, portalboxes, sister project boxes, and anything else that uses role=navigation. These can various types of tags (div, table) and I don't specify a list of classes that they must match to allow for edge cases.
  • class Message: message boxes in all their other forms (<table> with one of these classes: ambox, fmbox, ombox, imbox, cmbox, tmbox). This one gets tricky though -- ideally I'd like to also include anything else that uses role=presentation but in practice that role seems to be used for other things too. For instance, there's a <table role="presentation"> that is nested within article-stub boxes which would make them a Message nested within a Note? and also see the edge-case below which I presume happens in other forms too. And while dmbox would seem to fit this based on its classname, it actually is just a div element and is considered a Note.

Note: alternatively, we could keep it mostly as it is and only introduce the roles as a filter in plaintext functions. I prefer the full refactoring though as I think it's more consistent and therefore simpler and I don't currently like the messiness of the current taxonomy of table-like classes anyways.

Edge-cases that wouldn't be covered: