Uniquify identical completions

[meedstrom]

How? There's the concept of a "tofu", an invisible character added to each completion candidate, to be stripped later. Prior art in Consult.

[semilin]

Why not include a unique property in identically named candidates? The CREATED property would make most sense to me, but you could also show the ID. Either way, I think distinguishing these candidates somehow would be better than showing them as identical.

[meedstrom]

True, good to distinguish them. It could also be done with an annotation-function (like org-node-affixation-fn, just with shorter annotations for completion-at-point).

Either way, I guess the hard part will be stripping the tofu or unique property, prior to returning the candidate. Haven't learned to do that.

[fleimgruber]

For inspiration (org-ql was also inspiration for org-roam-ql mentioned in #87) maybe look at: https://github.com/alphapapa/org-ql/blob/98c62ab0a6c084ae4132110e24d9fe1ace91d363/org-ql-completing-read.el#L191?

[krisbalintona]

Question: why are completion candidates node titles instead of the nodes themselves? For instance, org-roam-node-read has org-roam nodes as completing-read candidates.

[meedstrom]

You'd think so, it's just confusingly written. It's passed thru org-roam-node-read--to-candidate, so what completing-read gets is an alist where the keys are strings. As far as completing-read is concerned, it just uses the keys of the alist: strings.