Continuing the discussion from We're live!:
I included Blake as the primary dev on SC integration plugin.
I have looked into existing solutions for tweaking the discourse fulltext search and, as Ayya Vimala mentioned before, there is not much interest among devs. I was then considering a way to do some sort of search hijack combined with regex parsing of search string into (multiple) SC IDs and remainder of keywords but it seemed too complicated. Next I considered a fulltext hack directly on PostgreSQL or DB level but this usually turns into a maintenance nightmare where every site upgrade might break something.
I almost gave up but then a very simple idea popped up: we have a solution implemented already, we just didn’t realize it
So, we take Blake’s autolink plugin, modify it to add a tag with the newly generated id to the thread in which it is invoked, and Bob’s your uncle!
We can then search on D&D with
tags:id1,id2 for posts containing id1 or id2 and with
tags:id1+id2 for posts containing both id1 and id2. And we can also send the exact same search string from SC to D&D
Of course it would be just too boring if this just worked, so I found some issues to fix.
search with tags containing dot doesn’t work as expected (other tags with sc ids work ok)
AN3.65 and MA16 Kālāma: Comparative translation, by Piya Tan is tagged with an3.65
https://discourse.suttacentral.net/search?q=tags%3Aan3.65 returns 0 results
We would of course need to bulk update all existing threads with all ids already mentioned in the posts within the thread, but there seem to be some easy solutions for that:
SC discussion button should generate https://discourse.suttacentral.net/search?q=tags:mn10 instead of https://discourse.suttacentral.net/search?q="mn10%20"
autolink plugin could be updated so that it contains a popup link to sutta card + additional link to EN trans. by Sujato
this is needed for suttas that have many-to-one mapping
autolink plugin could also be updated to include occurrences like an2,12 (typo) and an 2:12 (alternative style)
Then there is the issue of suttas on SC that don’t have a one-to-one mapping to URI
an2.11-20 should expand to tags:an2.11,an2.12,an2.13,an2.14,an2.15,an2.16,an2.17,an2.18,an2.19,an2.20 (user can then manually correct search string for specific sutta)
This can be done with a bit of code with all URIs that contain a hyphen or with a mapping table (maybe the same one that will be needed for the existing links pointing to the previous version of SC).
Perhaps this could be baked into an expansion table (we know that new sutta ids wil not be added so this is a one time job) and would fix SC search problems for suttas like an2.12
This could also be extended for para references in suttas: first link always points to sutta card, second to exact paragraph or start of the sutta within the range
/dn1 -> /dn1 (/en/sujato/dn1)
/dn1.2 -> /dn1 (/en/sujato/dn1#8)
/dn1.2.1 -> /dn1 (/en/sujato/dn#8)
/dn1.2.3 -> /dn1 (/en/sujato/dn1#10)
an2.11 -> an2.11-20 (/en/sujato/an2.11-20#sc11.1)
an2.12 -> an2.11-20 (/en/sujato/an2.11-20#sc12.1)
this would also work with the workaround for old links /en/dn1 and pi/dn1 -> /dn1
I noticed that there are two types of #bookmarks in the suttas, #sc12 and #12 for dn and mn, but just #sc12 for an.
Maybe #sc12 could be used to mark just the positions (for sections and start of the sutta, without highlighting) and #12 could highlight the referenced para? Then sections and paragraphs could be differentiated like this:
/dn1.2 -> /dn1 (/en/sujato/dn1#sc8)
/dn1.2.1 -> /dn1 (/en/sujato/dn#8)
The same principle should work for agamas and the rest