Conversation

Notices

1. LS (lain@lain.com)'s status on Saturday, 22-Feb-2025 16:45:37 JST LS
   <-- json-in-postgres-expert
   • LS (lain@lain.com)'s status on Saturday, 22-Feb-2025 17:14:27 JST LS

     in reply to

     • Johann150 ⁂ :ipv6: :open_access: ☮
     @Johann150 yes, it's a good fit, activities are documents. the performance characteristics are the same as if you had a fully normalized db, as longs as you have indexes on the actually relational parts. I would not split objects and activities n the database, though.
   • Johann150 ⁂ :ipv6: :open_access: ☮ (johann150@genau.qwertqwefsday.eu)'s status on Saturday, 22-Feb-2025 17:14:28 JST Johann150 ⁂ :ipv6: :open_access: ☮

     in reply to

     @lain@lain.com if pleroma was a new project started today, would you use json in postgres again?

   • LS (lain@lain.com)'s status on Saturday, 22-Feb-2025 17:33:05 JST LS

     in reply to

     • Johann150 ⁂ :ipv6: :open_access: ☮
     • feld
     @feld @Johann150 yes, but this doesn't happen if you only join / query over indexed expressions, because those use the statistics of the index, that's what we're doing mostly. i forgot about the extended statistics, good to know.
   • feld (feld@friedcheese.us)'s status on Saturday, 22-Feb-2025 17:33:06 JST feld

     in reply to

     • Johann150 ⁂ :ipv6: :open_access: ☮
     @lain @Johann150 we do gotta confront the bad query plans because of statistics tho. I just realized there's a manual way to work around it if you can identify it and I don't recall seeing this in our schema
     
     https://pganalyze.com/blog/5mins-postgres-planner-jsonb-selectivity
   • LS (lain@lain.com)'s status on Saturday, 22-Feb-2025 18:23:39 JST LS

     in reply to

     • Johann150 ⁂ :ipv6: :open_access: ☮
     • feld
     @feld @Johann150 actually looking at the original post that was linked in that post, it seems that they DID have a btree index on it. I should take a look again at slow queries.