2026-01-28 datalevin | Clojure Slack Archive

datalevin

Huahai 2026-01-28T16:18:08.207789Z

0.10.3 is released with query optimizer improvements https://github.com/datalevin/datalevin/blob/master/CHANGELOG.md#0103-2026-01-27

🎉 2

2026-01-28T21:52:55.970149Z

Quick question that I'm stumped on and I think I'm just missing something silly. This should probably be an or-join but I can't seem to make that work. The following rule works but the performance is bad (2s):

[(entity-area' ?area ?room)
     (or
       ;; Check for room within continuous area w/ quadrants
       (and
         [?area-or-quad :locations ?room]
         [?area :quadrants ?area-or-quad])
       ;; Check for room within discrete area locations (no quadrant)
       (and
         [?area-or-quad _]
         [?area :locations ?room]
         [?area :area-type :discrete]))]

I know the [?area-or-quad _] is the performance problem. How do I rework this? (or-join [?area ?room]... runs quickly but returns nil .

Huahai 2026-01-28T22:01:23.370939Z

what does this do [?area-or-quad _] you probably don't need it?

Huahai 2026-01-28T22:04:52.651389Z

Use or-join and remove [?area-or-quad _]

Huahai 2026-01-28T22:06:01.900919Z

or-join branches can use different sets of variables, so you do't need to include ?area-or-quad

Huahai 2026-01-28T22:18:07.303889Z

The way to think of or-join, is to think each of its branches as a sub-query. The interface between these sub-queries and the main query is the argument list. These sub-queries are also independent.

Huahai 2026-01-28T22:19:22.435069Z

or-join is optimized, so it normally performs much better than actual sub-queries, where a full query is run for each tuple, so it's quite expensive.

2026-01-29T01:03:56.109279Z

Ok thanks so much! I was able to solve the rule with:

[(entity-area ?area ?room)
     (or-join [?area]
              ;; Check for room within discrete area locations (no quadrant)
              (and [?area :locations ?room])
              ;; Check for room within continuous area w/ quadrants
              (and [?area :quadrants ?area-or-quad]
                   [?area-or-quad :locations ?room]))]

Huahai 2026-01-29T01:06:46.974329Z

you can get rid of the first and, as it only has one triple pattern, a naked triple pattern there works

2026-01-29T01:11:35.163929Z

haha right of course. I think it was leftover from something else I was trying. Appreciate it. Really excited for the new releases. I've been working on my project a long time and need to upgrade

2026-01-29T01:11:52.952079Z

It is fully persistent MUD using datalevin as the database

Huahai 2026-01-29T01:12:13.478599Z

cool:+1:

Huahai 2026-01-28T05:47:16.707159Z

Wrote a piece on SQLite https://yyhh.org/blog/2026/01/sqlite-in-production-not-so-fast-for-complex-queries/

🚀 3

Clojurians Log v2

datalevin