Fork me on GitHub
#datalevin
<
2024-02-14
>
andersmurphy17:02:10

Not sure if this is relevant to the new query engine. But I saw this recent change in datascript. https://github.com/tonsky/datascript/pull/459

Huahai17:02:18

Yes, it is applicable

🎉 1
Huahai17:02:10

Won't have substantial saving, but it does for some cases. In our case, most clauses are handled by the planner, but some clauses are not. This change helps those.

Huahai17:02:27

For now planner handles "normal" clauses, we will gradually move more types of clauses to it.

👍 1
Huahai19:02:48

version q1 q2 q3 q4 q5 qpred1 qpred2 latest-datomic 6.2 10.6 14.3 18.0 156.9 22.2 25.3 latest-datascript 0.63 3.0 4.4 6.7 352.7 7.3 16.4 0.8.26-datalevin 0.70 2.9 3.4 5.1 308.7 7.9 8.8 latest-datalevin 0.66 0.68 3.0 3.3 177.5 3.6 2.8

Huahai19:02:05

I have not finished the optimization work, but the preliminary results are promising. q5 should substantially improve after I am done.

Huahai20:02:05

In any case, queries in this datascript benchmark are too simple, we will port Join-order-benchmark from SQL in order to see how it fairs with complex queries.

andersmurphy10:02:48

Looks very promising!