Mops package: https://mops.one/rxmodb
Old thread: https://forum.dfinity.org/t/motoko-database/20759
Setup
I’ll keep things simple here and not get into details explaining how you set up your documents.
Example setup https://github.com/infu/RxMoDb/blob/main/test/hero.mo
import Hero "./hero";
stable let hero_store = Hero.init();
let hero = Hero.use(hero_store);
Usage
Insert or Update (if using PK)
hero.db.insert({
id=123;
updatedAt= 1;
score= 3;
name= "B";
level= 1;
skills= ["Fireball","Blizzard","Thunder"];
deleted= true
});
Delete by PK
hero.pk.delete(123);
Index find
hero.score.find(0, ^0, #fwd, 10);
// [{id=123; name="B" ...}, ...]
PK get
hero.pk.get(123);
// ?{id=123; name="B" ...}
Under the hood, it has a plugin architecture. The only data structure that is fixed is the main storage using Vector. This way we can have fast idx:Nat → Document access. Indexes have customizable key type and point to that idx:Nat.
Utilizing the plugins a developer can decide what data structures to use. Currently, there are only PrimaryKey:Btree and Index:Btree plugins. (Anyone can make them).
Going to get more tests done. So far I was able to insert 10mil documents with 2 indexes.
// Motoko compiler 0.9.3 --incremental-gc
// 10mil documents (took 66m 44.7s to insert in local replica)
// 2 indexes. Nat64 -> Nat
// upgrades ok
"documents": "10 000 000",
"memory_size": "3 758 096 384",
"max_live_size": "0",
"heap_size": "3 644 987 000",
"total_allocation": "3 644 987 304",
"reclaimed": "0"
Thanks to @icme for BTree, @timo for Vector, @claudio for helping figure things out @skilesare for helping with design @luc-blaeser for Incremental Gc @ZenVoich for mops and the test framework
