tensorchord / pgvecto.rs

Scalable, Low-latency and Hybrid-enabled Vector Search in Postgres. Revolutionize Vector Search, not Database.
https://docs.pgvecto.rs/getting-started/overview.html
Apache License 2.0
1.77k stars 73 forks source link

Diesel Support #115

Open kj3moraes opened 1 year ago

kj3moraes commented 1 year ago

Will there be support added for Diesel ? It is easier to use Diesel to generate SQL queries.

cutecutecat commented 1 year ago

Diesel only supports binary format https://github.com/diesel-rs/diesel/issues/2087 now and it seems unchanged yet. But Pgvecto.rs not support binary result format now. Might related: https://github.com/tensorchord/pgvecto.rs/issues/100 For now, You would meet ERROR: no binary output function available for type vector.

Here is a partial finished example for Diesel, insert is fine while select is still ill.

#[macro_use]
extern crate diesel;

use std::{env, vec};

use chunks::emb;
use diesel::expression::{AsExpression, Expression, TypedExpressionType};
use diesel::{insert_into, Connection, PgConnection, QueryDsl, RunQueryDsl};
use diesel::{pg::Pg, sql_types::*};

diesel::infix_operator!(Emb, " <-> ", backend: Pg);

pub fn dis<T, U, ST>(left: T, right: U) -> Emb<T, U::Expression>
where
    T: Expression<SqlType = ST>,
    U: AsExpression<ST>,
    ST: SqlType + TypedExpressionType,
{
    Emb::new(left, right.as_expression())
}

table! {
    chunks {
        id -> Integer,
        emb -> Array<Float>,
    }
}

#[derive(Insertable, Queryable, Identifiable, Debug, PartialEq, Selectable)]
#[diesel(table_name = chunks)]
#[diesel(check_for_backend(Pg))]
struct VectorTable {
    id: i32,
    emb: Vec<f32>,
}

fn main() {
    // let database_url = env::var("DATABASE_URL").expect("DATABASE_URL must be set");
    let database_url = "postgresql://postgres:mysecretpassword@localhost:5432".to_string();
    let mut connection = PgConnection::establish(&database_url)
        .unwrap_or_else(|_| panic!("Error connecting to {}", database_url));

    let values: Vec<VectorTable> = vec![
        VectorTable {
            id: 0,
            emb: vec![0.1, 0.2, 0.3],
        },
        VectorTable {
            id: 1,
            emb: vec![0.1, 0.2, 0.3],
        },
    ];
    let _ = insert_into(chunks::table)
        .values(&values)
        .execute(&mut connection)
        .expect("insert failed");
    let filter = chunks::table
        .select(emb)
        .order(dis(emb, vec![0.1, 0.2, 0.3]))
        .load::<Vec<f32>>(&mut connection)
        .expect("select failed");
    for v in filter {
        println!("{:?}", v)
    }
}