#orm #sql-database #sqlite #database-migrations #ormlite #sql-query #migration

app ormlite-cli

An ORM for people who love SQL. Use the ormlite crate, not this one.

31 releases (13 breaking)

0.22.2 Oct 8, 2024
0.21.0 Oct 4, 2024
0.17.5 Nov 19, 2023
0.17.1 Jul 30, 2023
0.12.1 Mar 27, 2023

#1255 in Database interfaces

MIT license

135KB
3K SLoC

GitHub Contributors Stars Build Status Downloads Crates.io

ormlite

ormlite is an ORM in Rust for developers that love SQL. Let's see it in action:

use ormlite::model::*;
use ormlite::sqlite::SqliteConnection;

#[derive(Model, Debug)]
pub struct Person {
    pub id: i32,
    pub name: String,
    pub age: i32,
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    /// Start by making a database connection.
    let mut conn = SqliteConnection::connect(":memory:").await.unwrap();

    /// You can insert the model directly.
    let mut john = Person {
        id: 1,
        name: "John".to_string(),
        age: 99,
    }.insert(&mut conn).await?;

    println!("{:?}", john);

    /// After modifying the object, you can update all its fields.
    john.age += 1;
    john.update_all_fields(&mut conn).await?;

    /// Query builder syntax closely follows SQL syntax, translated into chained function calls.
    let people = Person::select()
        .where_("age > ?").bind(50)
        .fetch_all(&mut conn).await?;

    println!("{:?}", people);
}

You might like ormlite because:

  • It auto-generates migrations from Rust structs. To my knowledge, it is the only Rust ORM with this capability.
  • The join API (in alpha) has far fewer moving pieces than any other Rust ORM. It only relies on the table structs themselves, and does not rely on relation traits (SeaORM) or modules (Diesel).
  • There's little to no query builder syntax to learn. The query builder basically joins together &str fragments of raw SQL. It strikes the right level of abstraction between composability, and having near-zero learning curve for anyone who already knows SQL.

Quickstart

Installation

Install with cargo:

# For postgres
cargo add ormlite --features postgres
# For sqlite
cargo add ormlite --features sqlite

Or update your Cargo.toml:

[dependencies]
# For postgres
ormlite = { version = "..", features = ["postgres"] }
# For sqlite
ormlite = { version = "..", features = ["sqlite"] }

Other databases and runtimes are supported, but are less tested. Please submit an issue if you encounter any.

Environment Setup

You need DATABASE_URL in your environment. We recommend a tool like just, which can pull in a .env file, but for simplicity, here we'll use shell directly.

export DATABASE_URL=postgres://postgres:postgres@localhost:5432/postgres

Migrations

If you are querying a static database and don't need migrations, skip this section. If you want migrations, keep reading.

First, install ormlite-cli. Currently, the CLI only supports Postgres. While ormlite-cli is separate from sqlx-cli, they are 100% compatible with each other. sqlx-cli does not support auto-generating migrations or snapshots (to rollback in development without writing down migrations), but it is less bleeding edge and supports more database types.

cargo install ormlite-cli

Next, create the database and the migrations table. init creates a _sqlx_migrations table that tracks your migrations.

# Create the database if it doesn't exist. For postgres, that's:
# createdb <dbname>
ormlite init

Let's see migrations in action. Create a Rust struct with #[derive(Model)], which the CLI tool detects to auto-generate migrations:

# src/models.rs

use ormlite::model::*;

#[derive(Model, Debug)]
pub struct Person {
    pub id: i32,
    pub name: String,
    pub age: i32,
}

Next, auto-generate the migration.

ormlite migrate initial

This creates a plain SQL file in migrations/. Let's review it before we execute it:

cat migrations/*.sql

Once you're satisfied reviewing it, you can execute it:

ormlite up

By default, up also creates a snapshot, so you can rollback using ormlite down if need be. There's also an option to generate paired up/down migrations instead of only up migrations.

That's the end of setup. Let's now look at how to run queries.

Insert & Update

The insert and update syntax at the top of the README is most effective for UUID primary key tables.

use ormlite::model::*;
use uuid::Uuid;

#[derive(Model, Debug)]
pub struct Event {
    pub id: Uuid,
    pub name: String,
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut conn = ormlite::sqlite::SqliteConnection::connect(":memory:").await.unwrap();

    let mut event = Event {
        id: Uuid::new_v4(),
        name: "user_clicked".to_string(),
    }.insert(&mut conn).await?;

    println!("{:?}", event);
}

This syntax has two possible issues. First, id is not Option, so it must be set, causing problems for autoincrement id fields. Second, the struct cannot track which fields are modified, so the update method must update all columns.

Insertion Struct

To work around the autoincrement issue, you can use an insertion struct, shown here, or a builder, shown below.

use ormlite::types::Json;
use serde_json::Value;

#[derive(Model, Debug)]
#[ormlite(insert = "InsertPerson")]
pub struct Person {
    pub id: i32,
    // Because the other fields are the primary key, and marked as default and default_value respectively,
    // `name` is the only field in the InsertPerson struct.
    pub name: String,
    // This field will not be part of the InsertPerson struct,
    // and rows will take the database-level default upon insertion.
    #[ormlite(default)]
    pub archived_at: Option<DateTime<Utc>>,
    // This field will not be part of the InsertPerson struct, 
    // which will always pass the provided value when inserting.
    #[ormlite(default_value = "serde_json::json!({})")]
    pub metadata: Json<Value>,
}

async fn insertion_struct_example(conn: &mut SqliteConnection) {  
    let john: Person = InsertPerson {
        name: "John".to_string(),
    }.insert(&mut conn).await?;

    println!("{:?}", john);
}

If the derived struct doesn't meet your needs, you can manually define a struct that only contains the fields you want, specifying table = "<table>" to route the struct to the same database table.

#[derive(Model, Debug)]
#[ormlite(returns = "User")]
pub struct InsertPerson {
    pub name: String,
    pub age: i32,
}

Insert Builder & Update Builder

You can also use builder syntax for insertion or to update only certain fields.

#[derive(Model, Debug)]
pub struct Person {
    pub id: i32,
    pub name: String,
    pub age: i32,
}

async fn builder_syntax_example() {    
    // builder syntax for insert
    let john = Person::builder()
        .name("John".to_string())
        .age(99)
        .insert(&mut conn).await?;

    println!("{:?}", john);

    // builder syntax for update
    let john = john.update_partial()
        .age(100)
        .update(&mut conn).await?;

    println!("{:?}", john);
}

Upsert

You can handle insertion on conflict using OnConflict (docs).

use ormlite::{
    model::*,
    query_builder::OnConflict,
};

#[derive(Debug, Model)]
pub struct Users {
    #[ormlite(primary_key)]
    pub id:    i32,
    pub name:  String,
    pub email: String,
}

async fn upsert_example(conn: &mut PgConnection) {
    Users {
        id: 1,
        name: String::from("New name"),
        email: String::from("New email"),
    }
        .insert(&mut conn)
        // update values of all columns on primary key conflict
        .on_conflict(OnConflict::do_update_on_pkey("id"))
        .await
        .unwrap();
}

Select Query

You can use Model::select to build a SQL query using Rust logic.

Note: Postgres's approach of using numbered dollar sign placeholders quickly breaks down when building queries. Instead, even with Postgres, use ? for parameters, and ormlite will replace the ? placeholders with $ placeholders when it constructs the final query.

#[derive(Model, Debug)]
pub struct Person {
    pub id: i32,
    pub name: String,
    pub age: i32,
}

async fn query_builder_example() {
    let people = Person::select()
        .where_("age > ?")
        .bind(50i32)
        .fetch_all(&mut conn)
        .await?;
    println!("All people over 50: {:?}", people); 
}

Raw Query

You can fall back to raw queries if the ORM methods don't work for you. You can include handwritten strings, or if you want a lower-level query builder, you can use sqlmo, the underlying engine that powers ormlite's query builder & migration auto-generation.

async fn model_query_example() {
    // Query using the Model to still deserialize results into the struct
    let _person = Person::query("SELECT * FROM person WHERE id = ?")
        .bind(1)
        .fetch_one(&mut conn)
        .await?;
}

async fn raw_query_example() {
    // You can also use the raw query API, which will return tuples to decode as you like
    let _used_ids: Vec<i32> = ormlite::query_as("SELECT id FROM person")
        .fetch_all(pool)
        .await
        .unwrap()
        .into_iter()
        .map(|row: (i32, )| row.0)
        .collect();
}

Table Customization

Attributes are defined in these structs.

This example shows them in action:

#[derive(Model, Debug)]
#[ormlite(table = "people", insert = "InsertPerson")]
pub struct Person {
    #[ormlite(primary_key)]
    pub id: i32,
    pub name: String,
    #[ormlite(column = "name_of_db_column")]
    pub age: i32,
}

Joins

Join support is alpha stage. Right now, ormlite only support many-to-one relations (e.g. Person belongs to Organization). Support for many-to-many and one-to-many is planned. If you use this functionality, please report any bugs you encounter.

#[derive(Model, Debug)]
pub struct Person {
    pub id: Uuid,
    pub name: String,
    pub age: i32,
    
    // Note that we don't declare a separate field `pub organization_id: Uuid`.
    // It is implicitly defined by the Join and the join_column attribute.
    #[ormlite(join_column = "organization_id")]
    pub organization: Join<Organization>,
}

#[derive(Model, Debug)]
pub struct Organization {
    pub id: Uuid,
    pub name: String,
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Note we don't need to insert it.
    let org = Organization {
        id: Uuid::new_v4(),
        name: "Acme".to_string(),
    };
    
    let user = Person {
        id: Uuid::new_v4(),
        name: "John".to_string(),
        age: 99,
        organization: Join::new(org),
    };
    
    let mut conn = ormlite::sqlite::SqliteConnection::connect(":memory:").await.unwrap();

    let user = user.insert(&mut conn).await?;
    assert_eq!(user.organization.loaded(), true);
    println!("{:?}", user);
    
    // You can choose whether you want to load the relation or not. The value will be Join::NotQueried if you don't 
    // opt-in to loading it.
    let users = Person::select()
        .join(Person::organization())
        .fetch_all(&mut conn)
        .await?;

    for user in users {
        assert!(user.organization.loaded());
        println!("{:?}", user);
    }
}

Features & Data Types

Uuid, Chrono, & Time

If you want Uuid or DateTime, combined with serde, you need to depend directly on uuid, time or chrono, and add the serde feature to each of them.

# Cargo.toml
[dependencies]
uuid = { version = "...", features = ["serde"] } 
chrono = { version = "...", features = ["serde"] }
time = { version = "...", features = ["serde"] }
use ormlite::model::*;
use serde::{Serialize, Deserialize};
use ormlite::types::Uuid;
use ormlite::types::chrono::{DateTime, Utc};

#[derive(Model, Debug, Serialize, Deserialize)]
pub struct Person {
    pub uuid: Uuid,
    pub created_at: DateTime<Utc>,
    pub name: String,
}

Json/Jsonb Columns

You can either use ormlite::types::Json for JSON or JSONB fields, or you can use the json attribute. For unstructured data, use serde_json::Value as the inner type. Use a struct with Deserialize + Serialize as the generic for structured data.

use ormlite::model::*;
use ormlite::types::Json;
use serde_json::Value;

#[derive(Debug, Serialize, Deserialize)]
pub struct JobData {
    pub name: String,
}

#[derive(Model, Serialize, Deserialize)]
pub struct Job {
    pub id: i32,
    pub structured_data: Json<JobData>,
    pub unstructured_data: Json<Value>,
    #[ormlite(json)]
    pub unstructured_data2: Value,
    #[ormlite(json)]
    pub structured_data2: JobData,
}

Logging

You can log queries using sqlx's logger: RUST_LOG=sqlx=info

Roadmap

  • Insert, update, delete directly on model instances
  • Builder for partial update and insertions
  • User can create insert models that ignore default values
  • Select query builder
  • Build the derive macro
  • Get() function for fetching a single entity.
  • Ability to specify the name of a table and name of primary column
  • Automatically generate insert models
  • Automatically generate migrations
  • Eliminate need for FromRow macro
  • Many to one joins
  • created_at should naturally default to now()
  • id: i32 should default to identity by default
  • Autogenerate indexes for migrations
  • Many to many joins
  • One to many joins
  • Make sure features are wired up correctly to support mysql and different runtimes & SSL libraries.
  • Macro option to auto adjust columns like updated_at
  • Upsert functionality
  • Bulk insertions
  • Query builder for bulk update
  • Handle on conflict clauses for bulk update
  • Benchmarks against raw sql, sqlx, ormx, seaorm, sqlite3-sys, pg, diesel
  • Support for patch records, i.e. update with static fields.
  • Consider a blocking interface, perhaps for sqlite/Rusqlite only.

Contributing

Open source thrives on contributions, and ormlite is a community project. We welcome you to file bugs, feature requests, requests for better docs, pull requests, and more!

Dependencies

~62MB
~1M SLoC