SQLX migrator

A Rust library for writing SQLX migrations using Rust instead of SQL.

License	Crates Version	Docs

Supported Databases:

• PostgreSQL
• SQLite
• MySql
• Any

Installation

Add sqlx_migrator to your Cargo.toml with the appropriate database feature:

sqlx_migrator = { version = "0.16.2", features=["postgres"] }

OR

sqlx_migrator = { version = "0.16.2", features=["mysql"] }

OR

sqlx_migrator = { version = "0.16.2", features=["sqlite"] }

OR

sqlx_migrator = { version = "0.16.2", features=[
    "any",
    # Plus any one of above database driver
    ] }

Usage

To use sqlx_migrator, implement the Operation trait to define your migration logic. Here's an example using PostgreSQL:

use sqlx_migrator::error::Error;
use sqlx_migrator::operation::Operation;

pub(crate) struct FirstOperation;

#[async_trait::async_trait]
impl Operation<sqlx::Postgres> for FirstOperation {
    // Up function runs apply migration
    async fn up(&self, connection: &mut sqlx::PgConnection) -> Result<(), Error> {
        sqlx::query("CREATE TABLE sample (id INTEGER PRIMARY KEY, name TEXT)")
            // NOTE: if you want to use connection multiple times pass `&mut *connection`
            // as a parameter instead of `connection`
            .execute(connection)
            .await?;
        Ok(())
    }

    // down migration runs down migration
    async fn down(&self, connection: &mut sqlx::PgConnection) -> Result<(), Error> {
        sqlx::query("DROP TABLE sample").execute(connection).await?;
        Ok(())
    }
}

After defining your operations, you can create a migration:

use sqlx_migrator::error::Error;
use sqlx_migrator::migration::Migration;
use sqlx_migrator::operation::Operation;

pub(crate) struct FirstMigration;

impl Migration<sqlx::Postgres> for FirstMigration {
    // app where migration lies can be any value
    fn app(&self) -> &str {
        "main"
    }

    // name of migration
    // Combination of migration app and name must be unique to work properly expects for virtual migration
    fn name(&self) -> &str {
        "first_migration"
    }

    // Use the parent function to add parents of a migration.
    // If you cannot access or create the parent migration easily, you can also use
    // `(A,N) where A: AsRef<str>, N: AsRef<str>` where A is the app name
    // and N is the name of the migration.
    fn parents(&self) -> Vec<Box<dyn Migration<sqlx::Postgres>>> {
        vec![]
        // vec![("main", "initial_migration"), AnotherInitialMigration]
    }

    // use operations function to add operation part of migration
    fn operations(&self) -> Vec<Box<dyn Operation<sqlx::Postgres>>> {
        vec![Box::new(FirstOperation)]
    }

    // Migration trait also have multiple other function see docs for usage
}

This migration can be represented in a simpler form using macros:

use sqlx_migrator::vec_box;
sqlx_migrator::migration!(
    sqlx::Postgres,
    FirstMigration,
    "main",
    "first_migration",
    vec_box![], 
    vec_box![FirstOperation]
);
// OR
sqlx_migrator::postgres_migration!(
    FirstMigration,
    "main",
    "first_migration",
    vec_box![], 
    vec_box![FirstOperation]
);

If your up and down queries are simple strings, you can simplify the implementation:

sqlx_migrator::postgres_migration!(
    FirstMigration,
    "main",
    "first_migration",
    sqlx_migrator::vec_box![],
    sqlx_migrator::vec_box![
        (
            "CREATE TABLE sample (id INTEGER PRIMARY KEY, name TEXT)",
            "DROP TABLE sample"
        )
    ]
);

Finally, create a migrator to run your migrations:

use sqlx_migrator::migrator::{Info, Migrate, Migrator};
use sqlx::Postgres;

#[tokio::main]
async fn main() {
    let uri = std::env::var("DATABASE_URL").unwrap();
    let pool = sqlx::Pool::<Postgres>::connect(&uri).await.unwrap();
    let mut migrator = Migrator::default();
    migrator.add_migration(Box::new(FirstMigration));
}

Running Migrations

You can run migrations directly or integrate them into a CLI:

Programmatic Execution

use sqlx_migrator::migrator::Plan;
let mut conn = pool.acquire().await?;
// use apply all to apply all pending migration
migrator.run(&mut *conn, Plan::apply_all()).await.unwrap();
// or use revert all to revert all applied migrations
migrator.run(&mut *conn, Plan::revert_all()).await.unwrap();
// If you need to apply or revert to certain stage than see `Plan` docs

CLI Integration

To integrate sqlx_migrator into your CLI, you can either use the built-in MigrationCommand or extend your own CLI with migrator support. Below are examples of both approaches:

Built-in Migration Command

use sqlx_migrator::cli::MigrationCommand;

MigrationCommand::parse_and_run(&mut *conn, Box::new(migrator)).await.unwrap();

Extending Your Own CLI with Migrator Support

#[derive(clap::Parser)]
struct Cli {
    #[command(subcommand)]
    sub_command: CliSubcommand
}

#[derive(clap::Subcommand)]
enum CliSubcommand {
    #[command()]
    Migrator(sqlx_migrator::cli::MigrationCommand)
}

impl Cli {
    async fn run() {
        let cli = Self::parse();
        // create connection
        match cli.sub_command {
            Migrator(m) => {
                m.run(&mut conn, Box::new(migrator)).await.unwrap()
            }
        }
    }
}

Migrate from sqlx default sql based migration

To migrate from sqlx sql based migration to rust migration the recommended approach is to rewrite your SQL migrations as Rust operations and migrations as explained above. After rewriting your SQL migrations, you need to mark them as applied without re-executing them. This step ensures that the migration state aligns with the existing database. There are two ways to perform a fake apply:

Programmatic Fake Apply

Use the fake option with the Plan::apply_all() function:

use sqlx_migrator::migrator::Plan;

migrator.run(&mut *conn, Plan::apply_all().fake(true)).await.unwrap();

CLI-Based Fake Apply

If you're using a CLI, use the --fake flag with the apply command: <migrator_cli_command> apply --fake

Note: Before writing any other migrations

Before adding new migrations for future updates, ensure you complete the above steps to mark existing migrations as applied. Run the fake apply only once to align the migration state. After this, remove the fake(true) option or the --fake flag to allow new migrations to execute normally.

By following these steps, you can seamlessly transition from SQLX SQL-based migrations to Rust migrations while maintaining an accurate migration state and ensuring compatibility for future updates.