5 releases
0.0.5 | Mar 12, 2021 |
---|---|
0.0.4 | Feb 5, 2021 |
0.0.3 | Dec 31, 2020 |
0.0.2 | Dec 11, 2020 |
0.0.1 | Oct 10, 2020 |
#2866 in Database interfaces
12KB
202 lines
sqlx-simple-migrator
This crate is a very lightweight migration framework. It simply runs a series of sql commands in succession. It is not sophisticated.
Let's take a look at the built-in migration that the crate uses to create the table:
pub fn migration() -> Migration {
Migration::new(NAME)
.with_up(
r#"
CREATE TABLE migrations (
name TEXT NOT NULL PRIMARY KEY,
executed_at TIMESTAMPTZ NOT NULL DEFAULT now()
)
"#,
)
.with_down(
r#"
DROP TABLE IF EXISTS migrations
"#,
)
}
Right now, it's hard coded against TIMESTAMPTZ making this crate only suitable for Postgres. @ecton is only using this crate with Postgres, but would welcome any contributions to make this more generic.
Each with_up
call is executed in the order it is added to the Migration structure. When rolling back a migration, the with_down
instructions are operated in reverse order. This allows you to write with_up
and with_down
on a single-structure basis like the example above shows, keeping the up and down logic close together.
If you're working on a migration and want it to execute every time, just add .debug()
to the builder pattern before returning it. debug()
is not enabled on builds without cfg(debug_assertions)
ensuring that if you build with --release
for deploying, you will never accidentally deploy a migration that was still marked as being debugged.
Lastly, if you want to test rebuilding the database from scratch, you can use .nuclear_debug()
instead, which will force every run to undo all migrations and redo them.
The pattern for executing migrations looks like this:
pub fn migrations() -> Vec<Migration> {
vec![
migration_0001_accounts::migration(),
// ...
]
}
pub async fn run_all() -> Result<(), MigrationError> {
let pool = connect_to_postgres();
Migration::run_all(&pool, migrations()).await
}
Dependencies
~20–30MB
~546K SLoC