5 releases

Uses old Rust 2015

0.3.1 Jan 21, 2018
0.3.0 Nov 20, 2016
0.2.2 Jan 4, 2016
0.2.1 Sep 24, 2015
0.2.0 Sep 20, 2015

#2325 in Database interfaces

LGPL-3.0

26KB
313 lines

Trek Travis build status

Database migration management for Rust programs.

All code in the src folder is licensed under LGPL v3+. All code in the examples folder is available under the MIT license.

Documentation

To use Trek from crates.io in your project, add it as a dependency in your Cargo.toml:

[dependencies]
trek = "0.3.1"

Usage

The heart of Trek is the MigrationIndex class, which keeps track of all migrations and manages applying and rolling them back. It's recommended that you add an implementation of the Default trait in your program for MigrationIndex. This allows creating a single source of truth for the list of database migrations that exist for your program. Unfortunately Rust doesn't allow implementing traits on structs from other crates, so you'll have to create a MigrationIndex wrapper in your own program and implemennt the Default trait on that. Fortunately the boilerplate code for this is simple and can be copied from the example program's implementation at examples/migration_index.rs. Remember after copying to edit the Default trait in that file and set it to an empty migration list. When you add new migrations, you'll add them to the Default impl in this file.

You'll also need a folder in your source tree to hold the migrations, and a mod.rs file that exports them. Each migration is its own struct in its own file in this folder. The example program uses the examples/migrations/ folder to store its migrations.

For ease of use, check out the example program at examples/example.rs to see how to hook Trek into your own program so you can use Trek's migration management through your own program's CLI interface.

Creating Migrations

Note: Trek expects migration names to be snake-cased like my_new_migration.

If you integrated Trek into your program's CLI interface like the example program does, then adding a new migration is as easy as cargo run my_program -- trek g migration new_migration_name (if you want to try it out with the example program, cargo run --example example -- trek g migration new_migration_name). Otherwise you'll have to call Trek::create_migration() programmatically, passing in the path to your migrations folder and the migration's snake-cased name. Either way, you'll get a new migration skeleton in your migrations folder. With the skeleton generated, there are a couple manual steps to turning into a fully-prepared migration:

  1. Fill out the new migration skeleton with your SQL. The up method provides the SQL to apply the migration, and the down method provides the SQL to undo it.
  2. Add a pub mod <migration file name> line to the mod.rs file in your migrations folder. This exports your new migration so it can be used in step 3.
  3. Update your MigrationIndex's Default impl to include the new migration. For an example, see the bottom of examples/migration_index.rs.

Running Migrations

The example program provides sample code at example/example.rs for integrating Trek's facilities for applying and rolling back migrations. It's recommended that you copy this code into your own program so that you can apply or roll back migrations from your own program's CLI interface.

Test Setup

Trek expects an empty PostgreSQL database that it can test against. To set it up for testing:

  1. Create a new, empty database named trek_test: connect to your postgres database and run create database trek_test;
  2. Set environment variables giving connection information for the new database: there's a skeleton shell script at tests/env_vars_example.sh. Copy this file to tests/env_vars.sh and edit env_vars.sh to have your database name, user name, and other information filled in. The connection parameters are expected to be in the format used by the rust-postgres crate.
  3. Load environment variables by running source tests/env_vars.sh
  4. Run tests with cargo test

Dependencies

~8MB
~219K SLoC