sequel-activerecord_connection

This is a database extension for Sequel that makes it to reuse an existing Active Record connection for database interaction.

This can be useful if you want to use a library that uses Sequel (e.g. Rodauth or rom-sql), or you're transitioning from Active Record to Sequel, or if you just want to use Sequel for more complex queries, and you want to avoid creating new database connections.

It works on ActiveRecord 4.2+ and fully supports PostgresSQL, MySQL and SQLite, both the native adapters and JDBC (JRuby). The SQL Server external adapter is supported as well (tinytds in Sequel), and there is attempted support for Oracle enhanced (oracle and in Sequel). Other adapters might work too, but their integration hasn't been tested.

Why reuse the database connection?

At first it might appear that, as long as you're fine with the performance impact of your database server having to maintain additional open connections, it would be fine if Sequel had its own database connection. However, there are additional caveats when you try to combine it with Active Record.

If Sequel and Active Record each have their own connections, then it's not possible to combine their transactions. If we executed a Sequel query inside of an Active Record transaction, that query won't actually be executed inside a database transaction. This is because transactions are tied to the database connection; if one connection opens a transaction, this doesn't affect queries executed on a different connection, even if both connections are used in the same ruby process. With this library, transactions and queries can be seamlessly combined between Active Record and Sequel.

In Rails context, there are additional considerations for a Sequel connection to play nicely. Connecting and disconnecting would have to go in lockstep with Active Record, to make commands such as rails db:create and rails db:drop work. You'd also need to find a way for system tests and the app running in the background to share the same database connection, which is something Sequel wasn't designed for. Reusing Active Record's connection means (dis)connecting and sharing between threads is all handled automatically.

Framework Agnostic

The only hard dependencies are:

• ActiveRecord
• Sequel
• after_commit_everywhere

...which means you can use it with any Rack / Ruby based framework: Rails / Roda / Sinatra etc. or even without a framework.

Installation

Add the gem to your project:

$ bundle add sequel-activerecord_connection

Usage

Assuming you've configured your ActiveRecord connection, you can initialize the appropriate Sequel adapter and load the activerecord_connection extension: e.g.

# Place in relevant initializer
# e.g. Rails: config/initializers/sequel.rb

require "sequel"
DB = Sequel.postgres(extensions: :activerecord_connection) # postgres

Now any Sequel operations that you make will internaly be done using the ActiveRecord connection, so you should see the queries in your ActiveRecord logs.

DB.create_table :posts do
  primary_key :id
  String :title, null: false
  Stirng :body, null: false
end

DB[:posts].insert(
  title: "Sequel::ActiveRecordConnection",
  body:  "Allows Sequel to reuse ActiveRecord's connection",
)
#=> 1

DB[:posts].all
#=> [{ title: "Sequel::ActiveRecordConnection", body: "Allows Sequel to reuse ActiveRecord's connection" }]

DB[:posts].update(title: "sequel-activerecord_connection")
#=> 1

The database extension supports postgresql, mysql2 and sqlite3 ActiveRecord adapters, just make sure to initialize the corresponding Sequel adapter before loading the extension.

Sequel.postgres(extensions: :activerecord_connection) # for "postgresql" adapter
Sequel.mysql2(extensions: :activerecord_connection)   # for "mysql2" adapter
Sequel.sqlite(extensions: :activerecord_connection)   # for "sqlite3" adapter

If you're on JRuby, you should be using the JDBC adapters:

Sequel.connect("jdbc:postgresql://", extensions: :activerecord_connection) # for "jdbcpostgresql" adapter
Sequel.connect("jdbc:mysql://", extensions: :activerecord_connection)      # for "jdbcmysql" adapter
Sequel.connect("jdbc:sqlite://", extensions: :activerecord_connection)     # for "jdbcsqlite3" adapter

Transactions

This database extension keeps the transaction state of Sequel and ActiveRecord in sync, allowing you to use Sequel and ActiveRecord transactions interchangeably (including nesting them), and have things like ActiveRecord's and Sequel's transactional callbacks still work correctly.

ActiveRecord::Base.transaction do
  DB.in_transaction? #=> true
end

Sequel's transaction API is fully supported:

DB.transaction(isolation: :serializable) do
  DB.after_commit { ... } # executed after transaction commits
  DB.transaction(savepoint: true) do # creates a savepoint
    DB.after_commit(savepoint: true) { ... } # executed if all enclosing savepoints have been released
  end
end

When registering transaction hooks, they will be registered on Sequel transactions when possible, in which case they will behave as described in the Sequel docs.

# Sequel: An after_commit transaction hook will always get executed if the outer
# transaction commits, even if it's added inside a savepoint that's rolled back.
DB.transaction do
  ActiveRecord::Base.transaction(requires_new: true) do
    DB.after_commit { puts "after commit" }
    raise ActiveRecord::Rollback
  end
end
#>> BEGIN
#>> SAVEPOINT active_record_1
#>> ROLLBACK TO SAVEPOINT active_record_1
#>> COMMIT
#>> after commit

# Sequel: An after_commit savepoint hook will get executed only after the outer
# transaction commits, given that all enclosing savepoints have been released.
DB.transaction(auto_savepoint: true) do
  DB.transaction do
    DB.after_commit(savepoint: true) { puts "after commit" }
    raise Sequel::Rollback
  end
end
#>> BEGIN
#>> SAVEPOINT active_record_1
#>> ROLLBACK TO SAVEPOINT active_record_1
#>> COMMIT

In case of (a) adding a transaction hook while Active Record holds the transaction, or (b) adding a savepoint hook when Active Record holds any enclosing savepoint, Active Record transaction callbacks will be used instead of Sequel hooks, which have slightly different behaviour in some circumstances.

# ActiveRecord: An after_commit transaction callback is not executed if any
# if the enclosing savepoints have been rolled back
ActiveRecord::Base.transaction do
  DB.transaction(savepoint: true) do
    DB.after_commit { puts "after commit" }
    raise Sequel::Rollback
  end
end
#>> BEGIN
#>> SAVEPOINT active_record_1
#>> ROLLBACK TO SAVEPOINT active_record_1
#>> COMMIT

# ActiveRecord: An after_commit transaction callback can be executed already
# after a savepoint is released, if the enclosing transaction is not joinable.
ActiveRecord::Base.transaction(joinable: false) do
  DB.transaction do
    DB.after_commit { puts "after savepoint release" }
  end
end
#>> BEGIN
#>> SAVEPOINT active_record_1
#>> RELEASE SAVEPOINT active_record_1
#>> after savepoint release
#>> COMMIT

Model

By default, the connection configuration will be read from ActiveRecord::Base. If you want to use connection configuration from a different model, you can can assign it to the database object after loading the extension:

class MyModel < ActiveRecord::Base
  connects_to database: { writing: :animals, reading: :animals_replica }
end

DB.activerecord_model = MyModel

Normalizing SQL logs

Active Record injects values into queries using bound variables, and displays them at the end of SQL logs:

SELECT accounts.* FROM accounts WHERE accounts.email = $1 LIMIT $2  [["email", "[email protected]"], ["LIMIT", 1]]

Sequel interpolates values into its queries, so by default its SQL logs include them inline:

SELECT accounts.* FROM accounts WHERE accounts.email = '[email protected]' LIMIT 1

If you want to normalize logs to group similar queries, or you want to protect sensitive data from being stored in the logs, you can use the sql_log_normalizer extension to remove literal strings and numbers from logged SQL queries:

Sequel.postgres(extensions: [:activerecord_connection, :sql_log_normalizer])

SELECT accounts.* FROM accounts WHERE accounts.email = ? LIMIT ?

Note that the sql_log_normalizer extension opens a database connection while it's being loaded. If you're setting up Sequel in a Rails initializer, you'll probably want to handle the database not existing, so that commands such as rails db:create continue to work.

DB = Sequel.postgres(extensions: :activerecord_connection)
begin
  DB.extension :sql_log_normalizer
rescue ActiveRecord::NoDatabaseError
end

Tests

You'll first want to run the rake tasks for setting up databases and users:

$ rake db_setup_postgres
$ rake db_setup_mysql

Then you can run the tests:

$ rake test

When you're done, you can delete the created databases and users:

$ rake db_teardown_postgres
$ rake db_teardown_mysql

Support

Please feel free to raise a new disucssion in Github issues, or search amongst the existing questions there.

License

The gem is available as open source under the terms of the MIT License.

Code of Conduct

Everyone interacting in this project's codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.