Rails 4 alternate primary key with MySQL - mysql

Question

I have a strange problem I just cannot figure out.
I want to use the clustering ability of mysql to store related records beside each other on disk. Mysql clusters by the primary key on the table, which for a default rails model is ID.
However, for a lot of tables, it may make sense for the primary key of the table to be, for example, user_id, subscription_id, clustering the related records beside each, and making for a very efficient lookup when you ask the database for all of a user's subscriptions.
To do this, I created a mysql table like:
execute('create table subscriptions (
id integer not null auto_increment,
user_id integer not null,
feed_id integer not null,
folder_id integer,
created_at datetime,
updated_at datetime,
primary key (user_id, feed_id),
key id (id)
) engine=InnoDB default charset=utf8');
Notice that my PK is user_id, feed_id but I still have the ID column present, and I want rails to still use that as what it believes is the PK for the table.
First off, this didn't work at all until I set:
class Subscription < ActiveRecord::Base
self.primary_key = 'id'
...
end
Now comes the strange part.
When I run my tests, I get a strange error:
Mysql::Error: Field 'id' doesn't have a default value: INSERT INTO `subscriptions`
However - if I stick the application in development mode and do operations through the webpage, it works just fine.
After a lot of googling, I found a monkey patch to stop Rails setting MySQL into a stricter mode:
class ActiveRecord::ConnectionAdapters::MysqlAdapter
private
alias_method :configure_connection_without_strict_mode, :configure_connection
def configure_connection
configure_connection_without_strict_mode
strict_mode = "SQL_MODE=''"
execute("SET #{strict_mode}", :skip_logging)
end
end
If I add this, my test suit appears to work (for most tests, but not all), but any models that get created have an ID of zero.
Again in production mode, through the webpage things work just fine, and the models get an auto_increment ID as expected.
Has anyone got any ideas on what I can do this make my test suite work correctly in this setup?

Answer 1

I figured out what is going on.
What I did not remember, is that the development database is created by running migrations against the database, which also generates the schema.rb file. The schema.rb file is then used to load the test database.
So while my development database looked as I expected, the test database looked different - it would seem that the code which generates the schema.rb file cannot understand the database format I created and does not create a schema.rb that reflects my migrations correctly.
If I load my test database with:
$ rake db:migrate RAILS_ENV=test
And then run my test suite with:
$ rake test:all
Things work correctly. This is because the test:all task does not reload the database before running the test suite.
So what I described in the question to create an alternative primary key while maintaining the rails ID key works, except for the schema.rb part.