Use default collation

[dereuromark]

This is a (multiple allowed):

• bug

• enhancement

• feature-discussion (RFC)

• CakePHP Version: 3.4 dev (dev-3.next)

• Migrations plugin version: dev-master

• Database server (MySQL, SQLite, Postgres): MySQL 5.6

• PHP Version: 5.6

What you did

utf8mb4_unicode_ci is the default database collation.
I have a migration file including new tables with a normal setup:

        $this->table('countries')
            ->addColumn('name', 'string', [
                'default' => null,
                'limit' => 40,
                'null' => false,
            ])
            ...
            ->create();

The datasources connection is also set up as

'encoding' => 'utf8mb4',

What happened

It generated tables with "utf8_general_ci".

Expected Behavior

It should have used the connection/database default "utf8mb4_unicode_ci".

[HavokInspiration]

The main concern I have is that it will generate migrations with a collation that is not "universal" between database vendor, making the migration "not universal".
But looking at phinx adapter classes, apparently only Mysql and SqlServer has support for collation on table creation.

Since we have our own Table class (extending the Table class from Phinx) we could very well fetch the collation from the datasource connection config and inject it in the table options before creating it.
So it might be doable.

There's just the concern I outlined at the beginning of the post to take into consideration.

[HavokInspiration]

Well actually, no, if we get the encoding from the datasource configuration and inject it, there's nothing in the migrations file, it will be hidden from the user and everything will be running in the background. So there might not be a problem after all.

[dereuromark]

Well, the problem is there, but it might not be an issue of the plugin, but the connection to the database itself, or the default setting inside the DB.
So maybe this issue is out of scope for the plugin to solve.

[HavokInspiration]

No, it comes from the MySQL adapter of phinx which defined utf8_general_ci as default collation when creating a table (see this link)

But when we create the table we can fetch the encoding from the datasource that is migrating the schema and inject it to the table object options in order to force the adapter to run using the good encoding. This needs some testing though but it seems doable on paper.