Greetings wise brothers. I seek your wisdom on how to be confident that all are different without the need to remember every part.

I am working DBIx::Class, and I would like to add a unique constraint of limited key length to a much longer varchar column

My DBIC schema definition code looks like this:

__PACKAGE__->table("classification_rules");
__PACKAGE__->add_columns(
    "id",
    { data_type => "integer", is_auto_increment => 1, is_nullable => 0
+ },
    ...
    "path",
    { data_type => "varchar", is_nullable => 0, size => 600 },
);

__PACKAGE__->set_primary_key("id");
__PACKAGE__->add_unique_constraint( ['path'] );
[download]

This generates MySQL like this:

CREATE TABLE `classification_rules` (
  `id` integer NOT NULL auto_increment,
...
  `path` text NOT NULL,
  PRIMARY KEY (`id`),
  UNIQUE `classification_rules_path` (`path`)
) ENGINE=InnoDB;
[download]

The problem is, that when I deploy, MySQL correctly reports that you can not have a unique constraint on a varchar that long without limiting the key length. What I would like to do, is have it generate SQL like UNIQUE `classification_rules_path` (path(250)) but I cannot work out how.

I have already stepped through the add_unique_constraint code in DBIx::Class::ResultSource to try to find an undocumented feature or hook I can use and also added breakpoints in SQL::Translator::Producer::MySQL to try and spot where the SQL is being generated from the other end.

Is there another solution to this? Is there a hook to add literal SQL to the definition for my table?

Thank you.