NAME

DBIx::Class::Migration::Tutorial::SecondMigration - Upgrade the Database

GOAL

In this section you will change your database and learn how to create upgrade files. You will also create some custom fixture configurations and learn about creating seed data.

Changing Requirements

So far your MusicBase application is doing fine, but some new requirements come down. In addition to storing the Artist's name, we now want to store some geographical information, in this case a Country that can be associated with the Artist. Right now we are just going to care about Artists that can be located in one of the three following Countries:

    Canada
    Mexico
    USA

We'll need to seed that data to the database and also track it via some custom fixture configurations. That way if you need to start a new database from go you can get both the table structure and the essential running information (in this case the list of countries we care about could be considered system / domain data, not transactional data). Let's do it!

Change the Schema

We will need to add a new table and then link that table to Artist. From the project home directory (containing the dist.ini file) execute the following command in your terminal shell.

    touch lib/MusicBase/Schema/Result/Country.pm

Then open this file in your text editor of choice and add the following code:

    package MusicBase::Schema::Result::Country;

    use strict;
    use warnings;

    use base 'DBIx::Class::Core';

    __PACKAGE__->table('country');

    __PACKAGE__->add_columns(
      'country_id' => {
        data_type => 'integer',
      },
      'name' => {
        data_type => 'varchar',
        size => '96',
      });

    __PACKAGE__->set_primary_key('country_id');
    __PACKAGE__->add_unique_constraint(['name']);
    __PACKAGE__->has_many(
      'artist_rs' => "MusicBase::Schema::Result::Artist",
      {'foreign.country_fk'=>'self.country_id'});

    1;

This will establish a new Result class with a one to many relationship to the Artist Result class. You probably notice that we are referencing a column in the Artist Result class that does not yet exist. Lets add that as well

Open lib/MusicBase/Schema/Result/Artist.pm in your editor and change it to look like this:

    package MusicBase::Schema::Result::Artist;

    use strict;
    use warnings;

    use base 'DBIx::Class::Core';

    __PACKAGE__->table('artist');

    __PACKAGE__->add_columns(
      artist_id => {
        data_type => 'integer',
      },
      country_fk => {
        data_type => 'integer',
      },
      name => {
        data_type => 'varchar',
        size => '96',
      });

    __PACKAGE__->set_primary_key('artist_id');

    __PACKAGE__->belongs_to(
      'has_country' => 'MusicBase::Schema::Result::Country',
      {'foreign.country_id'=>'self.country_fk'});

    __PACKAGE__->has_many(
      'cd_rs' => 'MusicBase::Schema::Result::Cd',
      {'foreign.artist_fk'=>'self.artist_id'});

    1;

So we added the foreign key field country_fk and we added the other end of our new relationship to Country (called has_country).

Great, we've altered our schema to fit the new requirement. Let's up the $VERSION and commit if you are using a repository.

Alter file lib/MusicBase/Schema.pm

    package MusicBase::Schema;

    use strict;
    use warnings;

    use base 'DBIx::Class::Schema';

    our $VERSION = 2;

    __PACKAGE__->load_namespaces;

    1;

So we just change $VERSION from 1 to 2, which is how the default configuration of DBIx::Class::DeploymentHandler works.

Prepare the new migration

Now that we've altered the schema and upped the version, we use the prepare command to create the version 2 files. First, let's verify the status of our system:

    dbic-migration -Ilib status

Should return

    Schema is 2
    Deployed database is 1
    

That looks correct. Let's get the database up to date. First prepare the files.

    dbic-migration -Ilib prepare

That will generate a bunch of new files. Lets see the directory structure now.

    /share
      /fixtures
        /1
          (Same as before)
        /2
          /conf
            all_tables.json
      /migrations
        /_common
          (Same as before)
        /_source
          (system files for DBIC::DeploymentHandler.  Leave alone!)
        /SQLite
          /deploy
            /1
              (Same as before)
            /2
              001-auto-_VERSION.sql
              001-auto.sql
          /downgrade
            /2-1
              001-auto.sql
          /upgrade
            /1-2
              001-auto.sql
      musicbase-schema.db

Ok, we got a bunch more files. You can see that we've created a new all_tables.json fixture configuration, which if you peek inside, you will see has been updated to include your new Country Result class.

Additionally you can see we have a new version 2 directory under deploy which contains the full DDL for you new schema, as well as the special metadata table that DBIx::Class::DeploymentHandler uses to manage deployment history. You should take a quick look inside those as well, and see that the new country table has been added.

So far all this is similar to Version 1. What is different is you have the new downgrade and upgrade directories. For simplicity I will not deal with downgrades for this section of the tutorial, and instead focus on the upgrade path.

Lets take a closer look at /upgrade:

    /upgrade
      /1-2
        001-auto.sql

Since our schema was at version 2 and the database was at version 1, we created a 1 to 2 upgrade path, which will allow us to get the database in sync with our schema. DBIx::Class::DeploymentHandler will introspect your schema and database using SQL::Translator and try to suggest some DDL for this. You should treat this initial 001-auto.sql file as a suggestion and as a guide. You will need to make changes to it based on your data change need (SQL::Translator knows about your table structure, but not your data) as well as your performance and uptime needs. For example, when changing a table that has 1 million rows you might need to take an alternative approach than what is suggested.

This is the part of the job where you need to exercise the most planning and good judgment, and you might need to involve your DBA to help you craft a change set that can work for your particular needs. In this case we have a very small database using sqlite, so our needs are similarly small, however that should not excuse us from planning!

Let's look at the proposed DDL change. In the editor of your choice open share/migrations/SQLite/upgrade/1-2/001-auto.sql

    BEGIN;

    CREATE TABLE country (
      country_id INTEGER PRIMARY KEY NOT NULL,
      name varchar(96) NOT NULL
    );

    CREATE UNIQUE INDEX country_name ON country (name);

    ALTER TABLE artist ADD COLUMN country_fk integer NOT NULL;

    CREATE INDEX artist_idx_country_fk ON artist (country_fk);

    COMMIT;

So the first part of this that adds the country table is pretty straight forward. Lets break that part out into its own upgrade step. Generally even if the DDL change proposed is perfect, I prefer to move the code to a file name other than the default 001-auto.sql since if I need to prepare the upgrade several times (as you might if you are building a new version and realize you make a mistake and need to re prepare it, as we'll see in a later step) each time you do it will overwrite that file, blowing away any customization you made. So lets bust out the first part:

    touch share/migrations/SQLite/upgrade/1-2/001-add_country.sql

And then open it in your text editor and add the following (copied from the 001-auto.sql file).

    BEGIN;

    ;
    CREATE TABLE country (
      country_id INTEGER PRIMARY KEY NOT NULL,
      name varchar(96) NOT NULL
    );

    CREATE UNIQUE INDEX country_name ON country (name);

    ;
    COMMIT;

Great, now we have a step that makes a table. Next, we need to add the list of default countries. For this we will use a Perl script similar to the one we did for version 1. You could do this in SQL if you wanted, just if I can use a Perl script I would prefer that since it would be more portable across other databases (and eventually you will need something other than SQLite for production). Lets add that now:

    mkdir share/migrations/_common/upgrade
    mkdir share/migrations/_common/upgrade/1-2
    touch share/migrations/_common/upgrade/1-2/002-insert_countries.pl

We are putting this in the _common migrations directory so that later on if we add another database (such as MySQL) we'd be able have it run for both. Since Perl migration run files are going to be database agnostic, it makes sense to do this. You could have just as easily created a file:

    ## example, you don't need to do this!
    touch share/migrations/SQLite/upgrade/1-2/002-insert_countries.pl

And that would have made a script that would only run on SQLite installs.

Lets edit our Perl run file:

share/migrations/_common/upgrade/1-2/002-insert_countries.pl

    use strict;
    use warnings;
    use DBIx::Class::Migration::RunScript;

    migrate {
      shift->schema
        ->resultset('Country')
        ->populate([
          ['name'],
          ['Canada'],
          ['Mexico'],
          ['USA'],
      ]);
    };

Since the populate method uses bulk insertion, its generally my favored way to insert rows for migrations. Its going to be much faster than doing separate inserts. In this case we don't have a lot of data, so it didn't make a big difference, just that later on if you data needs are larger it would have an impact.

Ok, so let's look at the remaining part of the 001-auto.sql file that we have not yet examined:

    ALTER TABLE artist ADD COLUMN country_fk integer NOT NULL;

    CREATE INDEX artist_idx_country_fk ON artist (country_fk);

So that only does part of the job. First of all, we can't really add a NOT NULL column without default values, since that's going to be a constraint. Also the foreign key is missing. That's because it seems SQLite doesn't let you alter a table to add that kind of constraint. Let's look at the DDL for the new artist table from the full ddl in share/migrations/SQLite/deploy/2/001-auto.sql:

    CREATE TABLE artist (
      artist_id INTEGER PRIMARY KEY NOT NULL,
      country_fk integer NOT NULL,
      name varchar(96) NOT NULL,
      FOREIGN KEY(country_fk) REFERENCES country(country_id)
    );
    CREATE INDEX artist_idx_country_fk ON artist (country_fk);

Quite often when pondering a DDL change, it is valuable to glance at the full DDL that we give you in the deploy directory, so that you can understand better what the target is. In this case it is really clear the diff is only getting you partway there.

Alright, lets try to fix it. Lets create a file to hold our code:

    touch share/migrations/SQLite/upgrade/1-2/003-change_artist.sql

So this is going to be the third step for this upgrade (the second one was to add the country rows we need). Open the file in your text editor and enter:

    BEGIN;

    CREATE TEMPORARY TABLE artist_temp_alter (
      artist_id INTEGER PRIMARY KEY NOT NULL,
      name varchar(96) NOT NULL
    );

    INSERT INTO artist_temp_alter SELECT artist_id, name FROM artist;

    DROP TABLE artist;

    CREATE TABLE artist (
      artist_id INTEGER PRIMARY KEY NOT NULL,
      country_fk integer NOT NULL DEFAULT 1,
      name varchar(96) NOT NULL,
      FOREIGN KEY(country_fk) REFERENCES country(country_id)
    );

    CREATE INDEX artist_idx_country_fk ON artist (country_fk);

    INSERT INTO artist SELECT artist_id, 1, name FROM artist_temp_alter;

    DROP TABLE artist_temp_alter;

    COMMIT;

So this was my personal best shot, but of course there is going to be more than one way to do it. I decided it would be best to create a temporary table to hold the current artist info, build a new artist table with the correct FK constraint and new column, and then copy back from the temp table. Now, if I had a million rows this wouldn't work out very well, but with a database that big you'd probably have moved on to something that would actually allow you to add an FK in an alter statement. So I'll go with this since it does the job.

Notice that when I copy back to the new artist table from the temporary table I make all the artists live in the country matching country_id = 1 (Canada). I'll leave it this way for the section of the tutorial, just to keep it simple but in reality you would probably need to copy things more careful, or maybe add an "UNKNOWN" country option to the list of countries. We'll do an example of something like that in a later section. I just want to point out that when you craft your migration code you need to interpret the suggested DDL change and keep in mind how you data is put together. That way you can craft a good change set that keeps you database well constrained and organized.

Now that your are done converting the suggested DDL change, you should delete the generated auto file, otherwise when later you run the migration, you will end up doing more changes and get some errors.

    rm share/migrations/SQLite/upgrade/1-2/001-auto.sql 

Ok, so now you have a good set of migrations to move from version 1 to 2. Lets try that out now.

Upgrade the database

Once you have good migrations, its time to upgrade. Typically I first make sure the database is in a good known state:

    dbic-migration -Ilib drop_tables

That will blow away your tables, so be careful when using this command! You should see:

    Dropping table cd
    Dropping table track
    Dropping table artist
    Dropping table dbix_class_deploymenthandler_versions

Great, now the database is clear. Let's install version 1

    dbic-migration -Ilib install --to_version 1

Since the schema is version 2 we need to explicitly mention the target version we are deploying. Otherwise DBIx::Class::Migration will want to deploy a version that matches the current schema. We want to do an upgrade, not an install, since we need to modify both the tables AND our data, then we can build some good new fixtures. Lets do the upgrade:

    dbic-migration -Ilib upgrade
    dbic-migration -Ilib status

That should give you:

    Schema is 2
    Deployed database is 2

Let's peek in the database and see if everything is good:

    sqlite3 share/musicbase-schema.db
    sqlite> .tables

Should give you:

    artist                  dbix_class_deploymenthandler_versions
    cd                      track                                
    country   

So there's our new table. Lets make sure our data is correct:

    sqlite> select * from country;
    1|Canada
    2|Mexico
    3|USA

    sqlite> select * from artist;
    1|1|Michael Jackson
    2|1|Eminem

So looks like out migration worked out. Later on we'll learn how to write some test cases for our database, and I would highly suggestion that you write some tests that make sure your migration worked as expected, rather than doing the manual inspect as above

You are done with the upgrade, lets move on to make new fixtures. Exit the sqlite shell:

    sqlite> .q

Customize Fixtures and Serializing Fixtures

When you prepared the new version, in addition to creating the database structure migrations, we also got a new version 2 directory for the fixtures.

    /share
      /fixtures
        /1
          (Same as before)
        /2
          /conf
            all_tables.json

By default we build you a fresh all_tables.json that should reflect any new or removed tables. Additionally, if there were any custom fixture configurations, we would have copied those from the version 1. We didn't have any custom files so nothing was copied. Lets make a custom fixture config now:

    touch share/fixtures/2/conf/countries.json

We will make a fixture configuration to dump all the countries. Since the list of countries is part of your domain data (and not transaction data) you should create fixtures for it so that when someone is setting up a new version 2 database they can install a version that has both the structure and the data needed to have a functioning database. Data that is part of the domain we call 'seed' data. You'll need to create fixtures for all types of seed data, such as country lists, roles, product names, etc.

So you can use fixtures for a variety of jobs, from created demo databases to show clients, databases for developer, fixtures for running tests, and fixtures to preserve necessary domain data. You can then install a database and setup whatever list of fixtures is needed for the job at hand. Additionally since our migrations perform updates on the database data, we can just dump new fixtures after each clean update, and expect to have properly updated fixtures.

Here's our fixture configuration for share/fixtures/2/conf/countries.json:

    {
       "sets" : [
          {
             "quantity" : "all",
             "class" : "Country"
          }
       ],
       "might_have" : {
          "fetch" : 0
       },
       "belongs_to" : {
          "fetch" : 0
       },
       "has_many" : {
          "fetch" : 0
       }
    }

This will dump all the countries, and you will be able to load them later. You could review DBIx::Class::Fixtures for a better understanding of the way fixture rule sets are creating, but this is basically saying: "Dump everything in the Country source, and don't follow any relationships from Country or to Country."

Now that you have the new, custom fixture, lets dump all the sets:

    dbic-migration -Ilib dump_all_sets

Please note that if you 'played' with the database after doing the update, any new additional data will now become part of your fixtures. This may or may not be desirable. Let's see the list of new fixtures:

    /share
      /fixtures
        /1
          (Same as before)
        /2
          /all_tables
            _dumper_version
            /artist
              1.fix
              2.fix
            /cd
              (1-3.fix)
            /country
              (1-3.fix)
            /track
              (1-7.fix)
          /countries
            _dumper_version
            /country
              (1-3.fix)
          /conf
            all_tables.json
            countries.json

So now you have two sets of fixtures, the all_tables and the countries sets!

SUMMARY

    dbic-migration -Ilib status
    Schema is 2
    Deployed database is 2

You've successful crafted a migration to move your database structure and your data from version 1 to version 2. You've also updated your fixtures and created a custom fixture configuration for managing your seed data. You now have a good system where a new developer can walk in and run one or two commands to get a solid working database! Congrats.

NEXT STEPS

Proceed to DBIx::Class::Migration::Tutorial::Testing.

AUTHOR

See DBIx::Class::Migration for author information

COPYRIGHT & LICENSE

See DBIx::Class::Migration for copyright and license information