John Napiorkowski > DBIx-Class-Migration-0.008 > DBIx::Class::Migration::FAQ


Annotate this POD


Open  3
View/Report Bugs
Source   Latest Release: DBIx-Class-Migration-0.058


DBIx::Class::Migration::FAQ - Answers to Frequently Asked Questions

When I run the commandline, I see a lot of debugging output ^

    SV = IV(0x7fb4e24dc858) at 0x7fb4e24dc868
      REFCNT = 1
      RV = 0x7fb4e0b0ddf8
        SV = PVHV(0x7fb4e24c48a0) at 0x7fb4e0b0ddf8
          REFCNT = 1
          STASH = 0x7fb4e2174a80        "DBI::db"
          ARRAY = 0x0
          KEYS = 0
          FILL = 0
          MAX = 7
          RITER = -1

There's some debugging code somewhere in the DBIx::Class::DeploymentHander dependency chain. We've looked and can't find it :( A case of beer at the next YAPC we meet at to whoever can figure it out.

ALthough you see this, there's nothing wrong. The command will work as in the documentation. The only issue is that if there's a lot of debugging scroll by, you might need to page up in your terminal to catch any real error messages.

Contributing ^

Contributing to the project is easy. First you'd fork the project over at Github (, clone the repo down to your work environment and install project dependencies:

    cpanm Dist::Zilla
    dzil authordeps | cpanm
    dzil listdeps | cpanm

You should first have a Modern Perl setup, including local::lib. If you need help installing Perl and getting started, please take a look at the Learning Perl website:

Dist::Zilla? ^

I realize Dist::Zilla seems to invoke feelings on nearly religious vigor (both for and against). After considering the options I felt using it, but carefully constraining the use of plugins to the default set was a better option than what I've done on other projects, which is to have a custom wrapper on top of Module::Install (you can peek at that if you want over here: I've decided I'd rather not continue to spend my limited time dealing with dependency and installation management tools, when there's a rational solution that many people already embrace available. The only other option is to continue to build and maintain code for this purpose that nobody else uses, and possibly nobody else understands.

If a better option with equal or greater maturity and community acceptance emerges, I'll entertain changing.

If the requirement of typing cpanm Dist::Zilla and using the dzil command line tool for a limited number of build jobs prevents you from contributing, I think you are unreasonably stubborn.

If you do contribute to the project, please be aware that I'm not likely to accept patches that include significant changes to the way I'm using Dist::Zilla, including using plugins to weave pod, automagically guess dependencies and generate tests. I'd prefer to stick to the most simple and standard Perl practices for building and installing code for the present.

I don't want my database sandbox files in my source control repository ^

Having the database sandboxes automatically created in the share directory is a nice feature, but can clutter your repository history. You really don't need that in the source control, since installing and controlling your database is something each developer who checks out the project should do.

If you are using git you can modify your .gitignore. If you sandbox is share/myapp-schema.db or (if you are using the mysql or postgresql sandboxes) share/myapp-schema/ you can add these lines to your .gitignore


Other source control systems have similiar approaches (recipies welcome for sharing).

What's the command to run migrations on an existing database? ^

    dbic-migration -Ilib status \
      --dsn="DBI:mysql:database=test;host=;port=5142" \
      --user msandbox \
      --password msandbox

Would be the general approach.

Error: "Failed to find share dir for dist..." ^

You didn't specify a custom --target_dir but forgot to make the /share directory in your project application root.

By default we expect to find a /share directory in your primary project root directory (contains your Makefile.PL or dist.ini, and the lib and t directories for example) where we will create migrations. At this time we can't automatically create this /share directory in the same way we can create all the migration files and directory for you. You need to create that directory yourself:

    mkdir share

Patches to fix this, or suggestions welcomed.

Can I use this even if I don't want to use DBIx::Class? ^

Not everyone loves using an ORM. Personally I've found DBIx::Class to be the only ORM that gets enough out of my way that I prefer it over plain SQL, and I highly recommend you give it a go. However if you don't want to, or can not convince your fellow programers (yet :) ), here's one way to still use this migrations and fixtures system. Strictly speaking, we are stilling using DBIx::Class behind the scenes, just you don't have to know about it or use it.

You use a subclass of DBIx::Class::Schema::Loader in a namespace for your application, like:

    package MyApp::Schema;

    use strict;
    use warnings;

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

    our $VERSION = 1;

    __PACKAGE__->loader_options( );


You'd put that in lib/MyApp/ along with your other application code. then just use MyApp::Schema as is discussed in the documentation. This will dynamically build a schema for you, as long as you set the schema arguments to connect to your actual database. Then everytime someone changes the database you just up the $VERSION and take it from there. Obviously this is a bit more manual effort, but at least you can have the ability to populate to any given version, manage fixtures, do some sane testing, etc. Maybe you'll even be able to convince people to try out DBIx::Class!

By the way, if you wanted, you can always dump the generated version of your classes using make_schema (see "make_schema" in DBIx::Class::Migrations and "make_schema" in DBIx::Class::Migrations::Script).

I am using MySQL and when the migration fails, it doesn't ROLLBACK ^

That's because MySQL does not support transaction DDL. Even if you have a transaction, MySQL will silently COMMIT when it bumps into some DDL.

I need to dump fixtures from a existing database ^

You don't always have the luxury of building a new database from the start. For example, you may have an existing database that you want to start to create migrations for. In this case you probably want to dump some data directly from that existing database in order to create fixtures for testing or for seeding a new database.

DBIx::Class::Migration will let you run the dump_all_sets and dump_named_sets commands against an unversioned database. For example:

    dbic-migration -Ilib -SMyApp::Schema dump_all_sets /
      --dsn="dbi:mysql:database=myapp;host=;port=3306" /
      --username johnn /
      --password $PASSWORD

In this case let's say "dbi:mysql:database=myapp;host=;port=3306" is a database that you've been managing manually and it has some data that is useful for creating your fixture sets.

When you run these commands against an unversioned database you will be warned because we have no way of being sure what version of the fixture sets you should be dumping. We will just assume that whatever the Schema version is, is correct. This can of course lead to bad or undumpable fixtures should your Schema and the unversioned DB not match properly. Buyer beware!

Error: "`' is not a module name" ^

Sorry this error is vague and I am working on a fix. You will get this if you have failed to provide a schema_class, either by setting it with the -S or -schema_class commandline option flag:

    dbic-migration -Ilib -SMyApp::Schema
    dbic-migration -Ilib --schema_class MyApp::Schema

or by exporting the %ENV:


Or, if you have a custom version of DBIx::Class::Migration::Script as discussed in the tutorial, you are not providing a good schema_class value.

Error: "Attribute (schema_class) does not pass the type constraint" ^

You probably forgot to include your project lib directory in the Perl search path. The easiest way to fix this is to use the I or lib command line: option flag:

    dbic-migration -Ilib -SMyApp::Schema [command]

When using the Postgresql Sandbox, I get "FATAL: could not create shared memory segment" ^

There will be more like this:

    FATAL:  could not create shared memory segment: Cannot allocate memory
    DETAIL:  Failed system call was shmget(key=1, size=2138112, 03600).
    HINT:  This error usually means that PostgreSQL's request for a shared memory
    segment exceeded available memory or swap space, or exceeded your kernel's 
    SHMALL parameter.  You can either reduce the request size or reconfigure 
    the kernel with larger SHMALL.  To reduce the request size (currently 2138112
    bytes), reduce PostgreSQL's shared memory usage, perhaps by reducing 
    shared_buffers or max_connections.

The solution is as presented. Since I prefer not to change my system settings permanently, " just add the following to a little bash script in my application

    sudo sysctl -w kern.sysv.shmall=65536
    sudo sysctl -w kern.sysv.shmmax=16777216

I don't know enough about Postgresql to know if the above settings are good, but they work for my testing. Corrections very welcome! Ideally I'd try to find a way to offer a patch to Test::postgresql, although this seems to be limited to people using Macs.


See DBIx::Class::Migration for author information


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

syntax highlighting: