Seed Fu is an attempt to once and for all solve the problem of inserting and maintaining seed data in a database. It uses a variety of techniques gathered from various places around the web and combines them to create what is hopefully the most robust seed data system around.
Version 2.0.0 of Seed Fu introduced API changes. Seed::Writer
was been completely overhauled and will require you to update your scripts. Some other deprecations were introduced, and support is fully removed in version 2.1.0. Please see the CHANGELOG for details.
The API documentation is available in full at http://rubydoc.info/github/mbleigh/seed-fu/master/frames.
User.seed do |s|
s.id = 1
s.login = "jon"
s.email = "[email protected]"
s.name = "Jon"
end
User.seed do |s|
s.id = 2
s.login = "emily"
s.email = "[email protected]"
s.name = "Emily"
end
$ rake db:seed_fu
== Seed from /path/to/app/db/fixtures/users.rb
- User {:id=>1, :login=>"jon", :email=>"[email protected]", :name=>"Jon"}
- User {:id=>2, :login=>"emily", :email=>"[email protected]", :name=>"Emily"}
Just add gem 'seed-fu', '~> 2.1.0'
to your Gemfile
Seed Fu depends on Active Record, but doesn't have to be used with a full Rails app. Simply load and require the seed-fu
gem and you're set.
The current version is not backwards compatible with Rails 3.0. Please use gem 'seed-fu', '~> 2.0.0'
.
The current version is not backwards compatible with Rails 2.3. Please use gem 'seed-fu', '~> 1.2.0'
.
Constraints are used to identify seeds, so that they can be updated if necessary. For example:
Point.seed(:x, :y) do |s|
s.x = 4
s.y = 7
s.name = "Home"
end
The first time this seed is loaded, a Point
record will be created. Now suppose the name is changed:
Point.seed(:x, :y) do |s|
s.x = 4
s.y = 7
s.name = "Work"
end
When this is run, Seed Fu will look for a Point
based on the :x
and :y
constraints provided. It will see that a matching Point
already exists and so update its attributes rather than create a new record.
If you do not want seeds to be updated after they have been created, use seed_once
:
Point.seed_once(:x, :y) do |s|
s.x = 4
s.y = 7
s.name = "Home"
end
The default constraint just checks the id
of the record.
By default, seed files are looked for in the following locations:
Rails.root/db/fixtures
andRails.root/db/fixtures/Rails.env
in a Rails appdb/fixtures
when loaded without Rails
You can change these defaults by modifying the SeedFu.fixture_paths
array.
Seed files can be named whatever you like, and are loaded in alphabetical order.
When loading lots of records, the above block-based syntax can be quite verbose. You can use the following instead:
User.seed(:id,
{ :id => 1, :login => "jon", :email => "[email protected]", :name => "Jon" },
{ :id => 2, :login => "emily", :email => "[email protected]", :name => "Emily" }
)
Seed files can be run automatically using rake db:seed_fu
. There are two options which you can pass:
rake db:seed_fu FIXTURE_PATH=path/to/fixtures
-- Where to find the fixturesrake db:seed_fu FILTER=users,articles
-- Only run seed files with a filename matching theFILTER
You can also do a similar thing in your code by calling SeedFu.seed(fixture_paths, filter)
.
To disable output from Seed Fu, set SeedFu.quiet = true
.
Seed files can be huge. To handle large files (over a million rows), try these tricks:
- Gzip your fixtures. Seed Fu will read .rb.gz files happily.
gzip -9
gives the best compression, and with Seed Fu's repetitive syntax, a 160M file can shrink to 16M. - Add lines reading
# BREAK EVAL
in your big fixtures, and Seed Fu will avoid loading the whole file into memory. If you useSeedFu::Writer
, these breaks are built into your generated fixtures. - Load a single fixture at a time with the
FILTER
environment variable - If you don't need Seed Fu's ability to update seed with new data, then you may find that activerecord-import is faster
If you need to programmatically generate seed files, for example to convert a CSV file into a seed file, then you can use SeedFu::Writer
.
Please report them on Github Issues.
- Michael Bleigh is the original author
- Jon Leighton is the current maintainer
- Thanks to Matthew Beale for his great work in adding the writer, making it faster and better.
Copyright © 2008-2010 Michael Bleigh, released under the MIT license