I have a Padrino App called Gusy that specifies (Sequel) Models like
# gusy/models/seminar.rb
class Seminar < Sequel::Model
# hopefully irrelevant stuff defined here
end
I want to access this Model from either a second gem, or a script in bin/
.
Now, e.g. I require Gusy from a second gem "gusy_fill". The Gemfile is put up to set the path to a Gusy git repository. I can successfullly see the Gusy namespace (and e.g. print the apps Version Gusy::VERSION
) if interactively exploring with bundle console
.
How can I access the mapped models and where and how do I configure the database connection?
I see nothing relevant in the Padrino::
or Gusy::
modules.
An irb session might look like this:
require 'gusy'
Gusy::Seminar.create(:name => 'from gusy_fill' # => NameError: uninitialized constant Gusy::Seminar
I want to achieve this without creating a second Padrino App that mounts Gusy (for that, pointers are included in the generated gusy/README.md).
As initially statet, I would have the very same issue, if I would do what I want from within the same app: write a small script in gusy/bin
that talks with the database, really in the setting like when calling padrino console
.
Sorry to hear you're having trouble with this. It's good that you brought it up though because I've been trying to put my thoughts around the subject for a while now and this pushed me into it :). I've prepared a repo for you explaining how to do it with what we have now in Padrino.
The README (which I'm pasting afterwards), explains the reasoning behind it and puts some questions up for us to think about the way we've implemented them. I'd love to hear your thoughts about it :).
Gemified apps in Padrino
This repo intends to answer
How to access Padrino model and database in a “standalon” (bin/) script? and
How to access a gemified Padrino Apps Model from other gem that requires that App.
The issue
In short, there are two issues of the similar nature, both related to models defined in the gemified app:
- they need to be accessed from another gems/projects;
- they need to be accessed from the gemified app's
bin
, doing something else other than starting
the Padrino server.
The example
First there's gemified-app
. That's a Padrino app that is gemified. It also contains a model
called SomeModel
that has one field called property
.
Then there's access-gemified-app-without-padrino
; a ruby script that loads the gemified app to
access the model.
Finally, there's another-app
which is a regular Padrino app that just loads gemified-app
to use
its model (SomeModel
).
Problems with the current Padrino setup
Creating an app with padrino g project gemified-app --orm sequel --gem --tiny
will give you the
following gemspec
:
# -*- encoding: utf-8 -*-
require File.expand_path('../lib/gemified-app/version', __FILE__)
Gem::Specification.new do |gem|
gem.authors = ["Darío Javier Cravero"]
gem.email = ["dario@uxtemple.com"]
gem.description = %q{Padrino gemified app example}
gem.summary = %q{Padrino gemified app example}
gem.homepage = ""
gem.files = `git ls-files`.split($\)
gem.executables = gem.files.grep(%r{^bin/}).map{ |f| File.basename(f) }
gem.test_files = gem.files.grep(%r{^(test|spec|features)/})
gem.name = "gemified-app"
gem.require_paths = ["lib", "app"]
gem.version = GemifiedApp::VERSION
gem.add_dependency 'padrino-core'
end
The key points are gem.require_paths = ["lib", "app"]
and gem.add_dependency 'padrino-core'
.
gem.require_paths = ["lib", "app"]
explains why models/some_model.rb
isn't available when we
load the gem somewhere else. It simple isn't added to $LOAD_PATH
:(.
gem.add_dependency 'padrino-core'
hints us that something might be missing later on. What happens
with dependencies like the ORM or the renderer? Should we load those? I reckon that it's a matter
of what you want to achieve but I'd say that most times yes.
Our gemified app dependencies are still listed in our Gemfile
which will only be added in the
current scope and not in any gems requiring our gemified-app
gem.
A first attempt at solving this
For this to work there are two things we should do:
Add 'models'
to gem.require_paths = ["lib", "app"]
so that it becomes:
gem.require_paths = ["lib", "app", "models"]
.
That will make sure that anything inside the gemified-app/models
directory is included in your
gem.
To make it easier to test this, we'll use bundler
and in our access-gemified-app-without-padrino
test script we'll add a Gemfile
that looks like this:
source 'https://rubygems.org'
gem 'gemified-app', path: '../gemified-app'
gem 'pry'
Now in your new app, go to the REPL bundle exec pry
and try to require 'gemified-app'
.
Then try SomeModel.all
. It will fail. Why? Because you didn't require 'some_model'
.
It will still not work if you do that though. Why? Because none of the model's dependencies,
i.e. sequel
and sqlite3
(not a direct dependency but it is through the connection) are loaded.
Here you have two choices: you load them manually on your Gemfile
or you define them as
dependencies on gemified-app.gemspec
.
I regard the latter one as a better choice since you're already including the model and you're
expecting its dependencies to come with it. It would like this:
# gemified-app/gemified-app.gemspec
# ...
gem.add_dependency 'padrino-core'
gem.add_dependency 'padrino-helpers'
gem.add_dependency 'slim'
gem.add_dependency 'sqlite3'
gem.add_dependency 'sequel'
gem.add_development_dependency 'rake'
# ...
# gemified-app/Gemfile
source 'https://rubygems.org'
# Distribute your app as a gem
gemspec
You would have to explicitly include all the gems you will need. This may seem cumbersome but in
all fairness it gives you a greater understanding of what your app needs. Eventually you will
realise you don't even need bundler and the Gemfile :).
Alright, so, go ahead launch your REPL and type require 'gemified-app'
and require 'some_model'
.
Then try SomeModel.all
. And... It will fail :(. Why? Because Sequel::Base
isn't defined. Now you might be wondering:
what happened to the reference to sequel
I put in my gemified-app.gemspec
? Well, it's just that:
a reference and it won't require the gem for you.
This won't happen with Padrino either because we're using
require 'rubygems' unless defined?(Gem)
require 'bundler/setup'
Bundler.require(:default, RACK_ENV)
in our config/boot.rb
and that only loads required gems on our Gemfile
.
So the question is... Should we load that manually? And if so, where?
Well, since this is a gem itself, I believe that the best place to do so would be in lib/gemified-app.rb
.
Loading all the gems needed will make this file look like:
require 'padrino-core'
require 'padrino-helpers'
require 'slim'
require 'sqlite3'
require 'sequel'
module GemifiedApp
extend Padrino::Module
gem! "gemified-app"
end
Alright, so we're all set... Back to the REPL, do your requires
require 'gemified-app'
require 'some_model'
and try SomeModel.all
. And... It will fail :(. Again! :/ Why? Because there's no connection to the
database. Padrino was loading this for us through config/database.rb
.
Another question arises... Should we include config/database.rb
in the gem too?
The way I see it, we shouldn't. The way I see it, the database connection is something every app
should locally define as it may contain specific credentials to access it or stuff like that.
Our sample, access-gemified-app-without-padrino/do-somethin.rb
script will then look like this:
require 'gemified-app'
Sequel::Model.plugin(:schema)
Sequel::Model.raise_on_save_failure = false # Do not throw exceptions on failure
Sequel::Model.db = Sequel.connect("sqlite:///" + File.expand_path('../../gemified-app/db/gemified_app_development.db', __FILE__), :loggers => [logger])
require 'some_model'
SomeModel.all.each do |model|
puts %Q[#{model.id}: #{model.property}]
end
Yes, the connection code is pretty much the same than our Padrino app and we're reusing its database
for this example.
That was some ride :) but we finally made it. See the sample apps in the repo for some working
examples.
require some_model
:/
I don't know you but I don't like that at all. Having to do something like that means that I
really have to pick my models' names very carefully not to clash with anything I may want to use
in the future.
I reckon that modules are the answer to it but that's the current state of affairs. See the
conclusion for more on this.
An alternative approach
Separate your model layer into its own gem and require it from your (gemified or not) Padrino app.
This might probably be the cleanest as you can isolate tests for your models and even create
different models for different situations that may or may not use the same database underneath.
It could also encapsulate all of the connection details.
Conclusion
I think we should review Padrino's approach to gemified apps.
Should we use the gemspec instead of the Gemfile for hard dependencies?
Should we namespace the models (I know we had some issues in the past with this)?
Should we teach users to do explicit requires in their gems or to inspect the dependecies and
require them for them?
Should we teach our users how to load their dependencies and be more reponsible about it? At the end
of the day, if they went the gemified app route they are clearly much more proficient in Ruby and
should be aware of this kind of stuff.
Thoughts? :)