Aloha Pat!
I have solved the i18n issue. There were a couple of things to
consider. First, you have two init.rb files:
thinking-sphinx/init.rb and
thinking-sphinx/rails/init.rb
The content in both files is not the same.
In thinking-sphinx/init.rb you have:
require 'thinking_sphinx'
if Rails::VERSION::STRING.to_f < 2.1
ThinkingSphinx::Configuration.instance.load_models
end
if Rails::VERSION::STRING.to_f > 1.2
require 'action_controller/dispatcher'
ActionController::Dispatcher.to_prepare :thinking_sphinx do
ThinkingSphinx::Configuration.instance.load_models
end
end
in thinking-sphinx/rails/init.rb you have:
require 'thinking_sphinx'
require 'action_controller/dispatcher'
ActionController::Dispatcher.to_prepare :thinking_sphinx do
ThinkingSphinx::Configuration.instance.load_models
end
I would have expected the files to be the same.
I have identified a couple of ways to solve the problem. According to
the Rails Guide ( http://guides.rails.info/plugins.html )
"When rails loads plugins it looks for the
file named ‘init.rb’ or ‘rails/init.rb’."
However, if you put a couple of "raise" calls in "thinking-sphinx/
rails/init.rb" and "thinking-sphinx/init.rb" you will see that Rails
calls "thinking-sphinx/rails/init.rb" if it is present and then never
even calls "thinking-sphinx/init.rb".
If you are going to use "thinking-sphinx/rails/init.rb" then you need
this as its content:
require 'thinking_sphinx'
if Rails::VERSION::STRING.to_f < 2.1
ThinkingSphinx::Configuration.instance.load_models
end
if Rails::VERSION::STRING.to_f > 1.2
require 'action_controller/dispatcher'
ActionController::Dispatcher.to_prepare :thinking_sphinx do
I18n.backend.send(:init_translations) if
Rails::VERSION::STRING.to_f > 2.1
ThinkingSphinx::Configuration.instance.load_models
end
end
Note that your suggested change ( I18n.available_locales ) does not
work.
If you delete the "thinking-sphinx/rails/init.rb" file then no changes
are needed to "thinking-sphinx/init.rb" and the error doesn't even
happen (due to order in which things are loaded). However, the
convention when plugins can be used with multiple frameworks is to
include the 'rails/init.rb' file.
Therefore, my proposed fix is to change the content of both "thinking-
sphinx/rails/init.rb" and "thinking-sphinx/init.rb" to
require 'thinking_sphinx'
if Rails::VERSION::STRING.to_f < 2.1
ThinkingSphinx::Configuration.instance.load_models
end
if Rails::VERSION::STRING.to_f > 1.2
require 'action_controller/dispatcher'
ActionController::Dispatcher.to_prepare :thinking_sphinx do
I18n.backend.send(:init_translations) if
Rails::VERSION::STRING.to_f > 2.1
ThinkingSphinx::Configuration.instance.load_models
end
end
I have made these changes in my fork of TS. You are welcome to pull
from that if you like (1377ee8227dabf8b4917fc16e6586c024337196a).
Thanks to you and Karel for your help!
DrMark
On Apr 23, 8:42 pm, Pat Allan <[email protected]> wrote:
> On re-reading emails from RailsConf, I think I don't need to get those
> slides and such done until the actual conference, it's just materials
> attendees should have before the day. ie: Sphinx source and Thinking
> Sphinx. Much easier to take care of, so I've put that aside, and dug
> into the ActiveSupport source...
>
> While I18n loads the YAML files before Thinking Sphinx loads, it
> doesn't process them automatically. Lazy loading if you like, just
> like how Rails normally behaves with modesl... except then TS loads
> the model forcibly, without the YAML parsing being considered
> 'complete'.
>
> I guess one quick way of forcing it, in vendor/plugins/thinking_sphinx/
> rails/init.rb, add this line inside the block, before the load_models
> call:
> I18n.available_locales
>
> Which *should* force the full loading.
>
> It's not ideal, and I'd love some better ideas on how to approach
> this, but hopefully it'll do the job for the moment.
>
> --
> Pat
>
> On 24/04/2009, at 3:42 PM, Pat Allan wrote:
>
> > Hi Mark
>
> > Definitely sounds like something to do with the order of loading
> > (and yes, your demo app fails for me too). I'm *supposed* to be
> > sorting out my RailsConf tutorial content at the moment, so I really
> > shouldn't spend time digging into this at the moment.
>
> > But, a vague description of my approach: go hunting through Rails
> > source, find out when the environment is prepared (ie: when TS
> > manually reloads models), and when the I18n YAML file is parsed. I'm
> > assuming the former happens before the latter, and it really should
> > be the other way around... so if there's some later event hook we
> > can tie the TS loading to, that'd hopefully make things better.
>
> > --
> > Pat
>
> > On 24/04/2009, at 2:37 PM, DrMark wrote:
>
> >> Ok. I have identified an I18N issue caused by ThinkingSphinx and I
> >> would appreciate thoughts on what is causing the issue. It has driven
> >> me crazy all day. The problem wasn't obvious due to a large number of
> >> recent changes in my app, but I have finally tracked the problem to
> >> ThinkingSphinx.
>
> >> I use TS in my app and I use i18n extensively in a large app. I
> >> upgraded to the latest TS a short while ago (I was running my fork
> >> but
> >> switched to the main one). When running my unit tests today, many of
> >> my validation tests started failing with messages about "translation
> >> missing". This is odd because all of my tests had been passing.
>
> >> After many hours of debugging I have shown that TS is causing the
> >> problem. I have created a sample application that demonstrates the
> >> problem. (http://files.me.com/drmarklane/rrxgcm)
>
> >> In the sample app I have a simple User model with the following
> >> sample
> >> validations:
>
> >> validates_presence_of :first_name,
> >> :message => I18n.t
> >> ('users.errors.first_name_cant_be_blank')
> >> validates_format_of :last_name, :with=>/[^ \t\n\r\f\v ]/,
> >> :message=>"#{I18n.t
> >> ('users.errors.last_name_wrong_format')}"
>
> >> if you log in to the console then do:
> >>>> u = User.new
> >> => #<User id: nil, first_name: nil, last_name: nil, created_at: nil,
> >> updated_at: nil>
> >>>> u.valid?
> >> => false
> >>>> u.errors
> >> => #<ActiveRecord::Errors:0x21d0f2c @errors={"first_name"=>
> >> ["translation missing: en, users, errors, first_name_cant_be_blank"],
> >> "last_name"=>["translation missing: en, users, errors,
> >> last_name_wrong_format"]}, @base=#<User id: nil, first_name: nil,
> >> last_name: nil, created_at: nil, updated_at: nil>>
>
> >> Notice the "translation missing" errors in the errors hash. Any unit
> >> tests will also fail because of the translation missing errors.
> >> Adding
> >> default i18n settings in config.rb doesn't help.
>
> >> If you remove ThinkingSphinx, log out of the console, then repeat the
> >> above steps you will get:
>
> >>>> u = User.new
> >> => #<User id: nil, first_name: nil, last_name: nil, created_at: nil,
> >> updated_at: nil>
> >>>> u.valid?
> >> => false
> >>>> u.errors
> >> => #<ActiveRecord::Errors:0x2357da0 @errors={"first_name"=>["Your
> >> first name can't be blank"], "last_name"=>["Your first name is the
> >> wrong format"]}, @base=#<User id: nil, first_name: nil, last_name:
> >> nil, created_at: nil, updated_at: nil>>
>
> >> which is correct. Note that TS is not even called in the sample app.
> >> Simply placing TS in the plugins folder causes the problem.
>
> >> Does anyone have any ideas as to the cause of this issue? I think the
> >> issue is related to the class loading by TS. I would appreciate your
> >> advice.
>
> >> Thanks!
>
> >> DrMark
>
> >> PS. Pat, your copy of not-a-mock doesn't build unless you set the gem
> >> version of Rails to 2.2.2
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Thinking Sphinx" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/thinking-sphinx?hl=en
-~----------~----~----~----~------~----~------~--~---
