# Re: Configuration/Initialization question

**URL:** <https://www.ruby-forum.com/t/re-configuration-initialization-question/113106>\
**Category:** Engines\
**Created:** [August 27, 2007, 6:51pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106 "2007-08-27T18:51:56Z")\
**Posts on this page:** 12\
**Page:** 1

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 27, 2007, 6:51pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/1 "2007-08-27T18:51:56Z")

</div>

I too am annoyed by having to restart my server frequently when editing  
plugin assets. You symlink suggestion seems like it would work, but I  
have a question.

Doesn’t it just end up copying all the plugin asset files on-top of  
themselves at server start? If so, wouldn’t it be better to symlink and  
then somehow disable the whole copy process all-together?

Trevor Rowe

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 28, 2007, 9:32am UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/2 "2007-08-28T09:32:20Z")

</div>

On Mon, Aug 27, 2007 at 09:50:37AM -0700, Trevor Rowe wrote:

> I too am annoyed by having to restart my server frequently when editing  
> plugin assets. You symlink suggestion seems like it would work, but I  
> have a question.
> 
> Doesn’t it just end up copying all the plugin asset files on-top of  
> themselves at server start? If so, wouldn’t it be better to symlink and  
> then somehow disable the whole copy process all-together?

I hadn’t really considered that. It wouldn’t be a problem if it did,  
IMHO, but let’s look at what it really does.  
See engines/lib/init.rb, at the bottom of the file in the method  
mirror\_files\_from:

```
source_files.each do |file|
  begin
    target = File.join(destination, file.gsub(source, ''))
    unless File.exist?(target) && FileUtils.identical?(file, target)
      FileUtils.cp(file, target)
    end
  rescue Exception => e
    raise "Could not copy #{file} to #{target}: \n" + e
  end
end

```

So you see, it will never actually copy the file because it is identical  
to itself. (this method is called in engines/lib/engines/plugin.rb by  
mirror\_public\_assets)

Regards,  
Peter B.  
Solide ICT - [http://www.solide-ict.nl](http://www.solide-ict.nl)

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 12:41pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/3 "2007-08-29T12:41:01Z")

</div>

Hi,

Is there a special trick to getting observers to work with engines?

Putting config.active\_record.observers = :my\_model\_observer in the  
Initializer.run block throws an unitialized constant MyModelObserver  
error.

The my\_model\_observer.rb file is in vendor/plugins/my\_plugin/app/models,  
alongside the model being observered, my\_model.rb.

Tekin

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 12:48pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/4 "2007-08-29T12:48:01Z")

</div>

On 8/29/07, Tekin S. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> Is there a special trick to getting observers to work with engines?

You should be able to do this - the observers aren’t processed until  
after plugins are loaded (check railties/lib/initializer.rb to be  
sure).

What this _might_ mean is that there’s some error in either your model  
or your observer file that’s causing a silent failure when the class  
is loaded. I saw this recently when a model depended on a file that  
wasn’t required until the bottom of environment.rb - because the  
observer caused the model to be loaded before the rest of  
environment.rb had been processed, but the code which the model  
depended on hadn’t been loaded yet, it would silently fall over and  
the problem looked like it was in the observer.

Double check that everything your model needs is required before it’s  
class definition - this might help.

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 12:57pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/5 "2007-08-29T12:57:02Z")

</div>

On Wed, Aug 29, 2007 at 11:47:39AM +0100, James A. wrote:

> What this _might_ mean is that there’s some error in either your model  
> or your observer file that’s causing a silent failure when the class  
> is loaded. I saw this recently when a model depended on a file that  
> wasn’t required until the bottom of environment.rb - because the  
> observer caused the model to be loaded before the rest of  
> environment.rb had been processed, but the code which the model  
> depended on hadn’t been loaded yet, it would silently fall over and  
> the problem looked like it was in the observer.

Yeah, Rails has the nasty habit of sometimes eating up exceptions.  
Then everything seems fine until much later when something fails for  
some inexplicable reason.

Sometimes running script/console reveals these errors by displaying  
a message somewhere. Other times, it does not. Try removing the  
observer loading statement and run script/console and refer to  
the class. That should trigger it to autoload, and if there’s an  
error it will show up. Rails only drops errors in some circumstances,  
not in others.

By the way: Please don’t start a new topic by replying to another one.  
This confuses mailclients that support threading. (replying causes  
your mailclient to set an In-Reply-To header which tells my client  
that this is a reply in the previous thread and it displays it as  
such)

Regards,  
Peter B.  
Solide ICT - [http://www.solide-ict.nl](http://www.solide-ict.nl)

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 1:01pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/6 "2007-08-29T13:01:02Z")

</div>

On 8/29/07, Peter B. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> Sometimes running script/console reveals these errors by displaying  
> a message somewhere. Other times, it does not. Try removing the  
> observer loading statement and run script/console and refer to  
> the class. That should trigger it to autoload, and if there’s an  
> error it will show up. Rails only drops errors in some circumstances,  
> not in others.

You’ll need to be careful in this case, because it’s the fact that the  
observer causes the model to be loaded _before_ the rest of  
environment.rb has been processed that could be causing the problem.

But, yes, script/console is your friend. You could also try commenting  
out anything after the end of the initializer block to see if your  
model has any hidden dependencies that it isn’t explicitly declaring.

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 1:27pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/7 "2007-08-29T13:27:09Z")

</div>

> > Try removing the observer loading statement and run script/console  
> > and refer to the class.
> 
> > That should trigger it to autoload, and if there’s an error it will  
> > show up.
> 
> > Rails only drops errors in some circumstances, not in others.

This was the first thing I tried, both the model and the observer load  
fine at the console.

> > By the way: Please don’t start a new topic by replying to another  
> > one.
> 
> > This confuses mailclients that support threading. (replying causes  
> > your mailclient to set an In-Reply-To header which tells my client  
> > that this is a reply
> 
> > in the previous thread and it displays it as such)

Apologies, will remember this in future.

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 1:24pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/8 "2007-08-29T13:24:05Z")

</div>

> > Double check that everything your model needs is required before it’s

> > class definition - this might help.

I tried again, clearing the model so it was an empty, vanilla  
ActiveRecord model and still got the error.

One thing that may or may not be unusual, the error is being thrown  
during the loading of the first plugin, which is not the one containing  
the observer or the observered model:

loading plugin from script/…/config/…/vendor/plugins/raid\_core with  
engine additions  
Exiting  
/usr/local/lib/ruby/gems/1.8/gems/activesupport-1.4.2/lib/active\_support  
/dependencies.rb:266:in `load\_missing\_constant’: uninitialized constant  
MyModelObserver (NameError)

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 2:14pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/9 "2007-08-29T14:14:22Z")

</div>

OK, so I thought I’d go back to basics, fresh empty rails app with a  
single empty engines enable plugin and still no joy.

With my\_model.rb and my\_model\_observer.rb in app/models, everything  
starts up fine. If I move them to vendor/plugins/testplugin/app/models,  
I get the same unitialized constant error.

I then tried it without specifying the plugin load order (config.plugins  
= [‘engines’, ‘\*’]) and there is no error, despite the plugins loading  
in the same order. If I explicitly specify the plugin (config.plugins =  
[‘engines’, ‘testplugin’]), it also works.

Same applies to my app, with the plugin explicitly listed in the config,  
it now works.

Is this expected behaviour?

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 1:40pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/10 "2007-08-29T13:40:13Z")

</div>

On 8/29/07, Tekin S. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> One thing that may or may not be unusual, the error is being thrown  
> during the loading of the first plugin, which is not the one containing  
> the observer or the observered model:
> 
> loading plugin from script/…/config/…/vendor/plugins/raid\_core with  
> engine additions  
> Exiting  
> /usr/local/lib/ruby/gems/1.8/gems/activesupport-1.4.2/lib/active\_support  
> /dependencies.rb:266:in `load\_missing\_constant’: uninitialized constant  
> MyModelObserver (NameError)

At this point, I’d probably recommend throwing puts statements around  
every line in the init.rb file of the plugin that causes the error, so  
we can trace which line of code is causing the actual problem…

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 4:43pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/11 "2007-08-29T16:43:56Z")

</div>

> > behaviour, the engines plugin actually defers loading of any plugins  
> > that weren’t specified
> 
> > manually by using the after\_initialize mechanism. You’re seeing a  
> > problem because, when
> 
> > using that star-syntax, at the point where Rails is trying to  
> > instantiate your observer your
> 
> > testplugin hasn’t been loaded yet.

OK, that makes sense, thanks.

---

<div class="post-metadata">

**Author:** ![Trevor\_Rowe](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/t/d78d45/32.png) [@Trevor\_Rowe](https://www.ruby-forum.com/u/Trevor_Rowe)\
**Post date:** [August 29, 2007, 3:28pm UTC](https://www.ruby-forum.com/t/re-configuration-initialization-question/113106/12 "2007-08-29T15:28:40Z")

</div>

On 8/29/07, Tekin S. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote:

> I then tried it without specifying the plugin load order (config.plugins  
> = [‘engines’, ‘\*’]) and there is no error, despite the plugins loading  
> in the same order. If I explicitly specify the plugin (config.plugins =  
> [‘engines’, ‘testplugin’]), it also works.
> 
> Is this expected behaviour?

Aaaaah. OK. Here’s what’s happening.

To support the

config.plugins = %w(engines \*)

behaviour, the engines plugin actually defers loading of any plugins  
that weren’t specified manually by using the after\_initialize  
mechanism. You’re seeing a problem because, when using that  
star-syntax, at the point where Rails is trying to instantiate your  
observer your testplugin hasn’t been loaded yet.

The solution, for the moment, is to just add your observer-containing  
plugins explicitly to config.plugins - that way they will definitely  
be available before Rails starts dealing with observers. Another  
solution would be to call YourObserverClass.instance at the bottom of  
environment.rb, rather than using config.observers - either will work  
fine.