What I understand
Suppose I have a class with a handy validation like:
User < ActiveRecord::Base
validates :username, :format => {/regex/}, :message => :name_format
end
In this case, I can use i18n
to make the error message translatable, by including the following in my /config/locals/en.yml
:
en:
activerecord:
errors:
models:
user:
attributes:
username:
name_format: 'has the way-wrong format, bro!'
This is fine and generally really handy.
What I want to know:
My question is: What happens when I have subclasses that inherit from User:
UserSubclassOne < User
# extra stuff
end
UserSubclassTwo < User
# extra stuff
end
...
UserSubclassEnn < User
# extra stuff
end
Now the problem is that Rails can't find the translation user_subclass_one.attributes.username.name_format
.
It complains:
translation missing:
en.activerecord.errors.models.user_subclass_one.attributes.username.name_format
I'd hope that Rails would look up the hierarchy of UserSubclassOne
to User
when searching for a string in en.yml
and then notice when it gets a 'hit', but (unless I've done something horribly wrong) apparently that doesn't happen.
An obvious solution is to duplicate the data in en.yml.en.errors.models
for user
, user_subclass_one
, user_subclass_two
, etc, but my Rails-sense tells me that this is deeply wrong.
Any ideas, folks?
Potential Complication:
User
is defined in a gem MyGem
that is included in a Rails engine MyEngine
that is included in the full-on Rails app MyApp
that defines UserSubclassOne
, ..., UserSubclassEnn
. I don't think this should matter though, since the validations are running in MyGem::User
, which is where the en.yml
file lives -- just wanted to let people know in case it does.
Ultimate problem/solution:
So it turns out that the problem was namespacing. Recall that MyApp
(which defines UserSubclassOne
) uses MyGem
(which defines User
). It turns out User
is actually in the namespace MyGem
(this is not necessarily always the case), so the full declaration line at the beginning of User
is not:
User < ActiveRecord::Base
but rather
MyGem::User < ActiveRecord::Base
.
When the i18n gem looks up the class hierarchy, it notices this namespace and searches for my_gem/user
, rather than simply user
, my_gem.user
, my_gem: user
, etc.
Thus I had to change my en.yml
file to:
/config/locals/en.yml
:
en:
activerecord:
errors:
models:
my_gem/user:
attributes:
username:
name_format: 'has the way-wrong format, bro!'
and bingo!
So it turns out that the problem was namespacing. Recall that
MyApp
(which definesUserSubclassOne
) usesMyGem
(which definesUser
). It turns outUser
is actually in the namespaceMyGem
(this is not necessarily always the case), so the full declaration line at the beginning ofUser
is not:but rather
.
When the i18n gem looks up the class hierarchy, it notices this namespace and searches for
my_gem/user
, rather than simplyuser
,my_gem.user
,my_gem: user
, etc.Thus I had to change my
en.yml
file to:/config/locals/en.yml
:and bingo!
According to the Rails Guides for i18n regarding Error Message Scopes (5.1.1) for Active Record validation error messages, what you're attempting to do should work:
So, in your case, assuming your classes look something like this:
app/models/user.rb
app/models/user_subclass.rb
and your config/locales/en.yml looks something like:
then the message searching for a validation on
UserSubClass
should go:Assuming that your model files and yaml files look similar to what's above, then the potential complication you mentioned may be the issue, but obviously I can't be certain.