QUESTION: Is there any way to attach a listener that will fires when the
interpreter encounters an unknown Class and going to raise the
NameError?
For some reason all my classes are arranged the same way as files are
(i.e. A::C <=> A/B/C.rb, one to one)
So, theoretically i don’t really need any usage of the ‘require’
keyword.
My proposal is:
When the interpreter encounters unknown Class it raises the ‘NameError’.
I’m wondering is there any way to attach a listener right before it
happens (like method_missing)?
Another way:
begin
…
rescue NameError
load needed class
end
But, it’s not a solution, because in this case we loose the Point of
execution.
My proposal is:
When the interpreter encounters unknown Class it raises the ‘NameError’.
I’m wondering is there any way to attach a listener right before it
happens (like method_missing)?
Hi there. That’s a nice solution, I think. Thanks for sharing!
There might be some drawbacks in this approach, like now you cannot
require these files manually because it wouldn’t create the correct
module nesting, but for some applications I think your code is great.
Things that you probably could improve:
There’s a constant named TOPLEVEL_BINDING, you can use it instead of
your own constant.
You could backup the original const_missing and call it if the const
is really missing:
class Module
alias const_missing_orig const_missing
def const_missing(const)
…
else
const_missing_orig(const)
end
end
end
Also I think there might be a more elegant way to create the class
from the file in the correct nesting. There might be a way to do it
using some class_eval or something, but I’ll have to test it.
I’m hurry up, there are some bugs in current implementation. Do not use
it as it is.
Right now i’m using it in my project, so, stable solution should be
available, probably in one week :).
Not exactly understood what you mean, as i know ‘autoload’ works almost
the same as require - you should explicitly specify connection between
your classes and corresponding files.
So, instead of bunch of ‘require xxx’ there will be bunch of even longer
‘autoload xxx, xxx’ lines.
Class loader allows you to forget about it, just place your class file
wherever you want and it find it. It uses conventions to find your
classes and resources.
There is no much sense to make one or two files to be loaded by
ClassLoader, the point is - make the whole project loaded by it (and
forget about require) or don’t use it at all.
Maybe you add a little abstract how this is better than “autoload”.
Otherwise people might be left wondering why they should bother to add a
gem to their repository if they seem to get the same functionality out
of the box.
Not exactly understood what you mean, as i know ‘autoload’ works almost
the same as require - you should explicitly specify connection between
your classes and corresponding files.
So, instead of bunch of ‘require xxx’ there will be bunch of even longer
‘autoload xxx, xxx’ lines.
Not necessarily. Basically you need this only for the first constant
of a library. In the autoloaded file you can have more ‘autoload’
declarations. Or you require a single file which contains all the
initial autoload declarations and every subsequent file contains
further autoload declarations. If you provide this as a library then
usability for a user of that library is practically identical to your
approach - minus, you do not have to require another gem.
Class loader allows you to forget about it, just place your class file
wherever you want and it find it. It uses conventions to find your
classes and resources.
I know that approach from Java - and also all the issues that come
with automatic loading. You may end up loading other classes that you
intended to…
There is no much sense to make one or two files to be loaded by
ClassLoader, the point is - make the whole project loaded by it (and
forget about require) or don’t use it at all.
If I were to distribute a gem I would try to limit dependencies to
other gems. If I can achieve the same effect for my users with built
in features I would stick to them. I may be missing something here
but that’s the reason why I suggested you provide a short summary of
using your gem vs. built in features. If you want your gem to be used
you should probably do /some/ advertising.
Now i understood, you are absolutely right, explicit ‘require’ gives you
more control and less dependencies, and there can be some class conflict
problems with auto-reloading.
I think what approach to choose depends on what kind of projects you are
working on.
Thanks for your suggestion :), I’ll add comparison with standard
approach a bit later.
And one more thought - I also found it’s handy to use kind of IoC
container, so instead of thinking about file/class initialization and
dependency you just require it when you need it, here’s very basic
sample
class Logger
register_as :logger
end
class Application
inject :logger => :logger
def do_business
# now we can use injected component
logger.info 'done'
end
:C <=> A/B/C.rb, one to one)
:C
