Forum: RSpec Override tests (of a core rails app from a plugin)

Hello,

Is there some way to override a spec and more specifically parts of 
spec, ideally down to a specific example? I've been looking and googling 
for something like this, but I haven't found a good strategy for it.

The use-case is for a rails app that supports plugins, and those plugins 
may also tweak the behavior of core objects. For example (and it is a 
simplistic and bad example, but bear with me): Suppose a User object 
that has a boolean attribute show_mail_address that controls if the mail 
address of the user can be shown publicly, and that this defaults to 
true in the core app, but a security-savvy user of the software doesn't 
want that and creates a plugin to change the default to false. Running 
the core tests with the plugin will fail because the default won't be 
the one the tests expect. Is there any way to selectively override 
examples/tests/groups and/or extend tests (in the case the plugin adds 
functionality to the core object)?

Thanks,

Felix

On Oct 1, 2011, at 7:03 AM, Felix Schfer wrote:

> rspec-users mailing list
> rspec-users@rubyforge.org
> http://rubyforge.org/mailman/listinfo/rspec-users

Specs do not over-ride each other, so I can't think of an easy to do 
what you're trying to do. With that said, I have never come across a 
situation like this because usually the app has one or the other. 
Shouldn't the user delete their own code that your plugin accomplishes 
after they install it? Or, maybe your plugin overrides their core 
functionality, but doesn't provide tests - in which case it's up to the 
user to adjust their tests to your plugin.

Am 01.10.2011 um 22:17 schrieb Justin Ko:

> Specs do not over-ride each other

That's what I was able to gather so far, unfortunately.

> , so I can't think of an easy to do what you're trying to do. With that said, I 
have never come across a situation like this because usually the app has one or 
the other. Shouldn't the user delete their own code that your plugin accomplishes 
after they install it? Or, maybe your plugin overrides their core functionality, 
but doesn't provide tests - in which case it's up to the user to adjust their 
tests to your plugin.

The point is that a plugin can change core code but not core tests, 
making the core tests virtually useless. I'm aware that the preferred 
solution would probably be to test the core app and the plugins 
independently, or at least that the plugins only test what they change 
so that you could say "if the core tests without the plugin and the 
plugin tests with the plugin work, everything is fine", it doesn't work 
that well though. Changes introduced by plugins might slightly change 
other behaviors and have unintended (and thus not tested by the plugin's 
tests) side-effects which the core tests might catch, if those are known 
to break because of intentional changes to the core by the plugin, 
they're of no use.

Felix

I'm not sure what you mean by "core code" and "the user" here. Are you
writing an app, or a library, or a plugin? If you're writing an app, you
should test the app's behavior, which includes all the plugins and 
however
you configured them; if you're writing a library you should have 
scenarios
that test the various configuration options and basic functionality.

If you're writing a Rails plugin, you should test how that plugin works
inside apps. One way to do this is to generate some toy rails apps, 
stick
them deep inside your "spec" directory somewhere, have them include your
plugin and configure it in various ways, then test that the toy apps 
perform
as expected. (Technically those would be integration tests, not unit 
tests,
and you don't want to write too many of them if you can test the bulk of
your code through unit tests.)

you could say "if the core tests without the plugin and the plugin tests
> with the plugin work, everything is fine", it doesn't work that well though.


If your app is using a plugin, then why (and how) would you ever test 
your
app without the plugin?

On Sun, Oct 2, 2011 at 11:57 AM, Felix Schfer <f.schaefer@finn.de> 
wrote:
>
So, what you are saying is that you don't know how the core is
supposed to work after the user modifies it. So, you can't test that.
But, you need the core tests to prove that it works after they modify
it (Although, you don't know how it will work then.)

:-(

What you want seems like a logical impossibility to me.