Advice with Model tests

[Rodrigo Alvarez Fernández]

Hi!

I'm developing a rails applicaton with specify-before approach, with
nice results so far.

I use mocks and stubs in both Controller and views tests, but the main
issue is with model testing. Since I've seen many examples here and
there, I don't know if the right way is to use fixtures or not at all
(in model specs).

Atm, I'm using them, but I don't know if I should use mocks for
associations, all real fixture data...

This example http://rspec.rubyforge.org/documentation/rails/writing/models.html
uses both approachs (data generated on setup (but no mocks), and data
from fixtures).

Please, help me!

Apart from that, I really like this way of testing. I just want to do
it the right way.

Regards,
Rodrigo.
_______________________________________________
rspec-users mailing list
rspec-users@...
http://rubyforge.org/mailman/listinfo/rspec-users

[David Chelimsky-2]

On 12/21/06, Rodrigo Alvarez Fernández <papipo@...> wrote:

> Hi!
>
> I'm developing a rails applicaton with specify-before approach, with
> nice results so far.
>
> I use mocks and stubs in both Controller and views tests, but the main
> issue is with model testing. Since I've seen many examples here and
> there, I don't know if the right way is to use fixtures or not at all
> (in model specs).
>
> Atm, I'm using them, but I don't know if I should use mocks for
> associations, all real fixture data...
>
> This example http://rspec.rubyforge.org/documentation/rails/writing/models.html
> uses both approachs (data generated on setup (but no mocks), and data
> from fixtures).
>
> Please, help me!
>
> Apart from that, I really like this way of testing. I just want to do
> it the right way.

The "right" way doesn't really exist. It depends largely on the nature
and complexity of your app. That said, my personal preference is to
avoid fixtures and just create what I need in each spec. That's
because I like to see everything I need to understand the context.

This approach creates more duplication in specs, but less binding
between them. For me, the isolation is *usually* more important than
the duplication.

David


>
> Regards,
> Rodrigo.
> _______________________________________________
> rspec-users mailing list
> rspec-users@...
> http://rubyforge.org/mailman/listinfo/rspec-users
>
_______________________________________________
rspec-users mailing list
rspec-users@...
http://rubyforge.org/mailman/listinfo/rspec-users

[Rodrigo Alvarez Fernández]

On 12/21/06, David Chelimsky <dchelimsky@...> wrote:

> On 12/21/06, Rodrigo Alvarez Fernández <papipo@...> wrote:
> > Hi!
> >
> > I'm developing a rails applicaton with specify-before approach, with
> > nice results so far.
> >
> > I use mocks and stubs in both Controller and views tests, but the main
> > issue is with model testing. Since I've seen many examples here and
> > there, I don't know if the right way is to use fixtures or not at all
> > (in model specs).
> >
> > Atm, I'm using them, but I don't know if I should use mocks for
> > associations, all real fixture data...
> >
> > This example http://rspec.rubyforge.org/documentation/rails/writing/models.html
> > uses both approachs (data generated on setup (but no mocks), and data
> > from fixtures).
> >
> > Please, help me!
> >
> > Apart from that, I really like this way of testing. I just want to do
> > it the right way.
>
> The "right" way doesn't really exist. It depends largely on the nature
> and complexity of your app. That said, my personal preference is to
> avoid fixtures and just create what I need in each spec. That's
> because I like to see everything I need to understand the context.
>

What about models that need a lot of attributes? Do you create them in
the setup method?
I suppose that the answer is "yes".

> This approach creates more duplication in specs, but less binding
> between them. For me, the isolation is *usually* more important than
> the duplication.
>

Thinking about it... with fixtures the environment is shared between
specs and tries to satisfy all of them. In fact, you must write specs
that "fit" in the fixtures, or write/modify the current fixtures to
satisfy the new specs required environment.

It seems thar isolating is much better.

Thank you.

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

[Pat Maddox]

On 12/21/06, David Chelimsky <dchelimsky@...> wrote:

> On 12/21/06, Rodrigo Alvarez Fernández <papipo@...> wrote:
> > Hi!
> >
> > I'm developing a rails applicaton with specify-before approach, with
> > nice results so far.
> >
> > I use mocks and stubs in both Controller and views tests, but the main
> > issue is with model testing. Since I've seen many examples here and
> > there, I don't know if the right way is to use fixtures or not at all
> > (in model specs).
> >
> > Atm, I'm using them, but I don't know if I should use mocks for
> > associations, all real fixture data...
> >
> > This example http://rspec.rubyforge.org/documentation/rails/writing/models.html
> > uses both approachs (data generated on setup (but no mocks), and data
> > from fixtures).
> >
> > Please, help me!
> >
> > Apart from that, I really like this way of testing. I just want to do
> > it the right way.
>
> The "right" way doesn't really exist. It depends largely on the nature
> and complexity of your app. That said, my personal preference is to
> avoid fixtures and just create what I need in each spec. That's
> because I like to see everything I need to understand the context.

Out of curiosity, do you create AR records in your setup and save
them?  I almost always just create an object without saving it and
stub any finder methods.  I know AR is going to do its job and this
lets me hit the DB as little as possible, keeping my specs running
really fast.

Though sometimes I'm not sure if I'm using mocks and stubs too
much...trying to find a sweet spot.

I think I might write a tutorial on using RSpec to build Rails apps,
and then maybe you and Aslak could pick it apart and make
recommendations.  I could at least get started producing some
practical documentation for Rails though.

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

[David Chelimsky-2]

On 12/21/06, Pat Maddox <pergesu@...> wrote:

> On 12/21/06, David Chelimsky <dchelimsky@...> wrote:
> > On 12/21/06, Rodrigo Alvarez Fernández <papipo@...> wrote:
> > > Hi!
> > >
> > > I'm developing a rails applicaton with specify-before approach, with
> > > nice results so far.
> > >
> > > I use mocks and stubs in both Controller and views tests, but the main
> > > issue is with model testing. Since I've seen many examples here and
> > > there, I don't know if the right way is to use fixtures or not at all
> > > (in model specs).
> > >
> > > Atm, I'm using them, but I don't know if I should use mocks for
> > > associations, all real fixture data...
> > >
> > > This example http://rspec.rubyforge.org/documentation/rails/writing/models.html
> > > uses both approachs (data generated on setup (but no mocks), and data
> > > from fixtures).
> > >
> > > Please, help me!
> > >
> > > Apart from that, I really like this way of testing. I just want to do
> > > it the right way.
> >
> > The "right" way doesn't really exist. It depends largely on the nature
> > and complexity of your app. That said, my personal preference is to
> > avoid fixtures and just create what I need in each spec. That's
> > because I like to see everything I need to understand the context.
>
> Out of curiosity, do you create AR records in your setup and save
> them?

I only save them when saving them is necessary for the behaviour I'm
spec'ing. I'll also use fixtures on occasion for model classes that
have no interesting behaviour (i.e. validations, relationships, etc).

>I almost always just create an object without saving it and
> stub any finder methods.

If you're talking about non-model specs, I generally stub the finder
methods to return mocks (not instances of AR classes)

> I know AR is going to do its job and this
> lets me hit the DB as little as possible, keeping my specs running
> really fast.
>
> Though sometimes I'm not sure if I'm using mocks and stubs too
> much...trying to find a sweet spot.

That's the quest. And it's going to vary from app to app depending on
where the interesting behaviour exists. Sometimes it's more in the
models. Sometimes moreso in the controllers.

>
> I think I might write a tutorial on using RSpec to build Rails apps,
> and then maybe you and Aslak could pick it apart and make
> recommendations.  I could at least get started producing some
> practical documentation for Rails though.

That would be great! Thanks Pat.

David

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

[jshen]

"Since I've seen many examples here and
there, I don't know if the right way is to use fixtures or not at all
(in model specs)."

I've personally had a lot of problems with fixtures. One of my current rails apps uses test::unit and fixtures and the fixtures become like spaghetti code very fast. As your tests evolve over time it becomes very hard to manage your fixtures if they aren't super simple. When your writing a new test you don't know if you can add to a certain fixture association or not because it may break some other test in your app. If you delete a test that is no longer necessary you are unsure if you can delete the associated fixture data because some other test may depend on it.

I also agree with david in that it's nice to see everything in the context of the test.