Smalltalk › Pharo › Pharo Smalltalk Users

Conveying the usage of an object

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

8 messages Options

Sergio Fedi

Conveying the usage of an object

Hi,

In other smalltalks, which had the capability of marking a method with more than one category, I could make a method public/private AND it's functional category (accessor, printing, etc)

I did this to explicitly show the user which methods I expect to be sent to the object and which ones I don't.

It didn't seem to me I can assign more than one category in Pharo, so my question is:

How can I deal with this problem? Which is "the Pharo way" to accomplish this?

Mariano Martinez Peck

Re: Conveying the usage of an object

Sergio, a few times I also wanted that. And the only way of doing it I found was by doing defining protocols like 'public' , 'public - subcategoryX', 'public - subcategoryY' , etc... not sure powerful but simple.

Cheers,

On Mon, Apr 27, 2015 at 7:53 PM, Sergio Fedi <[hidden email]> wrote:

Hi,

In other smalltalks, which had the capability of marking a method with more than one category, I could make a method public/private AND it's functional category (accessor, printing, etc)

I did this to explicitly show the user which methods I expect to be sent to the object and which ones I don't.

It didn't seem to me I can assign more than one category in Pharo, so my question is:

How can I deal with this problem? Which is "the Pharo way" to accomplish this?

Mariano
http://marianopeck.wordpress.com

Sergio Fedi

Re: Conveying the usage of an object

The only other way I can think of doing this is via the virtual categories (which Pharo seems to have) and make them look for "Public" at the beggining of the first comment.

Not fancy, but practical.

Stephan Eggermont-3

Re: Conveying the usage of an object

In reply to this post by Sergio Fedi

On 28/04/15 00:53, Sergio Fedi wrote:
> How can I deal with this problem? Which is "the Pharo way" to accomplish
> this?

Uhm, you write unit tests for the public API? Showing what you expect to
sent to an object in a nice descriptive way?

Stephan

Sergio Fedi

Re: Conveying the usage of an object

I already do that.

(and I also make sure my tests have 100% coverage over the Public protocol)

But to know that you have to memorize the methods that are in the test cases.

Besides, I may want to test private methods too.

Stephan Eggermont-3

Re: Conveying the usage of an object

On 28/04/15 01:44, Sergio Fedi wrote:
> I already do that.
> (and I also make sure my tests have 100% coverage over the Public protocol)
>
> But to know that you have to memorize the methods that are in the test
> cases.
> Besides, I may want to test private methods too.

The categorization of methods is a weak way of showing the user,
as it is decoupled from the actual code, i.e. when I change the
behaviour, there is no automatic update of the category.

With Glamour/GTInspector extensions you could create an
API browser from e.g. marking API methods with a pragma
or using a TestCase subclass APITest

Stephan

jtuchel

Re: Conveying the usage of an object

Stephan,

isn't that a weak argument? If I change the bhavior of a method and forget to rename it to something intention revealing, there also is no guarantuee that the name tells a developer anything about it...

Joachim
> Stephan Eggermont <[hidden email]> hat am 28. April 2015 um 12:17 geschrieben:
>
>
> On 28/04/15 01:44, Sergio Fedi wrote:
> > I already do that.
> > (and I also make sure my tests have 100% coverage over the Public protocol)
> >
> > But to know that you have to memorize the methods that are in the test
> > cases.
> > Besides, I may want to test private methods too.
>
> The categorization of methods is a weak way of showing the user,
> as it is decoupled from the actual code, i.e. when I change the
> behaviour, there is no automatic update of the category.
>
> With Glamour/GTInspector extensions you could create an
> API browser from e.g. marking API methods with a pragma
> or using a TestCase subclass APITest
>
> Stephan
>
>
>

Stephan Eggermont-3

Re: Conveying the usage of an object

On 28/04/15 12:35, Joachim Tuchel wrote:
> isn't that a weak argument?

Yes, it is. And taking a look at the actual
categories gives me the impression that
naming is done better than categorization.

Stephan