Given this discussion, I would like to encourage you to take a look at
Phexample by Adrian Kuhn and Niko Schwarz, based on previous work of Markus
Gaelli. The basic idea is that there is no distinction between setup and tests.
They are all examples, and examples are built on top of each other.
Phexample is completely compatible with SUnit. It also comes with a fancier
API, but even if you would like to debate this API, I believe it is a nice
thing that we should look more closely into.
Here are two examples checking an implementation of Stack. The first one
creates the stack and checks that it should be empty, and the second one checks
the pushing behavior:
ForExampleStack>>shouldBeEmpty
| stack |
stack := Stack new.
stack isEmpty should beTrue.
stack size should = 0.
^ stack
ForExampleStack>>shouldPushElement
| stack |
stack := self given: #shouldBeEmpty.
stack push: 42.
stack isEmpty should not beTrue.
stack size should = 1.
stack top should = 42.
^ stack
More information can be found here:
http://smalltalkthoughts.blogspot.com/2009/11/phexample-because-examples-expand-on.html
You can get it by simply:
Gofer new
squeaksource: 'phexample';
addPackage: 'Phexample';
load
Cheers,
Doru
On 24 Nov 2010, at 08:28, laurent laffont wrote:
> I agree with you Dennis. TDD practices are far ahead in Ruby / Java / Python
> world.
>
> BUT IMHO can be a real cool neaty great platform for TDD because we have:
> dynamic development.
>
> What I think we miss:
> - put a Mock framework in PharoCore - Mocks should be a standard tool even
> for low-level stuff
> - introduce pragmas to declare tests / setUp / tearDown methods so I can write
>
> AddressBookWithLaurentAndDennis>>createAddressBook
> <setUpTest>
> bla bla
>
> AddressBookWithLaurentAndDennis>>shouldHaveNumberOfContactsAnswersTwo
> <test>
> blabla
>
> AddressBookWithLaurentAndDennis>>ifDennisRemovedNumberOfContactsShouldAnswersOne
> <test>
> bla bla
>
>
>
> - (harder) interface to / creation of an executable requirements / BDD
> framework like Cucumber http://cukes.info/
>
> Laurent
>
>
> On Wed, Nov 24, 2010 at 4:31 AM, Dennis Schetinin <[email protected]> wrote:
> As for me, I'm simply unable to establish smooth test-driven process without
> Mocks. I think it does not depend on language at all.
>
> Without mocks I have to divide the task I have into smaller independent
> parts. It's actually an up-front design. I can make a mistake (I usually do);
> even worse it's often just impossible to accomplish because too many
> subcomponents usually interact thoroughly and I simply don't know which one
> to implement first :). So, without mocks I have to code bottom-up, from
> details to general. This is unnatural and error-prone.
>
> With mocks I can start directly from the task I have. I get a use case and
> write a test for the one. Creating the test I detect subcomponents needed,
> understand how they should interact and document that in the same test. As
> those subcomponents do not exist yet, I have to specify them and their
> behavior right in the test (if I don't want to break the process I've just
> started). That's why I really need mocks. Then I can take another use case or
> implement one of the mocks I used as a real object and so on. Thus, mocks
> allow me to go top-down and implement only the things I really need, not the
> things I think I'll need.
>
> Are there other, maybe better, approaches?
>
> Concerning TDD popularity… I'm not sure TDD plays the role it should in
> Smalltalk community.
>
> Messing with Smalltalk for about 7–8 years, I can't say I often see real TDD
> code. About 50–70% (depending on Smalltalk dialect) of frameworks I use have
> no tests at all. The rest often looks being tested a posteriori. Perhaps,
> Pharo is the only exclusion…
>
> And if TDD is really popular and widely used, why do we have so much
> regression bugs even in core subsystems even in commercial Smalltalks?
>
> Please, correct me if I'm wrong.
>
> 2010/11/23 laurent laffont <[email protected]>
>
> On Tue, Nov 23, 2010 at 5:39 PM, Luc Fabresse <[email protected]> wrote:
> Hi Dennis,
>
> Yes, TDD is highly popular and encouraged in the Smalltalk community.
> And mocks are not necessary in most cases.
>
>
> Hi Luc,
>
> why Mocks are less necessary in Smalltalk than in other
> (Ruby/Python/whatever...) languages ? Do you have some examples / link ?
>
> Sometimes I think a mocking library should be in PharoCore so I don't need to
> write my own mocks.....
>
>
> Cheers,
>
> Laurent
>
>
>
>
> #Luc
>
>
> 2010/11/23 Dennis Schetinin <[email protected]>
>
> How is it compared to Mocketry (http://www.squeaksource.com/Mocketry.html)?
> SMock?
>
> BTW, isn't TDD without mocks a bit incomplete? And is TDD popular among
> smalltalkers actually?
>
> 2010/11/22 Mariano Martinez Peck <[email protected]>
>
> This is cool Anindya. I don't need mocks right now, but I will keep it in
> mind!
>
> I cc'ed Pharo-dev mailing list.
>
> Cheers
>
> Mariano
>
> ---------- Forwarded message ----------
> From: Anindya Haldar <[email protected]>
> Date: Mon, Nov 22, 2010 at 10:11 AM
> Subject: [Pharo-users] TestMock: a full featured mock testing framework now
> available in SqueakSource
> To: [email protected]
>
>
> Dear Pharoers,
>
> TestMcok is a full featured mock based test framework that I am very glad to
> contribute to the community. It is in squeaksource under the project name
> TestMock. Please feel free to try it, and feedback will be greatly
> appreciated. The wiki section includes a brief how-to guide with code
> examples.
>
> Regards,
> Anindya Haldar
>
>
>
>
> _______________________________________________
> Pharo-users mailing list
> [email protected]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-users
>
>
>
>
>
> --
> Dennis Schetinin
>
>
>
>
>
> --
> Dennis Schetinin
>
--
www.tudorgirba.com
"In a world where everything is moving ever faster,
one might have better chances to win by moving slower."
