"David T. Lewis" <lewis at mail.msen.com> wrote:
> On Fri, Dec 21, 2001 at 10:27:08PM +0300, Daniel Vainsencher wrote:
> This is a very good idea, I like it. 
Hey David. Thanks.

> As long as it is OK to propose things
> that create work for other people, I have some suggestions ;)

Hey - no fair - you want to make suggestions, you're supposed to also
cough up some helpful code :-) In the spirit of Goran's brave commitment
(Go CSOTD!), give us a test, or implement at least one of the
suggestions...

> - It would be good if ParagraphEditor>>createTestFor: could suggest the
> right class name and selector to use. This would encourage consistency
> in naming, and would help folks who are not familiar with the naming
> conventions.
Attached a modified version that gives some helpful hints and offers to
create the class if it doesn't exist. Maybe someone can make it show a
list of subclasses?

> - This deserves its own 'make test case' menu choice.
Sure, I just really don't like doing UI integration under the current
scheme with ParagraphEditor and friends. Never quite got the hang of the
MVC/Morphic compatibility scheme. I'm hoping someone who does will take
it up. For my part, I factored most of the functionality into a small
class, so it's not too messy, and easy to call from wherever is the
right place.

> - Perhaps some creative person could do a UI that generates the test
> cases. I could imagine one pane in the UI which would contain a (possibly
> empty) list of expressions which should succeed, and another pane
> containing expressions which should fail. The end result might look
> like a "TestCaseBrowser" which could be used for browsing classes and
> interactively generating test cases.
If it could also interactively run your tests and show you how they
fail, that'd be very nice. A guy called Peter demoed an interesting
test-based development tool at ESUG last year, for one of the other
Smalltalk platforms.

I think SUnit is just begging to have a smarter tool suite built around
it - something to avoid bogging down in irrelevant, long running tests
that are certain not to succeed when your basic ParserTest isn't even
running, let alone passing, for example...
 
> This is very nice, thanks!

Daniel

from preamble:

"Change Set:		TestMaker
Date:			21 December 2001
Author:			Daniel Vainsencher

This changeset replaces print-it with a test creation utility. Select
something that raises an exception or returns the wrong answer, answer a
few questions, and voila, a new test.

If someone posts tests and it helps them do it, I'll polish this scrap
into an ENH..."!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: TestMaker.cs.gz
Type: application/octet-stream
Size: 1232 bytes
Desc: not available
Url : http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20011223/922a3763/TestMaker.cs.obj