Stefan Matthias Aust wrote:
> 
> Vassili,
> 
> I don't really care whether
>     [a > 0] assert.
> or
>     self assert: [a > 0].
> is used.  It's important to have that concept.  I added a trivial changeset
> to my update-proposals to allow both ways.   We can then simply could the
> occurrences of both alternatives and refactor towards to more often used
> alternative later.
> 
> With your approach, Dave's idea to add an descriptive text might be easier:
>     [a > 0] assertWith: 'illegal value for a'
> (I disliked the #assert:that: which has a negative assertion block)

Me too, I have possibly (don't know!) misinterpreted it in an earlier
mail...

>  Although,
> 
>     self failIf: [a > 0]
>     self failIf: [a > 0] with: 'illegal value for a'
> 
> might be alternatives.

Much more better!

And the variants
	self assert: [a < 0] failMessage: 'illegal value for a (should be < 0)'
and
	self failIf: [a >= 0] failMessage: 'illegal value for a (should be <
0)'
.

I would prefer to state the positive condition, which has to be
fulfilled, though.

But: It's always possible to write
	self assert: [(someCondition) not] failMessage: 'illegal value for
someCondition (should be false)'
, but it's difficult to read. In such a case I would also use the
>>failIf: variant.
		

Stephan

> 
> I don't see the necessarity to enable or recompile and disable assertions
> (I already mentioned the Pragmatic Programmer's point of view (a must-have
> book)) so IMHO we should for now go without that feature.  If others
> contradict, we should make your code an update.
> 
> bye

-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3