std.unittests for (final?) review

[Michel Fortin]

On 2011-01-06 08:44:33 -0500, Lutger Blijdestijn 
<lutger.blijdestijn at gmail.com> said:

> Michel Fortin wrote:
> 
>> The whole point of my proposal is to make the regular asserts print a
>> message similar to yours *by default*, when you don't specify any
>> message. This is possible because assert is not a function, it's a
>> language construct handled by the compiler. The compiler has access to
>> the whole expression tree, and it can rewrite the code to store
>> intermediate results in variables it can later give to the assertion
>> handler in case of failure.
> 
> I understood, but was reacting to the proposition that assertPred is not an
> improvement over the current situation, isn't that what you were saying?
> Just being pragmatic, it's unlikely such big changes to assert are going to
> land in dmd anytime soon. But if they are, yeah that would be cool.

assertPred is an improvement for error messages but it's not without a 
price. Basically you trade the very straightforward assertion syntax 
for a more verbose and less intuitive syntax. My argument is that this 
is an unnecessary tradeoff since we can just "fix" assert instead.

I think the consensus is that 'assert' can and should give us better 
error messages, but that assertPred is a good stopgap solution for most 
cases.


-- 
Michel Fortin
michel.fortin at michelf.com
http://michelf.com/