Hi everyone,

Right after the GSoC finished (I'm really sure I passed :)), I've just found more time to improve things drastically:

By using a nice combination of storing AST children both sorted by name in a dictionary and in normal list, I was able to increase the resolution performance extremely - that means it takes ~0.3 milliseconds instead of ~3ms to resolve a normal symbol like a local variable or a local class definition.
As a side effect, UFCS completion and preparation right after startup gained a maximum of performance boost! Tooltips and (later on) semantic highlighting could be done even more faster :)

Then I fixed several issues with template parameter deduction - and figured out several bugs or a kinda undefined behaviour of dmd:

class A(T) {
	class SubClass {}
}
class C(U: A!W.SubClass, W){}
class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?

Furthermore I fixed the parameter completion on this() and super() expressions, what implies a great usability bonus imho. Not to forget all the classic null-check issues and smaller bugs.



Cheers,
Alex

http://mono-d.alexanderbothe.com
https://github.com/aBothe/Mono-D/issues

alex wrote:
> Then I fixed several issues with template parameter deduction - and
> figured out several bugs or a kinda undefined behaviour of dmd:
>
> class A(T) {
>      class SubClass {}
> }
> class C(U: A!W.SubClass, W){}
> class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?

try marking SubClass with static keyword.

On Thursday, 23 August 2012 at 23:23:39 UTC, Piotr Szturmaj wrote:
> alex wrote:
>> Then I fixed several issues with template parameter deduction - and
>> figured out several bugs or a kinda undefined behaviour of dmd:
>>
>> class A(T) {
>>     class SubClass {}
>> }
>> class C(U: A!W.SubClass, W){}
>> class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?
>
> try marking SubClass with static keyword.

This works indeed, but nervertheless not for templated sub classes like
... { static class SubClass(U) {} }

http://i3.kym-cdn.com/entries/icons/original/000/004/128/BRILLIANT_.jpg

Great work as always, Alex :D

F i L wrote:
> http://i3.kym-cdn.com/entries/icons/original/000/004/128/BRILLIANT_.jpg
>
> Great work as always, Alex :D

Yes! This is really good project! :)

On Fri, 24 Aug 2012 01:18:25 +0200
"alex" <info@alexanderbothe.com> wrote:
> 
> class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?
> 

IIRC, I've had problems doing anything complex in a no-paren template parameter. I always figured if you're doing no-parens, it had to be a single token (Maybe I'm wrong?). Does it work if you do this?: A!(int.SubClass)

On 2012-08-24 01:18, alex wrote:
> Hi everyone,
>
> Right after the GSoC finished (I'm really sure I passed :)), I've just
> found more time to improve things drastically:
>
> By using a nice combination of storing AST children both sorted by name
> in a dictionary and in normal list, I was able to increase the
> resolution performance extremely - that means it takes ~0.3 milliseconds
> instead of ~3ms to resolve a normal symbol like a local variable or a
> local class definition.
> As a side effect, UFCS completion and preparation right after startup
> gained a maximum of performance boost! Tooltips and (later on) semantic
> highlighting could be done even more faster :)
>
> Then I fixed several issues with template parameter deduction - and
> figured out several bugs or a kinda undefined behaviour of dmd:
>
> class A(T) {
>      class SubClass {}
> }
> class C(U: A!W.SubClass, W){}
> class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?
>
> Furthermore I fixed the parameter completion on this() and super()
> expressions, what implies a great usability bonus imho. Not to forget
> all the classic null-check issues and smaller bugs.

That sounds great. I think it's time I give Mono-D a try.

-- 
/Jacob Carlborg

> IIRC, I've had problems doing anything complex in a no-paren template
> parameter. I always figured if you're doing no-parens, it had to be a
> single token (Maybe I'm wrong?). Does it work if you do this?:
> A!(int.SubClass)

primitive or identifier, iirc

On Friday, 24 August 2012 at 00:10:03 UTC, Piotr Szturmaj wrote:
> F i L wrote:
>> http://i3.kym-cdn.com/entries/icons/original/000/004/128/BRILLIANT_.jpg
>>
>> Great work as always, Alex :D
>
> Yes! This is really good project! :)

Thanks :D

On Friday, 24 August 2012 at 02:40:43 UTC, Nick Sabalausky wrote:
> On Fri, 24 Aug 2012 01:18:25 +0200
> "alex" <info@alexanderbothe.com> wrote:
>> 
>> class D : C!(A!int.SubClass) { } // is not allowed, but A!int is -- why?
>> 
>
> IIRC, I've had problems doing anything complex in a no-paren template
> parameter. I always figured if you're doing no-parens, it had to be a
> single token (Maybe I'm wrong?). Does it work if you do this?:
> A!(int.SubClass)

It's working just with the subclass marked as static and without any further template paramters, not with A!(int.SubClass) but with A!int.SubClass :)