On Wed, Jan 28, 2009 at 12:46 AM, Ary Borenszweig <ary@esperanto.org.ar> wrote:
> Ary Borenszweig wrote:
>
> Here's the video!
>
> http://www.youtube.com/watch?v=oAhrFQVnsrY
>
> :-)
>

By the way, in the vid you use a plain int param in your opApply delegate, instead of "ref int".  I think this will not work.  At least I seem to remember finding that opApply doesn't work unless I make all the delegate's arguments ref.

I think some of that came from an auto-complete code template, so if the code template doesn't include the 'ref' it should, as a hint to the programmer.

Another question -- I was wondering what it does for CTFE functions.
I'm guessing it evaluates them and spits out the result.  If so that
could be very very helpful.  Especially for code-building CTFE mixins.
 I don't think you had an example like that in the vid.

This compile time view could be a great debugging and learning aid for D.  I think it's quite exciting.  I wonder if someone can get it working on something like http://paste.dprogramming.com/

--bb

Ary Borenszweig wrote:

> ---
> int foo(int x) {
>   return x * 2 * 2;
> }
> 
> int bar(int x) {
>   return 2 * 2 * x;
> }
> ---
> 
> is transformed to this:
> 
> ---
> int foo(int x) {
>   return x * 2 * 2;
> }
> 
> int bar(int x) {
>   return 4 * x;
> }
> ---

Those should be shifts.

Bill Baxter escribió:
> On Wed, Jan 28, 2009 at 12:46 AM, Ary Borenszweig <ary@esperanto.org.ar> wrote:
>> Ary Borenszweig wrote:
>>
>> Here's the video!
>>
>> http://www.youtube.com/watch?v=oAhrFQVnsrY
>>
>> :-)
>>
> 
> By the way, in the vid you use a plain int param in your opApply
> delegate, instead of "ref int".  I think this will not work.  At least
> I seem to remember finding that opApply doesn't work unless I make all
> the delegate's arguments ref.
> 
> I think some of that came from an auto-complete code template, so if
> the code template doesn't include the 'ref' it should, as a hint to
> the programmer.

I'll correct that, then.

> 
> Another question -- I was wondering what it does for CTFE functions.
> I'm guessing it evaluates them and spits out the result.  If so that
> could be very very helpful.  Especially for code-building CTFE mixins.
>  I don't think you had an example like that in the vid.

It does! See the first part of the video, when I do:

int x = 1 + 2 + 3 + 4 + someFunc(5);

and it shows:

int x = 25;

It just evaluated someFunc. :-)

> 
> This compile time view could be a great debugging and learning aid for
> D.  I think it's quite exciting.  I wonder if someone can get it
> working on something like http://paste.dprogramming.com/

You mean, web? The java code can be used in the backend for that with little modification, I think that's quite possible.

> 
> --bb

On Wed, Jan 28, 2009 at 10:33 AM, Ary Borenszweig <ary@esperanto.org.ar> wrote:
> Bill Baxter escribió:
>>
>> On Wed, Jan 28, 2009 at 12:46 AM, Ary Borenszweig <ary@esperanto.org.ar> wrote:
>>>
>>> Ary Borenszweig wrote:
>>>
>>> Here's the video!
>>>
>>> http://www.youtube.com/watch?v=oAhrFQVnsrY
>>>
>>> :-)
>>>
>>
>> By the way, in the vid you use a plain int param in your opApply delegate, instead of "ref int".  I think this will not work.  At least I seem to remember finding that opApply doesn't work unless I make all the delegate's arguments ref.
>>
>> I think some of that came from an auto-complete code template, so if the code template doesn't include the 'ref' it should, as a hint to the programmer.
>
> I'll correct that, then.
>
>>
>> Another question -- I was wondering what it does for CTFE functions.
>> I'm guessing it evaluates them and spits out the result.  If so that
>> could be very very helpful.  Especially for code-building CTFE mixins.
>>  I don't think you had an example like that in the vid.
>
> It does! See the first part of the video, when I do:
>
> int x = 1 + 2 + 3 + 4 + someFunc(5);
>
> and it shows:
>
> int x = 25;
>
> It just evaluated someFunc. :-)

Rockin!


>> This compile time view could be a great debugging and learning aid for D.  I think it's quite exciting.  I wonder if someone can get it working on something like http://paste.dprogramming.com/
>
> You mean, web? The java code can be used in the backend for that with little modification, I think that's quite possible.

Yeh, web.  There's already a site where you can paste in D code and the server will compile and run the code for you.  A compile-time view button on that page would be great.  Not that you should be the one to do it or anything.  Just thinking it would be nifty, that's all.

--bb

Ary Borenszweig wrote:
> Bill Baxter escribió:
>> Another question -- I was wondering what it does for CTFE functions.
>> I'm guessing it evaluates them and spits out the result.  If so that
>> could be very very helpful.  Especially for code-building CTFE mixins.
>>  I don't think you had an example like that in the vid.
> 
> It does! See the first part of the video, when I do:
> 
> int x = 1 + 2 + 3 + 4 + someFunc(5);
> 
> and it shows:
> 
> int x = 25;
> 
> It just evaluated someFunc. :-)

Given that int x isn't a const expression, is someFunc(5) supposed to be evaluated at compile time like that?

Later,
Brad

Brad Roberts wrote:
> Ary Borenszweig wrote:
>> Bill Baxter escribió:
>>> Another question -- I was wondering what it does for CTFE functions.
>>> I'm guessing it evaluates them and spits out the result.  If so that
>>> could be very very helpful.  Especially for code-building CTFE mixins.
>>>  I don't think you had an example like that in the vid.
>> It does! See the first part of the video, when I do:
>>
>> int x = 1 + 2 + 3 + 4 + someFunc(5);
>>
>> and it shows:
>>
>> int x = 25;
>>
>> It just evaluated someFunc. :-)
> 
> Given that int x isn't a const expression, is someFunc(5) supposed to be evaluated at compile time like that?
> 
> Later,
> Brad

Hrm.. unless it's just bog standard inlining and const folding going on, which is quite possible.

Nevermind,
Brad

On Tue, 27 Jan 2009 21:19:36 +0200, Ary Borenszweig <ary@esperanto.org.ar> wrote:

> Where can I find a version of obj2asm for Windows to see what's going on in those cases?

Try the free version of IDA (Interactive DisAssembler) instead:
http://www.hex-rays.com/idapro/idadownfreeware.htm

-- 
Best regards,
 Vladimir                          mailto:thecybershadow@gmail.com

Brad Roberts escribió:
> Brad Roberts wrote:
>> Ary Borenszweig wrote:
>>> Bill Baxter escribió:
>>>> Another question -- I was wondering what it does for CTFE functions.
>>>> I'm guessing it evaluates them and spits out the result.  If so that
>>>> could be very very helpful.  Especially for code-building CTFE mixins.
>>>>  I don't think you had an example like that in the vid.
>>> It does! See the first part of the video, when I do:
>>>
>>> int x = 1 + 2 + 3 + 4 + someFunc(5);
>>>
>>> and it shows:
>>>
>>> int x = 25;
>>>
>>> It just evaluated someFunc. :-)
>> Given that int x isn't a const expression, is someFunc(5) supposed to be
>> evaluated at compile time like that?
>>
>> Later,
>> Brad
> 
> Hrm.. unless it's just bog standard inlining and const folding going on,
> which is quite possible.

Great observation! I remember functions like that not being evaluated unless the variable was declared const. I just debugged that and checked with DMD's source code, and it seems the expression in the initializer is evaluated if the variable declaration is not inside a function (check VarDeclaration::semantic2 and ExpInitializer::semantic, which invoked optimize at the end).

I wonder why is that behaviour defined like that...

> 
> Nevermind,
> Brad

Ary Borenszweig escribió:
> The Descent plugin for Eclipse provides an IDE for writing, launching and debugging code in D.
> 
> Explanations on how to get it from within Eclipse are here:
> 
> http://www.dsource.org/projects/descent
> 
> New features:
>  - Compile-time view (Window -> Show View -> Other -> D -> Compile-time View): allows you to see things from the compiler point of view, which applies some transformations to the source code. For example you can see what happens when you do a foreach, when you invoke an "extension method", when you do operator overloading; know what type has an auto variable; how struct fields are accessed?; compiler optimizations. It also removes conditionals that evaluate to false, and shows the results of mixins in-place. As always, this is far from perfect and a lot of things can be improved. For the lazies, I'll later upload a video about this. :-)
>  - Now hovering over a mixin (over the "mixin" keyword) shows it's result in a popup.
>  - Now hovering over a function or template instance with ctrl+shift shows it's compile-time result (with shift only: it's source code).

Just a small remark: the probability that these things stop working is bigger if the code is spread in many modules. That's because I did some optimizations and lazy loading of some symbols, and sometimes it doesn't work (I'll fix that, eveeeeeeentualy). So this should work better in a single module. That's why this should be a good aid for prototyping new functionality that heavily uses mixins, templates and compile-time evaluation. :-)

Ary Borenszweig wrote:

> The Descent plugin for Eclipse provides an IDE for writing, launching and debugging code in D.

This build is not compatible with Eclipse3.3 und SuseEnt10(SP2) as well. When I press CTRL+Space, I will see an error message:

----------------------------------------------------------------------------
The 'descent.ui.JavaNoTypeCompletionProposalComputer' proposal computer from the 'descent.ui' plug-in did not complete normally. The extension has thrown a runtime exception.

To avoid this message, disable the 'descent.ui' plug-in or disable the 'Other D Proposals' category on the content assist preference page.
----------------------------------------------------------------------------

There is no problem with Windows, but with SuseEnt10.
Is it possible to make it work again??

For more information, please another thread of mine in this group.

--Qian