Future of Descent and D Eclipse IDE

[Bruno Medeiros]

On 15/06/2010 14:43, Ellery Newcomer wrote:
> On 06/15/2010 08:16 AM, Bruno Medeiros wrote:
>> On 03/06/2010 14:59, Ellery Newcomer wrote:
>>> On 06/03/2010 08:36 AM, Bruno Medeiros wrote:
>>>> On 28/05/2010 15:44, Ellery Newcomer wrote:
>>>>> On 05/28/2010 05:49 AM, Bruno Medeiros wrote:
>>>>>>
>>>>>> What do mean slow side? And what things can it parse that DMD
>>>>>> can't? (I
>>>>>> wasn't aware there were things DMD could not parse)
>>>>>>
>>>>>
>>>>> There's a few corner cases. eg:
>>>>>
>>>>> int delegate()[] i = [{while(true){}}];
>>>>>
>>>>> I think they're all in bugzilla.
>>>>>
>>>>> If you want a gauge on speed, it usually takes 10 to 15 seconds to
>>>>> parse
>>>>> through all of tango's code. (As an LL(k) parser, ANTLR requires
>>>>> lookahead in places that dmd doesn't.)
>>>>
>>>> BTW, what project is that parser for? Is it internal (like commercial)
>>>> or is open source? Is it listed here:
>>>> http://prowiki.org/wiki4d/wiki.cgi?GrammarParsers ?
>>>>
>>>
>>> Hm. Never knew about that page :)
>>>
>>> There might be an old version of it in antlrd; I'll check; but it's a
>>> project I started a year or two ago with the intent to eventually build
>>> an IDE with.
>>
>> Could you elaborate a bit more? I think I read in another post that your
>> parser targets the Java language, I didnt expect that, so the IDE you're
>> planning is in Java as well?? Would that be for a Netbeans based IDE, or
>> is it something commercial you are thinking? Basicly I'm wondering why
>> you wouldnt join the Descent project, if is in Java as well :)
>>
>
> Yeah, netbeans.
>
> Currently, my semantic analysis is all vaporware. The parser is slow,
> and incomplete (unicode: surrogate pairs - antlr sucks), and not yet set
> up for being amenable for things like syntax highlighting. And I have a
> habit of working feverishly on it and then leaving it to rot for months.
>
> I never joined Descent because I don't care much for eclipse and Ary's
> choice to port dmd didn't agree with me.
>
> But if you're going to be redesigning semantic from scratch, and you're
> going to leave the door open to port Descent to netbeans, sure, I'd love
> to contribute.

Porting the whole of Descent to Netbeans doesn't make any sense, one 
would need to port Eclipse to Netbeans... :P

But as for the semantic engine, it does make sense: my idea is not just 
to separate it from the rest of descent.core (like I mentioned 
previously), but also to make it independent of Eclipse. Thus it becomes 
trivially resuable in any other Java project like a Netbeans IDE.

Whether I'll try to rewrite the parser, or keep using DMD's parser, 
that's something that would have to be thought about later.


-- 
Bruno Medeiros - Software Engineer