flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christofer Dutz <christofer.d...@c-ware.de>
Subject AW: AW: AW: Falcon and Antlr4
Date Tue, 22 Jul 2014 19:07:25 GMT
The way I understood JBurg is that it seems to have been designed to manipulate ASTs produced
by Antlr2 and Antlr3.
Therefore it has a compile and runtime dependency to Antlr (I also sort of dislike them bundling
the generator together with the runtime stuff) but nevertheless ... I can't really use JBurg
with Antlr4 in my project not without splitting the module up but I think that's rather ugly.

I too was thinking about making JBurg obsolete too ... the way I understand Antlr4 it looked
as if it provided a lot which could make life a lot easier. For example currently the ASParser
creates the AST objects and JBurg then performs it's magic on that AST tree. In Antlr4 I could
create a Parser-Listener that produces the AST object (Probably needed for IDE suppoer) but
could probably create a second listener that directly outputs flash bytecode. I think this
approach should be maximum performant, maximum simple and use only a fragment of the memory
the current solutions need ... Sort of XML Dom processing (old AST parsing) and SAX (Antlr4
direct Bytecode output).

But I also think that this would be quite an effort. Perhaps I should split up my falcon-antlr4
branch even more and sort of start with the CSS parser and as soon as that's up and running
get the output running ... then we could compare the performance of both and see if it's worth
going down that path. Don't want to waste time I could be working on Flexmojos, or the new
Flex-Maven-Plugin for a less performant solution. But if it was faster (I would expect it
to be) it could be quite breakthough ... after all I have never seen several datastructure
conversions be faster than direct output.

Chris

-----Urspr√ľngliche Nachricht-----
Von: Alex Harui [mailto:aharui@adobe.com] 
Gesendet: Dienstag, 22. Juli 2014 17:23
An: dev@flex.apache.org
Betreff: Re: AW: AW: Falcon and Antlr4

Interesting.  I did not know that Jburg used Antlr.

Another option is to remove Jburg as well and see if a simple AST walk like FalconJX uses
is faster or not.  I know for sure it would be easier to understand.

Also, IIRC, there compile-time dependencies and run-time dependencies.
Jburg is not used at run-time, IIRC. Would it totally defeat your goals to use Antlr2 for
Jburg during the build?

-Alex

On 7/22/14 5:01 AM, "Christofer Dutz" <christofer.dutz@c-ware.de> wrote:

>Ok ... so it seems that in order to procede here, I have to get in 
>touch with the JBurg guys.
>
>Currently JBurg is linked against Antlr2 and 3 but is incompatable with 
>Antlr4. So in order to safely use JBurg as code generator, I would have 
>to update JBurg to support Antlr4 :-(
>
>The project doesn't look that complex, but I think I should concentrate 
>on Flexmojos falcon support first :-(
>
>Chris
>
>________________________________________
>Von: Christofer Dutz <christofer.dutz@c-ware.de>
>Gesendet: Dienstag, 22. Juli 2014 10:06
>An: dev@flex.apache.org
>Betreff: AW: AW: Falcon and Antlr4
>
>Ok ... so I created a "falcon-antlr4" branch (Haven't comitted/pushed 
>that yet so don't look for it yet ;-) ).
>
>I am now trying to setup the compiler modules structure as a native 
>maven project and will try to get everything up and running with Antlr4
>
>One thing I did notice is that the parsers seem to consist of some 
>Antlr grammar files as well as JFlex .lex files. Is it possible that 
>both are actually doing the same thing?
>You said in a previous post that JFlex was used to make the Eclipse 
>suppot more responsive, but if it is only used for that I would like to 
>concentrate an an Anltr4 based parser and throwing out the JFlex stuff 
>making Eclipse use the Antlr4 parser instead after finishing the parser 
>itself ... would that be a valid approach?
>
>In recent projects I have had a lot of trouble and strange effects when 
>not using the standard maven directory layout because some third party 
>tools make assumptions that might not apply in non-maven-structures. On 
>the other hand making an Ant buildscript run with a maven directory 
>strucure is as easy as adjusting a few paths. So I would go that path 
>instead. Would that be ok for you?
>
>Currently this would result in the sources being placed in:
>src/main/java (Java Sources)
>src/main/antlr4 (Antlr Grammar Files)
>src/main/resources (Static Resources such as i18n property files, and 
>other static stuff)
>
>After reading the source, the ant build script and a lot of googling at 
>least I think I now know how to approach this task and I think I's 
>absolutely manageable ... really looking forward to this.
>
>Chris
>
>
>________________________________________
>Von: Gordon Smith <gsmithsf@hotmail.com>
>Gesendet: Donnerstag, 17. Juli 2014 17:11
>An: dev@flex.apache.org
>Betreff: Re: AW: Falcon and Antlr4
>
>> why are we mixing up several tools that all seem to be doing similar 
>>thing
>
>Falcon's lexer and parser were borrowed from the Flash Builder code 
>base, where they supported various edit-time code intelligence 
>features, but not compilation. I think the Flash Builder team had 
>determined that JFlex could tokenize faster than Antlr 2 could. But I 
>don't know whether that is still the case with Antlr 3 and 4.
>
>- Gordon
>
>Sent from my iPad
>
>> On Jul 16, 2014, at 3:14 PM, "Christofer Dutz"
>><christofer.dutz@c-ware.de> wrote:
>>
>> Ok so that would have been my second question ... why are we mixing 
>>up several tools that all seem to be doing similar things :-)
>>
>> I definitely like to do some cleaning up. But depending on if any or 
>>which talks are accepter for the ApacheCon I might have to finish some 
>>other things first ;-)
>>
>> Chris
>>
>> PS: Will be offline for a few days ...
>>
>> -----Urspr√ľngliche Nachricht-----
>> Von: Gordon Smith [mailto:gsmithsf@hotmail.com]
>> Gesendet: Mittwoch, 16. Juli 2014 17:19
>> An: dev@flex.apache.org
>> Betreff: Re: Falcon and Antlr4
>>
>> You might also want to look into eliminating JFlex and have Antlr 
>>handle tokenization as well.
>>
>> - Gordon
>>
>> Sent from my iPad
>>
>>> On Jul 16, 2014, at 8:06 AM, "Alex Harui" <aharui@adobe.com> wrote:
>>>
>>> Good luck.  I'm interested in anything that would speed up Falcon.
>>> Please work in a branch.
>>>
>>>> On 7/16/14 2:44 AM, "Christofer Dutz" <christofer.dutz@c-ware.de>
>>>>wrote:
>>>>
>>>> Hi,
>>>>
>>>>
>>>>
>>>> while I was havin a first look at the internals of Falcon, I was 
>>>> surprized to find a mixture of Antlr2 & Antlr3 grammars for 
>>>> creating the parsers.
>>>>
>>>> In a first moment I thought it would be a good idea to migrate the
>>>> Antlr2 grammars ASParser.g and MetadataParser.g to Antlr3 but after 
>>>> finding out that IntelliJ now has a neat Antlr4 plugin and reading 
>>>> a bit about the differences from 2 and 3 to 4 it sounded like a 
>>>> good idea to migrate all to Antlr4. To me it looks as if the way 
>>>> things are processed in Antlr4 would make the grammars a lot easier 
>>>> as well as implementing the rule logic. My gut-feeling tells me 
>>>> that an
>>>> Antlr4 parser should need less processing and be quite a bit faster.
>>>> I did experiment a little on the CSS grammar and successfully 
>>>> created an Antlr4 version of that ... so I guess it should be 
>>>> possible and it would clean up things quite dramatically.
>>>>
>>>>
>>>>
>>>> What I particularly liked, was that Antlr4 automatically generates 
>>>>a  Listener interface for any rule it finds generating an 
>>>>"enter{ruleneme}"
>>>> and "exit{rulename}" as well as a base-class implementing this 
>>>>interface.
>>>> Now all of the java code we had to enter in the rule-document can 
>>>>now  be defined in a FalconCssListener class that extends this 
>>>>CSSBaseListener.
>>>> This is where the Java code can be added to handle the rules and we  
>>>>can easily debug it (I know you could set breakpoints in the  
>>>>generated code, but I allways disliked that).
>>>>
>>>>
>>>>
>>>> What do you think? ... Would it be a good idea to give something 
>>>> like that a try? After all ... it's just 3 grammars (CSS, ASParser 
>>>> and MetadataParser). But I have to admit that the ASParser grammar 
>>>> looks way more complex than the CSS and the MetadataParser grammar.
>>>>
>>>>
>>>>
>>>> Chris
>>>


Mime
View raw message