Home » Modeling » TMF (Xtext) » OCL and XTEXT(Good documentation about Xtext)
| |
| Re: OCL and XTEXT [message #656017 is a reply to message #656014] |
Wed, 23 February 2011 18:52   |
 Sebastian Zarnekow
Messages: 3118
Registered: July 2009 |
Senior Member |
|
|
Hi Ed,
are these numbers with Xtext 2.0?
Regards,
Sebastian
Am 23.02.11 19:37, schrieb Ed Willink:
> Hi
>
> Thanks for the reference. A more direct access is:
>
> http://www.eclipse.org/modeling/mdt/ocl/docs/publications/MO DELS2010OCLWorkshop/oclxtext.pdf
>
>
> When I presented the paper, I partially retracted the Xtext/ANTLR is 10
> times slower than LPG conclusion. If you force ANTLR to emulate LPG,
> then yes there is a problem. However if you revisit the grammar to
> reduce syntactic precision at the cost of more semantic validation,
> the performance cost dropped to perhaps a factor of two. The grammar
> size also shrank by about 10%. The heavy cost arises when ANTLR must
> backtrack when LPG can shift-reduce.
>
> Regards
>
> Ed Willink
>
>
> On 08/12/2010 08:43, Pierre-Alain BOURDIL wrote:
>> Hi, I found this document very usefull to get a good understanding of
>> Xtext. It is complementary of the Xtext official documentation, and very
>> interesting. So i decided to share this link . This paper is written by
>> ed willink.
>> http://dev.eclipse.org/viewcvs/viewvc.cgi/www/modeling/mdt/o cl/docs/publications/MODELS2010OCLWorkshop/oclxtext.pdf?view =co&root=Eclipse_Website
>>
>>
>> Regards,
>> Pierre-Alain
>>
>
|
|
|
| Re: OCL and XTEXT [message #656029 is a reply to message #656017] |
Wed, 23 February 2011 19:07 |
Ed Willink
Messages: 7655
Registered: July 2009 |
Senior Member |
|
|
Hi Sebastian
No. In September 2010 there was no 2.0. The results were very narrowly
focussed on the parsing since beyond that it was too hard to draw fair
comparisons, and some aspects of Xtext, particularly save were dreadful;
seems to be fixed in 2.0; thanks.
The grammar size size seems only slightly better in 2.0. I would be
surprised if the adverse cost of backtracking is changed in 2.0. For
sensibly structured grammars, my current estimate of only a two-fold
cost for ANTLR rather than LPG surprises me, I would expect an optimized
parser to be much better.
Regards
Ed Willink
On 23/02/2011 18:52, Sebastian Zarnekow wrote:
> Hi Ed,
>
> are these numbers with Xtext 2.0?
>
> Regards,
> Sebastian
>
> Am 23.02.11 19:37, schrieb Ed Willink:
>> Hi
>>
>> Thanks for the reference. A more direct access is:
>>
>> http://www.eclipse.org/modeling/mdt/ocl/docs/publications/MO DELS2010OCLWorkshop/oclxtext.pdf
>>
>>
>>
>> When I presented the paper, I partially retracted the Xtext/ANTLR is 10
>> times slower than LPG conclusion. If you force ANTLR to emulate LPG,
>> then yes there is a problem. However if you revisit the grammar to
>> reduce syntactic precision at the cost of more semantic validation,
>> the performance cost dropped to perhaps a factor of two. The grammar
>> size also shrank by about 10%. The heavy cost arises when ANTLR must
>> backtrack when LPG can shift-reduce.
>>
>> Regards
>>
>> Ed Willink
>>
>>
>> On 08/12/2010 08:43, Pierre-Alain BOURDIL wrote:
>>> Hi, I found this document very usefull to get a good understanding of
>>> Xtext. It is complementary of the Xtext official documentation, and very
>>> interesting. So i decided to share this link . This paper is written by
>>> ed willink.
>>> http://dev.eclipse.org/viewcvs/viewvc.cgi/www/modeling/mdt/o cl/docs/publications/MODELS2010OCLWorkshop/oclxtext.pdf?view =co&root=Eclipse_Website
>>>
>>>
>>>
>>> Regards,
>>> Pierre-Alain
>>>
>>
>
>
|
|
|
| Re: OCL and XTEXT [message #656103 is a reply to message #656029] |
Thu, 24 February 2011 08:50 |
Jan Koehnlein
Messages: 760
Registered: July 2009
Location: Hamburg |
Senior Member |
|
|
Xtext 2.0 now supports a form of syntactic predicates, which could allow
you to get rid of global backtracking and thus further improve performance.
Am 23.02.11 20:07, schrieb Ed Willink:
> Hi Sebastian
>
> No. In September 2010 there was no 2.0. The results were very narrowly
> focussed on the parsing since beyond that it was too hard to draw fair
> comparisons, and some aspects of Xtext, particularly save were dreadful;
> seems to be fixed in 2.0; thanks.
>
> The grammar size size seems only slightly better in 2.0. I would be
> surprised if the adverse cost of backtracking is changed in 2.0. For
> sensibly structured grammars, my current estimate of only a two-fold
> cost for ANTLR rather than LPG surprises me, I would expect an optimized
> parser to be much better.
>
> Regards
>
> Ed Willink
>
>
>
> On 23/02/2011 18:52, Sebastian Zarnekow wrote:
>> Hi Ed,
>>
>> are these numbers with Xtext 2.0?
>>
>> Regards,
>> Sebastian
>>
>> Am 23.02.11 19:37, schrieb Ed Willink:
>>> Hi
>>>
>>> Thanks for the reference. A more direct access is:
>>>
>>> http://www.eclipse.org/modeling/mdt/ocl/docs/publications/MO DELS2010OCLWorkshop/oclxtext.pdf
>>>
>>>
>>>
>>>
>>> When I presented the paper, I partially retracted the Xtext/ANTLR is 10
>>> times slower than LPG conclusion. If you force ANTLR to emulate LPG,
>>> then yes there is a problem. However if you revisit the grammar to
>>> reduce syntactic precision at the cost of more semantic validation,
>>> the performance cost dropped to perhaps a factor of two. The grammar
>>> size also shrank by about 10%. The heavy cost arises when ANTLR must
>>> backtrack when LPG can shift-reduce.
>>>
>>> Regards
>>>
>>> Ed Willink
>>>
>>>
>>> On 08/12/2010 08:43, Pierre-Alain BOURDIL wrote:
>>>> Hi, I found this document very usefull to get a good understanding of
>>>> Xtext. It is complementary of the Xtext official documentation, and
>>>> very
>>>> interesting. So i decided to share this link . This paper is written by
>>>> ed willink.
>>>> http://dev.eclipse.org/viewcvs/viewvc.cgi/www/modeling/mdt/o cl/docs/publications/MODELS2010OCLWorkshop/oclxtext.pdf?view =co&root=Eclipse_Website
>>>>
>>>>
>>>>
>>>>
>>>> Regards,
>>>> Pierre-Alain
>>>>
>>>
>>
>>
>
--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com
---
Get professional support from the Xtext committers at www.typefox.io
|
|
|
Goto Forum:
Current Time: Tue Apr 23 12:51:25 GMT 2024
Powered by FUDForum. Page generated in 0.03444 seconds |
] 
Search
Help
Register
Login
Home
