Another related idea is to not use srcML for the main parsing, but to
keep it around for custom queries.
So, we could still keep the current solution with inFusion for the
main models, and have srcML for when we want to write queries that
need access to AST.
For example, if you do not have the right libraries, inFusion will not
create annotation objects. If I would have access to the AST, I could
write a query (even if it is expensive at first) that checks the AST
of class for the existence of such annotations.
I would use this. And maybe like this the srcML interface would grow
in time and get better. What do you think?
On 28 Apr 2010, at 14:44, Alexandre Bergel wrote:
As Simon said, the type resolution in C is easy. In C++ it is
On 28 Apr 2010, at 04:47, Simon Denier wrote:
> On 28 avr. 2010, at 10:36, Stéphane Ducasse wrote:
>> the problem is that meta model is different so you would have to
>> map everything.
> Besides, from what I have seen, the generated xml is sometimes not
> regular: i.e. similar things in the code does not have the same xml
> representation. This makes it cumbersome as it is not documented,
> so you have to decode the grammar from the xml and adapt to such
> In the end, you still have to maintain a xml parser for each
> language and you still dont have control about what is extracted.
> And you have to do the type resolution :)
>> On Apr 28, 2010, at 10:28 AM, Laval Jannik wrote:
>>> Hi all,
>>> I am looking at srcML, and I see that it make XML files from Java
>>> and C++ source code.
>>> Since we use it for C source code, it should be possible to make
>>> it works with C++ and java.
>>> This strategy allows us to not us a special tool for each language.
>>> What do you think about it ?
>>> Jannik Laval
>>> Moose-dev mailing list
>> Moose-dev mailing list
> Moose-dev mailing list
Alexandre Bergel http://www.bergel.eu
Moose-dev mailing list