SWIG

Thanks to your inspiration, I've already solve the problem.The solution is
just to complie another source file that provide all explicit instantiation
declaration for type Sequence::Fasta

On Thu, Jul 23, 2009 at 6:25 PM, William S Fulton
<wsf@...:

> Xin Shuai wrote:
>
>> In libsequence, there're a few .tcc files which defines templates. I think
>> I can just handle them as usual templates. That is , First parse the
>> definition , then instantiate them using defined type.
>>
>> Today I tried to use Alignment.tcc. So I revised my interface file:
>> ...
>> %include *<*../../Alignment.tcc*>*
>> ...
>> %template() Sequence::Alignment::IsAlignment*<*Sequence::Fasta*>*;
>> ....
>>
>>
>> Everything looks fine until I successfully created the wrapping code and
>> dynamic library. But when I want to import the dynamic library in python,
>> error appeared:
>> ...
>> ImportError: dlopen(../libsequence/biolib/_libsequence.so, 2): Symbol not
>> found:
>> __ZN8Sequence9Alignment7GetDataINS_5FastaEEERSiRSt6vectorIT_SaIS5_EES3_
>> Referenced from:
>> /Users/diavy/Documents/BioLib/biolib/src/mappings/swig/python/libsequence/biolib/_libsequence.so
>> Expected in: flat namespace
>>
>> It seems like my instantiated template code haven't been complied into my
>> dynamic library.
>>
>> Can anyone give me some advice?
>>
> Assuming the templates include the code bodies in the .tcc files, you need
> to add in a #include of the .tcc files into your wrappers. Take a look at
> any of the basic examples shipped with SWIG for details, where they normally
> #include the header files. Otherwise link to the compiled version of the
> template code.
>
> William
>
>


-- 
Xin Shuai (David)
PhD of Complex System in School of Informatics
Indiana University Bloomington
812-606-8019

Recently I've tried to map many function templates with specialization to
python. However it didn't work. For example:template<>bool
IsAlignment<http://molpopgen.org/software/libsequence/doc/html/namespaceSequence_1_1Alignment.html#359a699ee42255b8f0cf14c6ae633dce>
(const
std::vector< std::string > &data)

  Only when I remove the
'template<>' identifier and treated just like a general function. It
can be wrapped. But I consult the
swig documentation and it says template specilization is supported.

 Has anyone met similar situation?
-- 
Xin Shuai (David)
PhD of Complex System in School of Informatics
Indiana University Bloomington
812-606-8019

I'm using Python and it did provide library for std::list.I added a default
constructor to original code and it worked.

On Thu, Jul 23, 2009 at 6:19 PM, William S Fulton
<wsf@...:

> Xin Shuai wrote:
>
>> Hi:
>> I want to instantiate a list template like the following:
>>
>> %include <std_list.i>
>> %template(margList) std::list<Sequence::marginal>;#marginal is a class
>> defined in Sequence namespace
>>
>> when I complie, the error appeared :
>> */usr/include/c++/4.0.0/bits/stl_list.h:482: error: no matching function
>> for call to ‘Sequence::marginal::marginal()’*
>>
>> I searched on line and found a solution:
>> %ignore std::list<maginal>::list(size_type);
>> %include <std_list.i>
>> %template(margList) std::list<Sequence::marginal>;
>>
>> But the same error reemerged!
>>
>> PS: For the other classes defined in Sequence namespace, no error!
>>
>> Does anyone can give me some suggestion? Thanks.
>>
>>  Looks like marginal does not contain a default constructor and the
> solution you are trying should remove the wrapper methods that require the
> default constructor. Which language are you targeting? Maybe the target
> language hasn't defined the list constructor as you your code suggests it
> is.
>
>
> William
>



-- 
Xin Shuai (David)
PhD of Complex System in School of Informatics
Indiana University Bloomington
812-606-8019

Olly Betts wrote:
> On 2009-07-31, William S Fulton <wsf@...> wrote:
>> Apparently we can upgrade svn on the server to 1.5. Does anybody have 
>> knowledge of server side svn and the time to look at this? I can arrange 
>> for the SourceForge permissions to do the job. It would be good to do 
>> this as then we get the improved merging facilities of 1.5.
> 
> I'm no expert, but I updated Xapian's (non-sourceforge) SVN repo to 1.5 a few
> months back, so I can probably do it if nobody else volunteers.
> 
That would be great, thanks. I think you might have the permissions to 
go ahead. If not, let me know.

William

Hi,

This week:

Random fixes, pointed out by running foo.php in the testsuite when there
is no foo_runme.php (runtime test):

- fix for the director_using testcase
- Avoid calling is_resource() twice with no good reason
- fix for the allowexcept testcase
- Cleanups.
- fix for the cpp_namespace testcase

Next week:

I still have 6 failing testcase pointed out by the above method, I plan
to continue shrinking that list.