pygccxml-development

[pygccxml-development] Removing copy constructor

[Benoit L. <ben...@mo...>]

Hi,

I have a problem with a class I try to bind (how original!).

It has a boost::mutex member (which is non-copyable).
My class itself has no explicitely defined copy-constructor.
==> py++ tries to access this copy-constructor in the wrapper class 
copy-constructor.

class MyClass
{
public:
    MyClass(){....}
    // ....
protected:
    // ....
    boost::mutex m_mutex;
    // ....
};

struct MyClassWrapper : MyClass, bp::wrapper<MyClass> {
     MyClassWrapper(MyClass const & arg ) : MyClass( arg ), bp::wrapper< 
MyClass >()
     {
           // copy constructor
     }

     // other stuff....
     // ...
};

I tried to set the class.copyable to False, but py++ still generates the 
same code...

Any clue?

Thanks!

Benoit

Re: [pygccxml-development] Removing copy constructor

[Roman Y. <rom...@gm...>]

On Tue, Apr 13, 2010 at 2:40 PM, Benoit Leveau
<ben...@mo...> wrote:
> Hi,
>
> I have a problem with a class I try to bind (how original!).

I hope py++ is part of the solution. If you have comments how it could
be improved - I'll be glad to get them.

Some time ago I wrote "easy extending guide"(
http://language-binding.net/pyplusplus/troubleshooting_guide/easy_extending_guide/easy_extending_guide.html
). Only pretty simple code could be wrapped without problems.

> It has a boost::mutex member (which is non-copyable).
> My class itself has no explicitely defined copy-constructor.
> ==> py++ tries to access this copy-constructor in the wrapper class
> copy-constructor.
>
> class MyClass
> {
> public:
>    MyClass(){....}
>    // ....
> protected:
>    // ....
>    boost::mutex m_mutex;
>    // ....
> };
>
> struct MyClassWrapper : MyClass, bp::wrapper<MyClass> {
>     MyClassWrapper(MyClass const & arg ) : MyClass( arg ), bp::wrapper<
> MyClass >()
>     {
>           // copy constructor
>     }
>
>     // other stuff....
>     // ...
> };
>
> I tried to set the class.copyable to False, but py++ still generates the
> same code...
>
> Any clue?


No. I have the following test: ( see noncopyable_to_be_exported.hpp file )


C++ code:



class c_t : public boost::noncopyable{

public:

    static char get_c(){ return 'c'; }



};

struct e_t{

    virtual void do_smth() = 0;

private:

    c_t c;

};


The generated code (without tweaking) is the following:

namespace bp = boost::python;

struct e_t_wrapper : noncopyables::e_t, bp::wrapper< noncopyables::e_t > {

    virtual void do_smth(  ){
        bp::override func_do_smth = this->get_override( "do_smth" );
        func_do_smth(  );
    }

};


Can you provide more details?


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/

Re: [pygccxml-development] Removing copy constructor

[Benoit L. <ben...@mo...>]

Roman Yakovenko wrote:
> On Tue, Apr 13, 2010 at 2:40 PM, Benoit Leveau
> <ben...@mo...> wrote:
>> Hi,
>>
>> I have a problem with a class I try to bind (how original!).
> 
> I hope py++ is part of the solution. If you have comments how it could
> be improved - I'll be glad to get them.
So far, everything is going well ;)

>> It has a boost::mutex member (which is non-copyable).
>> My class itself has no explicitely defined copy-constructor.
>> ==> py++ tries to access this copy-constructor in the wrapper class
>> copy-constructor.
>>
>> class MyClass
>> {
>> public:
>>    MyClass(){....}
>>    // ....
>> protected:
>>    // ....
>>    boost::mutex m_mutex;
>>    // ....
>> };
>>
>> struct MyClassWrapper : MyClass, bp::wrapper<MyClass> {
>>     MyClassWrapper(MyClass const & arg ) : MyClass( arg ), bp::wrapper<
>> MyClass >()
>>     {
>>           // copy constructor
>>     }
>>
>>     // other stuff....
>>     // ...
>> };
>>
>> I tried to set the class.copyable to False, but py++ still generates the
>> same code...
>>
>> Any clue?
> 
> 
> No. I have the following test: ( see noncopyable_to_be_exported.hpp file )
> 
> 
> C++ code:
> 
> 
> 
> class c_t : public boost::noncopyable{
> 
> public:
> 
>     static char get_c(){ return 'c'; }
> 
> 
> 
> };
> 
> struct e_t{
> 
>     virtual void do_smth() = 0;
> 
> private:
> 
>     c_t c;
> 
> };
> 
> 
> The generated code (without tweaking) is the following:
> 
> namespace bp = boost::python;
> 
> struct e_t_wrapper : noncopyables::e_t, bp::wrapper< noncopyables::e_t > {
> 
>     virtual void do_smth(  ){
>         bp::override func_do_smth = this->get_override( "do_smth" );
>         func_do_smth(  );
>     }
> 
> };
 > Can you provide more details?

That's really interesting...I didn't know py++ was supposed to detect 
this case (when a class has a non-copyable member). I'm going to try to 
reproduce your example, then i'll introduce my types...hopefully at some 
point I'll see what's causing the problem.
I'll let you know what I can find.

Re: [pygccxml-development] Removing copy constructor

[Roman Y. <rom...@gm...>]

On Tue, Apr 13, 2010 at 3:11 PM, Benoit Leveau
<ben...@mo...> wrote:
> That's really interesting...I didn't know py++ was supposed to detect this
> case (when a class has a non-copyable member).

:-)

Take a look on:

http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/data/noncopyable_to_be_exported.hpp?view=markup
http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/noncopyable_tester.py?view=markup

and the attached generated file

> I'm going to try to reproduce
> your example, then i'll introduce my types...hopefully at some point I'll
> see what's causing the problem.
> I'll let you know what I can find.

Cool.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/

Re: [pygccxml-development] Removing copy constructor

[Benoit L. <ben...@mo...>]

Roman Yakovenko wrote:
> On Tue, Apr 13, 2010 at 3:11 PM, Benoit Leveau
> <ben...@mo...> wrote:
>> That's really interesting...I didn't know py++ was supposed to detect this
>> case (when a class has a non-copyable member).
> 
> :-)
> 
> Take a look on:
> 
> http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/data/noncopyable_to_be_exported.hpp?view=markup
> http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/noncopyable_tester.py?view=markup
> 
> and the attached generated file
> 
>> I'm going to try to reproduce
>> your example, then i'll introduce my types...hopefully at some point I'll
>> see what's causing the problem.
>> I'll let you know what I can find.
> 
> Cool.

Thanks a lot! I'll try this in a few minutes..
I should have looked in the unit tests, I didn't think about it...

Re: [pygccxml-development] Removing copy constructor

[Benoit L. <ben...@mo...>]

Benoit Leveau wrote:
> Roman Yakovenko wrote:
>> On Tue, Apr 13, 2010 at 3:11 PM, Benoit Leveau
>> <ben...@mo...> wrote:
>>> That's really interesting...I didn't know py++ was supposed to detect this
>>> case (when a class has a non-copyable member).
>> :-)
>>
>> Take a look on:
>>
>> http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/data/noncopyable_to_be_exported.hpp?view=markup
>> http://pygccxml.svn.sourceforge.net/viewvc/pygccxml/pyplusplus_dev/unittests/noncopyable_tester.py?view=markup
>>
>> and the attached generated file
>>
>>> I'm going to try to reproduce
>>> your example, then i'll introduce my types...hopefully at some point I'll
>>> see what's causing the problem.
>>> I'll let you know what I can find.
>> Cool.
> 
> Thanks a lot! I'll try this in a few minutes..
> I should have looked in the unit tests, I didn't think about it...
using your example, I indeed don't get a copy-constructor, even after 
"complexifying" the class so it looks more like mine... this is probably 
something tricky! I'm not sure I really have to bind that class, but 
I'll let you know if I find where the problem was

I was thinking about something. I had a few troubles now and then using 
the call_policies. I explain:
If you want to use "copy_non_const_reference" you use:
function.call_policies = 
call_policies.return_value_policy(call_policies.copy_non_const_reference)

but if you want to use return_internal_reference you use:
function.call_policies = call_policies.return_internal_reference()

I had to look into call_policies.py and from time to time I forget and 
get some errors (even sometimes py++ doesn't complain but the errors 
appear at compilation).
All these policies are listed in the same section, without any mention 
that one is a function, the others are string and need to be passed to 
return_value_policy().
It might be useful to add that to the documentation, if it's not already 
somewhere I missed.

Benoit

Re: [pygccxml-development] Removing copy constructor

[Roman Y. <rom...@gm...>]

On Tue, Apr 13, 2010 at 5:58 PM, Benoit Leveau
<ben...@mo...> wrote:

> using your example, I indeed don't get a copy-constructor, even after
> "complexifying" the class so it looks more like mine... this is probably
> something tricky! I'm not sure I really have to bind that class, but I'll
> let you know if I find where the problem was

Thanks. "copyable" is the most "fragile" part of the project. That's
why I has so many tests for it.

> I was thinking about something. I had a few troubles now and then using the
> call_policies. I explain:
> If you want to use "copy_non_const_reference" you use:
> function.call_policies =
> call_policies.return_value_policy(call_policies.copy_non_const_reference)
>
> but if you want to use return_internal_reference you use:
> function.call_policies = call_policies.return_internal_reference()
>
> I had to look into call_policies.py and from time to time I forget and get
> some errors (even sometimes py++ doesn't complain but the errors appear at
> compilation).
> All these policies are listed in the same section, without any mention that
> one is a function, the others are string and need to be passed to
> return_value_policy().
> It might be useful to add that to the documentation, if it's not already
> somewhere I missed.

I tried to implement the interface similar to Boost.Python library:

http://www.boost.org/doc/libs/1_42_0/libs/python/doc/tutorial/doc/html/python/functions.html#python.call_policies

    * with_custodian_and_ward: Ties lifetimes of the arguments
    * with_custodian_and_ward_postcall: Ties lifetimes of the
arguments and results
    * return_internal_reference: Ties lifetime of one argument to that of result
    * return_value_policy<T> with T one of:
          o reference_existing_object: naive (dangerous) approach
          o copy_const_reference: Boost.Python v1 approach
          o copy_non_const_reference:
          o manage_new_object: Adopt a pointer and hold the instance

where you replace "< T >" with "( T )", I hoped it will be enough to
look at Boost.Python documentation and understand how the code (py++)
should be written.

Will few comments in the source file fix this or we need to introduce
some validation checks?

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/

Re: [pygccxml-development] Removing copy constructor

[Benoit L. <ben...@mo...>]

Roman Yakovenko wrote:
> On Tue, Apr 13, 2010 at 5:58 PM, Benoit Leveau
> <ben...@mo...> wrote:
> 
>> using your example, I indeed don't get a copy-constructor, even after
>> "complexifying" the class so it looks more like mine... this is probably
>> something tricky! I'm not sure I really have to bind that class, but I'll
>> let you know if I find where the problem was
> 
> Thanks. "copyable" is the most "fragile" part of the project. That's
> why I has so many tests for it.
>> I was thinking about something. I had a few troubles now and then using the
>> call_policies. I explain:
>> If you want to use "copy_non_const_reference" you use:
>> function.call_policies =
>> call_policies.return_value_policy(call_policies.copy_non_const_reference)
>>
>> but if you want to use return_internal_reference you use:
>> function.call_policies = call_policies.return_internal_reference()
>>
>> I had to look into call_policies.py and from time to time I forget and get
>> some errors (even sometimes py++ doesn't complain but the errors appear at
>> compilation).
>> All these policies are listed in the same section, without any mention that
>> one is a function, the others are string and need to be passed to
>> return_value_policy().
>> It might be useful to add that to the documentation, if it's not already
>> somewhere I missed.
> 
> I tried to implement the interface similar to Boost.Python library:
> 
> http://www.boost.org/doc/libs/1_42_0/libs/python/doc/tutorial/doc/html/python/functions.html#python.call_policies
> 
>     * with_custodian_and_ward: Ties lifetimes of the arguments
>     * with_custodian_and_ward_postcall: Ties lifetimes of the
> arguments and results
>     * return_internal_reference: Ties lifetime of one argument to that of result
>     * return_value_policy<T> with T one of:
>           o reference_existing_object: naive (dangerous) approach
>           o copy_const_reference: Boost.Python v1 approach
>           o copy_non_const_reference:
>           o manage_new_object: Adopt a pointer and hold the instance
> 
> where you replace "< T >" with "( T )", I hoped it will be enough to
> look at Boost.Python documentation and understand how the code (py++)
> should be written.
> 
> Will few comments in the source file fix this or we need to introduce
> some validation checks?

I don't think introducing some validation checks is necessary.
The confusing thing is that the online doc shows a different indentation 
than what you just said:

* default_call_policies
* return_value_policy
    - return_opaque_pointer
    - copy_const_reference
    - return_by_value
    - copy_non_const_reference
    - return_internal_reference <= should be at the same level as 
return_value_policy, else we think we must use return_value_policy( 
return_internal_reference )
* return self

so this would just be a matter of changing the indentation in
docs/documentation/functions/call_policies/call_policies.rest

Re: [pygccxml-development] Removing copy constructor

[Roman Y. <rom...@gm...>]

On Wed, Apr 14, 2010 at 11:32 AM, Benoit Leveau
<ben...@mo...> wrote:
> * default_call_policies
> * return_value_policy
>   - return_opaque_pointer
>   - copy_const_reference
>   - return_by_value
>   - copy_non_const_reference
>   - return_internal_reference <= should be at the same level as
> return_value_policy, else we think we must use return_value_policy(
> return_internal_reference )
> * return self
>
> so this would just be a matter of changing the indentation in
> docs/documentation/functions/call_policies/call_policies.rest

Done: http://pygccxml.svn.sourceforge.net/pygccxml/?rev=1836&view=rev

Thank you.



-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/