From abrahams at mediaone.net  Fri Nov  3 01:09:47 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Thu, 2 Nov 2000 19:09:47 -0500
Subject: [C++-SIG] py_cpp maintenance release
Message-ID: <096501c0452a$6550c8f0$0500a8c0@dragonsys.com>

People turned up a few problems with the last release that seemed relatively
serious to me, so I have posted another update to
http://people.ne.mediaone.net/abrahams/downloads/py_cpp.html

Release notes:

2000-11-02 17:42

    Added a fix for interactions between default virtual function
    implementations and declare_base(). You still need to write your
    declare_base() /after/ all member functions have been def()d for the two
    classes concerned.  Many, many thanks to Ullrich Koethe
    <koethe at informatik.uni-hamburg.de> for all his work on this.

    Added missing conversions:
        to_python(float)
        from_python(const char* const&)
        from_python(const double&)
        from_python(const float&)

    Added a Regression test for a reference-counting bug thanks to Mark
Evans
    (<mark.evans at clarisay.com>)

    const-ify ClassBase::getattr()

    Add repr() function to Class<T>

    Add to_python/from_python conversions for PyPtr<T>

    Standardize set_item/get_item interfaces (instead of proxies) for Dict
and List

    Add Reprable<> template to newtypes.h

    Fix a bug wherein the __module__ attribute would be lost for classes
that have a
    default virtual function implementation.

    Remove extra ';' in module.cpp thanks to "Ralf W. Grosse-Kunstleve"
    <rwgk at cci.lbl.gov>

    Fix a bug in the code of example1.html




From abrahams at mediaone.net  Fri Nov  3 01:56:24 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Thu, 2 Nov 2000 19:56:24 -0500
Subject: [C++-SIG] test - please ignore
Message-ID: <099b01c04530$e3b39dc0$0500a8c0@dragonsys.com>





From abrahams at mediaone.net  Mon Nov  6 06:38:11 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Mon, 6 Nov 2000 00:38:11 -0500
Subject: [C++-SIG] Yet another py_cpp update
Message-ID: <0f5701c047b3$c64f0780$0500a8c0@dragonsys.com>

Available at http://www.egroups.com/files/boost/py_cpp/py_cpp_20001106.zip
or http://people.ne.mediaone.net/abrahams/downloads/py_cpp.html

-Dave

Changes since last release:

2000-11-06 0:22
    Lots of documentation updates

    added 4-argument template constructor to py::Tuple

    added "add" member function to ClassWrapper<> to allow arbitrary Python
    objects to be added to an extension class.

    gen_all.py now generates support for n argument member functions and n+1
    argument member functions at the suggestion of "Ralf W.
Grosse-Kunstleve"
    <rwgk at cci.lbl.gov>

    Added regression tests and re-ordered declare_base calls to verify that
the
    phantom base class issue is resolved.

2000-11-04 17:35

    Integrated Ullrich Koethe's brilliant from_python_experiment for better
    error-reporting in many cases.

    extclass.h, gen_extclass.py:
        removed special-case MSVC code
        added much commentary
        removed unused py_copy_to_new_value_holder

    init_function.h, gen_init_function.py:
        added missing 'template' keyword on type-dependent template member
usage
        removed special-case MSVC code
        added much commentary

2000-11-04 0:36

    Removed the need for the phantom base class that screwed up inheritance
    hierarchies, introduced error-prone ordering dependencies, and
complexified
    logic in many places!

    extclass.h: Added some explanatory comments, removed wasteful m_self
member
    of HeldInstance

    extclass_demo.cpp: Added #pragmas which allow compilation in ansi strict
    mode under Metrowerks

    functions.h: Added virtual_function as part of phantom base class
removal;
    expanded commentary

    pyptr.h: Added some missing 'typename's and a GCC workaround fix

    subclass.cpp: Added missing string literal const_cast<>s.

2000-11-03 10:58

    Fix friend function instantiation bug caught by Metrowerks (thanks
    Metrowerks!)

    Add proof-of-concept for one technique of wrapping function that return
a
    pointer

    Worked around MSVC optimizer bug by writing to_python(double) and
    to_python(float) out-of-line

2000-11-02 23:25

    Add /Zm200 option to vc6_prj to deal with MSVC resource limitations

    Remove conflicting /Ot option from vc6_prj release build






From Xavier.Juvigny at onera.fr  Wed Nov  8 14:28:38 2000
From: Xavier.Juvigny at onera.fr (Xavier JUVIGNY)
Date: Wed, 08 Nov 2000 14:28:38 +0100
Subject: [C++-SIG] Problem with last version of py_cpp
Message-ID: <3A095505.B196CF7A@onera.fr>

Hi,

I try to use the last version of py_cpp on a Unix station with gcc
2.95.1

I had a problem with a method with three arguments.
The folowing example :

#include <py_cpp/class_wrapper.h>

class A
{
public:
  A()
    {}
  void f(int,int,int,float,float,float)
    {
    }
};

void initMesh()
{
  try
  {
    // Create the object representing this extension module
    py::Module PyA("A");
    // Create the Python type object for our extension class.
    py::ClassWrapper<A> A_class(PyA, "PyA");
    //
    A_class.def(&A::f, "f");
  }
  catch(...)
  {
    py::handle_exception();
  }
}

give me the unexcepted error :

/data/Wksp/juvigny/Decoupeur.d/py_cpp/functions.h: In function `class
py::Function * py::new_wrapped_function<void (A::*)(int, int, int,
float, float, float)>(void (A::*)(int, int, int, float, float, float))':

/data/Wksp/juvigny/Decoupeur.d/py_cpp/extclass.h:371:   instantiated
from `py::ExtensionClass<A,py::HeldInstance<A> >::def<void (A::*)(int,
int, int, float, float, float)>(void (A::*)(int, int, int, float, float,
float), const char *)'
/data/Wksp/juvigny/Decoupeur.d/py_cpp/class_wrapper.h:35:   instantiated
from `py::ClassWrapper<A,py::HeldInstance<A> >::def<void (A::*)(int,
int, int, float, float, float)>(void (A::*)(int, int, int, float, float,
float), const char *)'
Py_A.C:22:   instantiated from here
/data/Wksp/juvigny/Decoupeur.d/py_cpp/functions.h:127: no matching
function for call to `return_value (void (A::* &)(int, int, int, float,
float, float))'

Has anyone a idea ?

Thanks for help.

Juvigny Xavier




From abrahams at mediaone.net  Thu Nov  9 15:41:12 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Thu, 9 Nov 2000 09:41:12 -0500
Subject: [C++-SIG] Re: Problem with last version of py_cpp
References: <973757716.3196@egroups.com>
Message-ID: <059f01c04a5b$1bb7bec0$0500a8c0@dragonsys.com>

>   From: Xavier JUVIGNY <Xavier.Juvigny at onera.fr>
> Subject: Problem with last version of py_cpp
>
> Hi,
>
> I try to use the last version of py_cpp on a Unix station with gcc
> 2.95.1
> <snip>
> Has anyone a idea ?

First, as your function takes more than 5 arguments, you'll need to run:

> gen_all.py xxx

where xxx >= the most arguments you need to handle in the py_cpp directory.
This will generate code to support more arguments.

I think you may want to grab an updated copy of gen_extclass.py first,
though, from:
http://cvs.sourceforge.net/cgi-bin/cvsweb.cgi/boost/development/py_cpp/gen_e
xtclass.py?cvsroot=boost

as I believe we fixed a bug in the past few days.

Also, I suggest you get gcc 2.95.2, as 2.95.1 is not supported with py_cpp.

Regards,
Dave




From tom at malcolmson.com  Wed Nov 15 17:53:22 2000
From: tom at malcolmson.com (Tom)
Date: Wed, 15 Nov 2000 11:53:22 -0500
Subject: [C++-SIG] Problems using iostreams or string with Python?
Message-ID: <LKBBKLLJKKBNCJKHAAJJEEFOGCAA.Tom@Malcolmson.com>

I have seen it reported that the C++ iostreams library (and the C++ string
class, which uses iostreams) should never be mixed with the C streams
library.

Given that Python uses the C streams library, does this mean that I can't
use C++ iostreams with Python related code?

Thanks,

Tom.




From abrahams at mediaone.net  Thu Nov 16 14:28:37 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Thu, 16 Nov 2000 08:28:37 -0500
Subject: [c++-sig] Digest Number 54
References: <974362228.71020@egroups.com>
Message-ID: <019e01c04fd1$209db720$0500a8c0@dragonsys.com>

>    From: tom at malcolmson.com (Tom)

> I have seen it reported that the C++ iostreams library (and the C++ string
> class, which uses iostreams) should never be mixed with the C streams
> library.

I've seen it reported that if you eat pop rocks and drink coke, your stomach
will explode ;-)

> Given that Python uses the C streams library, does this mean that I can't
> use C++ iostreams with Python related code?

No. On the other hand, if you want to write to stdout/stderr coherently,
using std::cout/std::cerr to do it may have surprising results, since they
are buffered separately from the 'C' streams. There's a lot more to C++
iostreams than just cout/cerr, though. For example, you can make an
ostreambuf that writes to a 'C' FILE stream, or you can write to an
ostringstream and then write its contents into stdout yourself (less
convenient, but easier to implement).

that's-what-happened-to-mikey-you-know-the-life-cereal-kid-ly y'rs
dave




From mark.mcmahon at eur.autodesk.com  Thu Nov 16 14:31:50 2000
From: mark.mcmahon at eur.autodesk.com (mark.mcmahon at eur.autodesk.com)
Date: Thu, 16 Nov 2000 14:31:50 +0100
Subject: [c++-sig] Digest Number 54
Message-ID: <C2AA61E5CAA7D211832F0008C7A4138E04B0A9E6@eurmsgneu01.eur.autodesk.com>

Or you could use sync_with_stdio

----- copy from MSDN -----------
ios_base::sync_with_stdio

static bool sync_with_stdio(bool sync = true);

The static member function stores a stdio sync flag, which is initially true. When true, this flag ensures that operations on the same file are properly synchronized between the iostreams functions and those defined in the Standard C library. Otherwise, synchronization may or may not be guaranteed, but performance may be improved. The function stores sync in the stdio sync flag and returns its previous stored value. You can call it reliably only before performing any operations on the standard streams. 
--------------------------------

-----Original Message-----
From: David Abrahams [mailto:abrahams at mediaone.net]
Sent: 16 November 2000 14:29
To: c++-sig at python.org
Subject: Re: [c++-sig] Digest Number 54



>    From: tom at malcolmson.com (Tom)

> I have seen it reported that the C++ iostreams library (and the C++ string
> class, which uses iostreams) should never be mixed with the C streams
> library.

I've seen it reported that if you eat pop rocks and drink coke, your stomach
will explode ;-)

> Given that Python uses the C streams library, does this mean that I can't
> use C++ iostreams with Python related code?

No. On the other hand, if you want to write to stdout/stderr coherently,
using std::cout/std::cerr to do it may have surprising results, since they
are buffered separately from the 'C' streams. There's a lot more to C++
iostreams than just cout/cerr, though. For example, you can make an
ostreambuf that writes to a 'C' FILE stream, or you can write to an
ostringstream and then write its contents into stdout yourself (less
convenient, but easier to implement).

that's-what-happened-to-mikey-you-know-the-life-cereal-kid-ly y'rs
dave


_______________________________________________
C++-SIG maillist  -  C++-SIG at python.org
http://www.python.org/mailman/listinfo/c++-sig



From abrahams at mediaone.net  Thu Nov 16 15:15:17 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Thu, 16 Nov 2000 09:15:17 -0500
Subject: [c++-sig] Digest Number 54
References: <C2AA61E5CAA7D211832F0008C7A4138E04B0A9E6@eurmsgneu01.eur.autodesk.com>
Message-ID: <01b601c04fd7$a5a54a90$0500a8c0@dragonsys.com>

That's a nonstandard and thus nonportable feature of one particular standard
library implementation. The OP didn't say which platform he was using.

-Dave
----- Original Message -----
From: <mark.mcmahon at eur.autodesk.com>


> Or you could use sync_with_stdio
>
> ----- copy from MSDN -----------
> ios_base::sync_with_stdio
>
> static bool sync_with_stdio(bool sync = true);
>
> The static member function stores a stdio sync flag, which is initially
true. When true, this flag ensures that operations on the same file are
properly synchronized between the iostreams functions and those defined in
the Standard C library. Otherwise, synchronization may or may not be
guaranteed, but performance may be improved. The function stores sync in the
stdio sync flag and returns its previous stored value. You can call it
reliably only before performing any operations on the standard streams.
> --------------------------------





From mark.mcmahon at eur.autodesk.com  Thu Nov 16 15:20:31 2000
From: mark.mcmahon at eur.autodesk.com (mark.mcmahon at eur.autodesk.com)
Date: Thu, 16 Nov 2000 15:20:31 +0100
Subject: [c++-sig] Digest Number 54
Message-ID: <C2AA61E5CAA7D211832F0008C7A4138E04B0A9EC@eurmsgneu01.eur.autodesk.com>

It is actually part of the standard  (of which I do not have) but

http://www.claas-solutions.de/iostreams/ios_base.html

M

-----Original Message-----
From: David Abrahams [mailto:abrahams at mediaone.net]
Sent: 16 November 2000 15:15
To: mark.mcmahon at eur.autodesk.com; c++-sig at python.org
Subject: Re: [c++-sig] Digest Number 54


That's a nonstandard and thus nonportable feature of one particular standard
library implementation. The OP didn't say which platform he was using.

-Dave
----- Original Message -----
From: <mark.mcmahon at eur.autodesk.com>


> Or you could use sync_with_stdio
>
> ----- copy from MSDN -----------
> ios_base::sync_with_stdio
>
> static bool sync_with_stdio(bool sync = true);
>
> The static member function stores a stdio sync flag, which is initially
true. When true, this flag ensures that operations on the same file are
properly synchronized between the iostreams functions and those defined in
the Standard C library. Otherwise, synchronization may or may not be
guaranteed, but performance may be improved. The function stores sync in the
stdio sync flag and returns its previous stored value. You can call it
reliably only before performing any operations on the standard streams.
> --------------------------------



_______________________________________________
C++-SIG maillist  -  C++-SIG at python.org
http://www.python.org/mailman/listinfo/c++-sig



From alexander.belopolsky at gs.com  Thu Nov 16 16:43:15 2000
From: alexander.belopolsky at gs.com (Belopolsky, Alexander)
Date: Thu, 16 Nov 2000 10:43:15 -0500
Subject: [c++-sig] Digest Number 54
Message-ID: <CD5A38A73620D311BCF60090271ECB2B05A8A1B4@gsny28e.et.gs.com>

Yes this is in the standard:

ISO/IEC 148882:1998(E) Section 27.4.2.4 [lib.ios.members.static]

--alex

-----Original Message-----
From: mark.mcmahon at eur.autodesk.com
[mailto:mark.mcmahon at eur.autodesk.com]
Sent: Thursday, November 16, 2000 9:21 AM
To: c++-sig at python.org
Subject: RE: [c++-sig] Digest Number 54


It is actually part of the standard  (of which I do not have) but

http://www.claas-solutions.de/iostreams/ios_base.html

M


> Or you could use sync_with_stdio



From abrahams at mediaone.net  Fri Nov 17 13:41:35 2000
From: abrahams at mediaone.net (David Abrahams)
Date: Fri, 17 Nov 2000 07:41:35 -0500
Subject: [c++-sig] Digest Number 55
References: <974448663.75752@egroups.com>
Message-ID: <03a801c05093$b900dcf0$0500a8c0@dragonsys.com>

> From: "Belopolsky, Alexander" <alexander.belopolsky at gs.com>
> Subject: RE: Digest Number 54
> 
> Yes this is in the standard:
> 
> ISO/IEC 148882:1998(E) Section 27.4.2.4 [lib.ios.members.static]

So it is. And a big "whoops, sorry!" to all.
-Dave