Re: [boost] py_cpp and vector<..>::const_iterator

From: David Abrahams (abrahams_at_[hidden])
Date: 2000-11-30 09:15:48

Ullrich,

Compilers w/o partial specialization have trouble with iterator_traits, but
I think the iterator_category() function always works, except for
user-defined iterators. In STLport, iterator_category() also works for
user-defined iterators if they are derived from std::iterator<>.

Please feel free to integrate this (on a branch).
Also, if you want to check in your build/configure system, that would be
lovely. I think it requires an update to building.html, though.

-Dave

----- Original Message -----
From: "Ullrich Koethe" <koethe_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Thursday, November 30, 2000 8:42 AM
Subject: Re: [boost] py_cpp and vector<..>::const_iterator

David,

good idea, no problem. Do you want me to integrate this into BPL (as a
branch at first)? If yes, I need to know how to find out the
iterator_category in a non-conforming compiler such as MSVC.

Regards
Ulli

David Abrahams wrote:
>
> Ullrich,
>
> it would be great if you could find a way to keep a reference to the
> container alive as long as the cursor was in existence. Otherwise, the
user
> risks crashing from Python because of a dangling reference.
>
> -Dave
> ----- Original Message -----
> From: "Ullrich Koethe" <koethe_at_[hidden]>
> To: <boost_at_[hidden]>
> Sent: Wednesday, November 29, 2000 4:09 PM
> Subject: Re: [boost] py_cpp and vector<..>::const_iterator
>
> > Anton Gluck wrote:
> > >
> > > Ullrich,
> > >
> > > On Mon, 27 Nov 2000, Ullrich Koethe wrote:
> > >
> > > > Petr Kocmid wrote:
> > > > >
> > > > > > Don't expose iterators. Python has its own iteration idioms
(e.g.
> > > > > > "for x in
> > > > > > S"), so just support the python sequence interface on your
vector.
> Since
> > > > > > your vector is a random-access container, this will be fast. I
> reccommend
> > > > > > this approach.
> > > > >
> > > > > I predict the problem Anton had will arise many times in the
future.
> Perhaps
> > > > > we can
> > > > > add some generic conversion of STL containers to sequence/mapping
> protocol
> > > > > to py_cpp.
> > > > >
> > > >
> > > >
> > > > I've implemented and successfully tested the following
> "Cursor"-feature
> > > > to wrap STL iterators:
> > > >
> > > > in C++:
> > > >
> > > >     typedef std::list<Foo> FooList;
> > > >     boost::python::class_builder<FooList> foolist_class(my_module,
> > > > "FooList");
> > > >     foolist_class.def(boost::python::constructor<>());
> > > >     foolist_class.def((void (FooList::*)(Foo const
> > > > &))&FooList::push_back, "append");
> > > >
> > > >     // define a cursor for FooLists
> > > >     my_module.def_cursor_for(foolist_class);
> > > >
> > > > Then in Python:
> > > >
> > > >     >>> fooList = FooList()
> > > >     >>> fooList.append(Foo(1))
> > > >     >>> fooList.append(Foo(2))
> > > >     >>>
> > > >     >>> for i in fooList.cusor(): # use cursor in a loop
> > > >     ...     i
> > > >     ...     print i.getData()
> > > >     ...
> > > >     <Foo object at 111070>
> > > >     1
> > > >     <Foo object at 1110e8>
> > > >     2
> > > >     >>>
> > > >     >>> c = fooList.cursor()
> > > >     >>> c[1].getData()   # use cursor to read and write the data
> > > >     2
> > > >     >>> c[1] = Foo(4)
> > > >     >>> c[1].getData()
> > > >     4
> > > >
> > > > Would this suit your needs ?
> > >
> > > I'm not sure if this was meant for me or Petr. I can see useful parts
> > > here when creating an iterator, but can't see how it helps exposing an
> > > existing iterator. I guess the magic is happening in
> > > "my_module.def_cursor_for(foolist_class);"?
> > >
> >
> > Sorry for being too vague. It was meant for you and Petr (and anyone
> > interested in wrapping iterators).
> >
> > Yes, the magic is happening in
> > my_module.def_cursor_for(foolist_class);". This line creates an
> > auxiliary class "FooList_cursor" that wraps FooList's iterator. In
> > addition, it adds the factory function "cursor()" to extension class
> > "FooList" itself.
> >
> > The auxiliary cursor class can be created for *any STL conforming
> > container* (code attached below). In particular, the following is
> > needed:
> >
> > FooList::iterator
> > FooList::value_type
> > FooList::begin()
> > FooList::size()
> > std::iterator_traits<FooList::iterator>::iterator_category
> >
> > The cursor implements random access functions (get_item() and
> > set_item()) for any iterator, but these will only be as efficient as the
> > underlying iterator allows. However, this is not a problem because 'for
> > i in fooList.cusor():' accesses the items in forward order anyway. Only
> > problem: I don't know how MSVC distinguishes iterator categories, so on
> > this compiler everything is treated as a forward iterator. Any hints?
> >
> > Regards
> > Ulli

--
 ________________________________________________________________
|                                                                |
| Ullrich Koethe  Universität Hamburg / University of Hamburg    |
|                 FB Informatik / Dept. of Computer Science      |
|                 AB Kognitive Systeme / Cognitive Systems Group |
|                                                                |
| Phone: [phone]                Vogt-Koelln-Str. 30 |
| Fax:   [phone]                D - 22527 Hamburg   |
| Email: u.koethe_at_[hidden]               Germany             |
|        koethe_at_[hidden]                        |
| WWW:   http://kogs-www.informatik.uni-hamburg.de/~koethe/      |
|________________________________________________________________|