[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

RE: inconsistency in Python SWIG return parameters

From: Marshall White <cscidork_at_yahoo.com>
Date: 2003-02-28 01:53:44 CET

(Speculation) My guess is that the SWIG interface functions must return SOME kind of Python
object. Setting "Py_None" right off the bat makes it the default in case of no other return
values. I would also guess that returning NULL would be bad... The code seems to do that on
error conditions.

I believe this whole problem would be relatively simple to work around IF a single-valued PyTuple
is okay. (Isn't that an oxymoron?) http://dictionary.reference.com/search?q=tuple

I am not brave, and I haven't tried it, but you could do something like the code below. The only
difference between it and "t_output_helper" is the middle "else if" clause.

It gives the effect of ignoring the "default" Py_None, and sticks ALL subsequent objects into the
tuple regardless of what they are. Just remember that if it is only called ONCE, it will give a
"single-valued PyTuple."

Please note that it would be extremely wise to see how Python would react before trying it. That
is go see if the Python source code would "eat your hard drive."

The "woman" is telling me that it is HER turn to use the computer now. Good night gentleman.

static PyObject* NEW_t_output_helper(PyObject* target, PyObject* o) {
    PyObject* o2;
    PyObject* o3;

    if (!target) {
        target = o;
    } else if (target == Py_None) {
        Py_DECREF(Py_None);
        target = PyTuple_New(1);
        PyTuple_SetItem(target, 0, o);
    } else {
        if (!PyTuple_Check(target)) {
            o2 = target;
            target = PyTuple_New(1);
            PyTuple_SetItem(target, 0, o2);
        }
        o3 = PyTuple_New(1);
        PyTuple_SetItem(o3, 0, o);

        o2 = target;
        target = PySequence_Concat(o2, o3);
        Py_DECREF(o2);
        Py_DECREF(o3);
    }
    return target;
}

--- Dale Hirt <dale@sbcltd.com> wrote:
> Well, now I just need to get up to speed on Python. So far, my big question
> is, why does resultobj get set to Py_None right away. That seems to me to
> be pretty silly. It should be set to NULL, otherwise, if we use a newer
> version, then it will return 3 args, instead of the 2 that we desire. Hmmm,
> let me think on that part, too.
>
>
> > -----Original Message-----
> > From: Marshall White [mailto:cscidork@yahoo.com]
> > Sent: Thursday, February 27, 2003 3:55 PM
> > To: Dale Hirt; kfogel@collab.net
> > Cc: dev@subversion.tigris.org
> > Subject: RE: inconsistency in Python SWIG return parameters
> >
> >
> >
> > --- Dale Hirt <dale@sbcltd.com> wrote:
> > > > -----Original Message-----
> > > > From: Marshall White [mailto:cscidork@yahoo.com]
> > > [snip]
> > > >
> > > > It looks to me like one possible culprit, this funky
> > > > "t_output_helper", is generated by SWIG.
> > > >
> > >
> > > You're reading my thoughts on this one. So far, if we're to change
> > > anything, this is where I would start. However, then it
> > becomes a matter of
> > > how do we understand that there is indeed more than one
> > return value, or
> > > that a None value is permissible. You were right in that
> > t_output_helper is
> > > lazy, but making it work for it's dinner is going to be hard without
> > > breaking a lot of code. Boy, talk about a nice complex issue.
> >
> > Well it looks like that "generated" was not the right word.
> > It is hard coded in a file called "fragments.i"
> >
> > Greg Stein just posted an email with a wonderful suggestion.
> > Perhaps a function similar to "t_output_helper" could be
> > written that "does the right thing."
> > In the "*.i" files this new function could be specified
> > instead of "t_output_helper".
> >
> >
> > >
> > > > From reading I see that Dale Hirt is looking into this.
> > > > I wish him the best of luck. Perhaps there are command line
> > > > options for SWIG that might change
> > > > the default behavior of this function. A very quick glance
> > > > didn't show anything obvious though.
> > > >
> > > > I don't want to *duplicate* his effort, but I will look into
> > > > this some myself. Maybe two sets of
> > > > eyes might be better than one.
> > > >
> > > Any and all help is always appreciated. Like I said
> > before, I'm a relative
> > > newcomer to Python, Python extensions, and SWIG, so having
> > you look also
> > > should help a lot.
> >
> > I am a newcomer as well. I have been looking at Python for
> > about a month now.
> >
> > >
> > > >
> > > > Marshall
> > >
> > > Thanks,
> > >
> > > Dale
> >

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Feb 28 01:54:48 2003

This is an archived mail posted to the Subversion Dev mailing list.