[geos-devel] Fwd: Prepared Geometry

[Paul Ramsey]

Another quarter heard from. I have applied this "ugly" patch
nonetheless, since it removes my leak in PostGIS.

P


---------- Forwarded message ----------
From: Hartmut Kaiser <hartmut.kaiser at gmail.com>
Date: Thu, Oct 9, 2008 at 6:55 AM
Subject: RE: Prepared Geometry
To: Paul Ramsey <pramsey at cleverelephant.ca>


Paul,

> Could you try big against small1 and small2?

Patch attached. But frankly, I don't like the solution (even if it adopted
the way it's currently handled all over the place in GEOS).

The way the SegmentString class is written and used requires the user of the
an instance of SegmentString to take care of the deletion of the embedded
point array, for instance:

   for ( size_t i = 0, ni = lineSegStr.size(); i < ni; i++ )
   {
       delete lineSegStr[ i ]->getCoordinates();    // <-- here
       delete lineSegStr[ i ];
   }

Which is just wrong, because this could be done in a better way if
SegmentString was taking care of this in the destructor. I tried to fix GEOS
(and the tests) in this direction, but apparently the buffers stored in the
SegmentString are shared in between different higher level structures in
addition, which makes it impossible (at least currently for me) to fix that
without introducing some kind of smart pointers (or investing a really large
amount of time, which I don't have) ...

For this reason I fixed this memory leak the straightforward (and ugly) way.
I'm sure there are more similar bugs to recover, though :-P

Regards Hartmut

PS: As impressive GEOS is it convinces me once more that it's not a good
idea to write C++ in a way similar to Java, especially if no special means
of memory management (i.e. smart ptrs) are introduced.


>
> Thanks!
>
> Paul
>
> On Wed, Oct 8, 2008 at 1:54 PM, Hartmut Kaiser
> <hartmut.kaiser at gmail.com> wrote:
> > Paul,
> >
> >> The reason I ask is because while contains and covers appear leak
> >> free, intersects seems to be leaking. I'm pretty sure it's not on
> the
> >> PostGIS side (I think).
> >>
> >> P.
> >>
> >> On Wed, Oct 8, 2008 at 12:25 PM, Paul Ramsey
> >> <pramsey at cleverelephant.ca> wrote:
> >> > Hartmut,
> >> >
> >> > When you tested for leaks, did you test prepared contains,
> >> intersects,
> >> > covers, coversproperly? Or just one of them?
> >
> > I was using the test program as attached to the ticket (which AFAIR
> > calls all of the topological attribute functions). This test program
> > was leak free for me. But I tried the geometries as attached to the
> > ticket only either, so your leaks might show themselves with certain
> geometries only.
> >
> > Regards Hartmut
> >
> >
> >
> >
-------------- next part --------------
A non-text attachment was scrubbed...
Name: memleak2.patch
Type: application/octet-stream
Size: 678 bytes
Desc: not available
Url : http://lists.osgeo.org/pipermail/geos-devel/attachments/20081009/92920944/memleak2.obj