queryByShape on PostGIS-Layer segfault in php-mapscript

[Uwe Junge]

Thank You very much Bart.

Unfortunatly, I rely at this time completely on the binary package made
available
at maptools.org; the latest version there is
mapserver-4.5-win32-php4.3.7.zip
built on March-15.  So let's wait ...

The function to do queries-by-shape are essential to our application.
Yesterday, I tried PostGis 1.0.0 and was really pleased (see further down) .
Currently, all the data, including dynamically generated data, are
sourced as shapefiles
in a multi-server arrangement.  The overhead to distribute dynamically
generated data  to serveral servers is quit significant at this time.

Of course, additional functionality can be developed around the application
when using PostGis as an additional source, in partiular when it comes
- to large Attribut-Datasets (for example Census SF1 & SF3 with about
16K Attributes) that needs to be sourced for thematic maps,
- or when using serveral mapservers in parallel with dynamic content
stored on one single PostGis-Server.
Developing a separate, scripted direct search on PostGis on the back-end
is not an option,
since the overhead to append the results one-by-one to the
mapserver-query-result-set (*.qyr), on which the application relies, is
way to much.

Just for interest, let me summarize my experience with mapserver:

I made several attempts to build php_mapscript.dll from the sources with
ms vc, together with
postgres/postgis on cygwin, spend a lot of time and never got it working.
So i focus on application development instead an use the binary packages
for the installations.
Fortunatly, i found yesterday the brand new installers for postgres 80
and postgis;
- it took me 4 hours to install, to write a script and to import 1Gig
worth of Tiger-shape files and
geocoded database content.
I'm **very** pleased about the speed of mapserver in conjunction with
postgis,
there is only a relatively small rendering-time difference (varying
between 0..30%)
for the colocated, postgres server out-of-the-box compared to reading
the indexed shape files (tiles) directly !
The PostGis-Team did a great job with their new Version 1.0.0.

Also, i'm **very** pleased about the very small foot-print of
php_mapscript compared to other
GIS-servers. Map-File and Layer-Managment are anyway custom, so these
additional
capabilities of  'competing products'  are of no use (for us) anyway.

To my surprise, rendering time (ca 30 Gig of mapdata browsed) on my new
notebook (mobile Athlon 3700+,1Gig Ram with a 7.200 rpm HD on USB2.0)
was significantly faster compared to development- or the production server
(Dual Xeon 2.4-2.8GHz, 4Gig RAM, SCSI Raid10 15K rpm HD)  !!!!!!
Of course I know when more user simultaniously access the services, this
situation will change.
But the lesson learned for mapserver is to spend the money for the
processors and not for the disks.

Last year we also made an attempt to to use Linux (32bit) for the
mapservers, apart from getting
php_mapscript working, rendering was significantly slower compared to
W2K on the same hardware.
But that should change when using 64bit versions and hardware.




Uwe Junge
CTO PMCentral Systems, Inc.
4201 Pinefield Court
Fairfax, VA 22033