[mapserver-users] testing mapserver large files rendering improvements

[Paul Ramsey]

I've been thinking about auto-building qix files... it would fix this
problem and also fix the problem of the shape file that gets updated
and out-of-synch with its index file.

I think programmers will hate the idea, but users, maybe they will
like it? What do you think?

On Thu, May 15, 2008 at 7:09 AM, Guillaume Sueur
<guillaume.sueur at neogeo-online.net> wrote:
> Thanks for theses informations Jeff.
> I'd like to suggest to emphasis the need of a .qix file a little bit more in
> the mapserver documentation. It's only explained in 'Utility Programs'
> Section and 'Optimizing vector data sources' doc, but the huge benefit of it
> (25 x faster in this case, but still 5x on small shapes) is not, IMHO,
> mentionned enough to make people use it on every shapefile.
>
> Regards,
>
> Guillaume
>
>
> Jeff McKenna a écrit :
>>
>> Hello, comments inline below:
>>
>>
>> On 15-May-08, at 9:27 AM, Guillaume Sueur wrote:
>>
>>> Hi,
>>>
>>> I've just been doing few tests on mapserver trunk version to get an idea
>>> of how much the rendering of large shapefiles has been improved by paul's
>>> hacks on shx handler (http://blog.cleverelephant.ca/2008/04/see_30.html)
>>>
>>> My shapefile is 1 208 668 features, 160 Mo for the .shp file itself with
>>> a .shx of 5.2 Mo
>>>
>>> here are the times for a 5x5 km map to be drawn with shp2img, in PNG (no
>>> AGG)
>>>
>>> NO QIX :
>>> mapserv 5.0.2 : 0.680 s
>>> mapserv trunk : 0.699 s (!)
>>>
>>> WITH QIX :
>>> mapserv 5.0.2 : 0.124 s
>>> mapserv trunk : 0.027 s
>>>
>>>
>>> it looks like if the improvement only appears with use of a .qix file,
>>> which is highly recommended but sometimes forgotten.
>>> Has anyone had the same kind of experiment ?
>>
>> yes: http://lists.osgeo.org/pipermail/mapserver-dev/2008-April/007170.html
>>
>>>
>>>
>>>
>>> By the way, I noticed that it doesn't seem mandatory anymore to write the
>>> mapfile's DATA statement without the .shp extension to make use of the .qix
>>> file, as the results are strictly identical on mapserver trunk if I write
>>> "streets.shp" or "streets". Can someone confirm that ?
>>>
>>
>> Confirmed.  This was fixed recently (ticket:
>> http://trac.osgeo.org/mapserver/ticket/590), verified, and documentation was
>> updated.  Therefore the next release of MapServer will no longer require you
>> to omit the '.shp' for the DATA parameter for .qix use.
>>
>> -jeff
>>
>>
>> _______________________________________________
>> mapserver-users mailing list
>> mapserver-users at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/mapserver-users
>>
>
> _______________________________________________
> mapserver-users mailing list
> mapserver-users at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-users
>