[OSM-dev] Many Layers Visible at Low Zoom-Levels for Rendering of Planetfile

Matt Kenny matthewkenny at gmail.com
Mon Apr 4 04:37:18 BST 2011


Hello,

If this isn't the proper mailing list to post this question, I
apologize in advance.

I'm currently working with a group of individuals from the Cascadia
Users of Geospatial Open Source (CUGOS) OS-Geo Chapter on creating our
own OSM planet server. We currently have the planet file imported and
updating, and are performing basic rendering using mod_tile + renderd
with mapnik2.

I'm experiencing some strange behavior in regards to rendering tiles
at low zoom-levels (e.g. Zoom Level 4). Specifically, it appears that
a number of layers are being rendered at these small-scales which
should actually not be rendered based on the 'MaxScaleDenominator'
values in my mapnik2 stylesheet. As all these layers are being
rendered across such a large spatial extent, render times are
extremely long.

Utilizing Dane Springmeyer's nik2img.py [1] utility in combination
with the osm.xml stylesheet [2] (our version is upgraded for mapnik2
syntax), I've produced the following log file [3]. AFAIK, a number of
these layers should not be rendering (reported as visible) at
zoom-scale 4. For example, the reported MapScaleDenominator via the
nik2img log is, '22363290.3714'. With this in mind, I went ahead and
searched my osm stylesheet for the 'MaxScaleDenominator' for one of
the layers I assumed should not be rendering at zoom-scale 4,
'piers-area'. The following snippet shows that the MaxScaleDenominator
parameter has a value much lower then the the reported
MapScaleDenominator at zoom-scale 4.

<Style name="piers-area">
    <Rule>
      <MaxScaleDenominator>200000</MaxScaleDenominator>
      <PolygonSymbolizer fill="#f2efe9"/></Rule>
</Style>

Additionally, this same behavior is occurring using a different
stylesheet, the osm-bright style by AJ Ashton of Development Seed. The
nik2img logfile for the same location and zoom can be found here [4].

So... after all that, my questions are essentially:
*    Is it correct in assuming that all these layers should _not_ be
listed as visible when rendering an image via nik2img at such a low
zoom-level?
*    If so, what might be causing the disconnect between the layers
that nik2img is reporting as visible for this mapscale versus the
layers' own MaxScaleDenominator values in my stylesheets? Could it be
a projection issue? Maybe something I'm introducing during the use of
upgrade_map_xml.py to convert mapnik -> mapnik2 syntax?

Thanks for any advice. I'd be happy to provide any further
supplementary/clarifying info.

-Matt

[1]: http://code.google.com/p/mapnik-utils/wiki/Nik2Img
[2]: http://svn.openstreetmap.org/applications/rendering/mapnik
[3]: http://pastebin.com/QDQ701nZ
[4]: http://pastebin.com/6HMpMvWb
--
m a t t h e w k e n n y
http://www.mkgeomatics.com



More information about the dev mailing list