search

PBrockmann / wms-pyferret

Slippy maps with WMS from pyferret
MIT License
3 stars 3 forks source link

tiles made from a fill command have holes #1

Open PBrockmann opened 8 years ago

PBrockmann commented 8 years ago

./slippy_map.bash 'fill/line/lev=(-inf)(30,40,1)(inf)/pal=mpl_PSU_inferno salt[k=1]'

holes

karlmsmith commented 8 years ago

If you use translucent colours you see that Ferret code sometimes overdraws polygons. So it might also be missing something as well.

Probably not related, but made me think of the following.... The "thin white lines" issue with PyFerret plots can be worked around using SET WINDOW /OUTLINE=0.75 which outlines polygons with that thickness of a line. The default of zero does not outline polygons (so faster).

The "thin white lines" issue is not actually a Ferret/PyFerret bug - lots of programs have this problem and everyone's work-around is to outline the polygons so they actually overlap just a bit. But one can see this jaggedness if one looks in closely.

PBrockmann commented 7 years ago

The white holes are related to "white space at the edge of colored contour plot" issue described in http://www.pmel.noaa.gov/maillists/tmap/ferret_users/fu_2005/msg00153.html

The request to the gunicorn server to produce the tiles can be seen as producing a small tile like this:

yes? use levitus_climatology
yes? shade/lev=(-inf)(20,30,.5)(inf)/x=-180:180/y=-90:90/nolab/nokey/hlim=-16.875:-11.25/vlim=-16.875:-11.25 temp[k=@max]
yes? fill/lev=(-inf)(20,30,.5)(inf)/x=-180:180/y=-90:90/nolab/nokey/hlim=-16.875:-11.25/vlim=-16.875:-11.25 temp[k=@max]

You get white spaces at the corners when using fill and not when using shade command. fill shade

Is there a better fix than (shade then fill/over) to solve this problem ?

karlmsmith commented 7 years ago

As you mention, well known artifact of fill; Ansley knows it well. The shade then fill/over is what we do in LAS; no better solution at this time.

PBrockmann commented 7 years ago

Even with this workaround (shade then fill/over), I get artefacts. shadethenfill

So I keep it as an open issue.