Hi Jeff,
On 3/15/21 2:35 PM, Jeff McKenna wrote:
> Thanks Stefan for this wonderful introduction, it's nice to hear of
> positive experiences to give us all more motivation to continue to
> helping to improve MapServer! Welcome to both you and Iliya to the
> community. Keep us posted on your activities!
>
> Oh, once your MapServer backend is plugged-in be sure to add your
> weather site to the MapServer gallery, so we can help new users see
> what can be done...
Thanks for your nice words! I now have mapserver and mapcache running
successfully and talking to each other in my home lab. I have set up a
number of test layers from different input sources pretty successfully.
The documentation and the configuration required some adaption of my
thinking processes because it's pretty different from anything else I
have ever done before but I love this kind of challenges!
Currently I have only two issues which prevent me from doing the swap
right now. Thanks to NOAA using a 0-360° grid instead of (-180)-180°,
there's some artefacts when reprojecting to epsg:3857. For raster maps
this is already addressed in #6106. I ran into an issue which LINE maps
with a CONTOUR source which probably is related and opened #6261.
Then I have a question... Although it's not strictly necessary, I would
like to paint those Hs and Ls in my atmospheric pressure map. Is there a
way to run scripts or something like that in mapserver. I could probably
run some GDAL stuff externally over the grid files in my download script
and create some point overlays which could be in a GROUP with the
isoline layer but doing this inside of mapserver would kinda be cooler.
And a second question: How is mapserver handling raster files
memory-wise? Does it try to load the whole file to the memory at once?
My biggest issue during the whole 14 years I run this website always has
been balancing preprocessing and map rendering. The more preprocessing I
have to do, the longer the whole process of map generation takes but of
course I couldn't afford to buy a $100k top-of-the-line server with
terabytes of RAM and plenty of CPU cores for the rendering...
I'm thinking about my input file strategy at the moment...
1.) One GRIB file for everything (around 1.9GB)
2.) One GRIB file per parameter for all time steps
3.) One GRIB file per time step including all parameters
4.) One GRIB file per parameter for each time step
4. would be the most complex to configure and also would be fun (irony!)
to make local forecasts from but probably take the least memory.
2. and 3. are a fairly good compromise with 2 being easier for
calculating local forecasts and 3 easier for calculating additional
stuff like my "severe weather warning level"
1. would be easiest for rendering and calculating but even at the moment
with GrADS I every now and then run into memory issues (leaks?).
This project really turns out to be a fun challenge!
Stefan
--
(o_ Stefan Gofferje | SCLT, MCP, CCSA
//\ Reg'd Linux User #247167 | VCP #2263
V_/_ https://www.gofferje.net | https://www.saakeskus.fi
_______________________________________________
mapserver-users mailing list
mapserver-users@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/mapserver-users
