darksidelemm / cusf_predictor_wrapper Goto Github PK

This repo is fairly standalone, and so can probably be packed up into something that can be installed using pip.

Will need to move the get_wind_data script into a module. Also need to check how the GDAL dependency breaks things when installing using pip.

Investigate switching to cfgrib: https://pypi.org/project/cfgrib/

This will be used to replace the GDAL calls in this function: https://github.com/darksidelemm/cusf_predictor_wrapper/blob/master/apps/get_wind_data.py#L246

Hopefully this will reduce the current difficulty in getting this wrapper going.

A few days ago the NOAA website used to get wind data changed their directory structure such that, for example /cgi-bin/filter_gfs_0p50.pl?dir=/gfs.######## no longer works. The change seems to only be the addition of a subdirectory for the data to be in called atmos. As such, changing the URL to /cgi-bin/filter_gfs_0p50.pl?dir=/gfs.########/atmos works.

I only use the get_wind_data.py script for my fork of this repo and I fixed this issue by modifying line 97 and adding /atmos to the value of _filter_params['dir'] as below:

_filter_params['dir'] = "/gfs.%s/atmos" % (_model_timestring)

I am unsure of how this would affect the main application as I do not implement it in the project I use.

Executing Command: sudo python3 setup.py install
Ends with the following:
Processing dependencies for cusfpredict==0.2.1
error: numpy 1.17.4 is installed but numpy>=1.18.5; platform_machine != "aarch64" and platform_machine != "arm64" and python_version < "3.10" is required by {'pandas'}

Running Ubuntu 20.04.4 LTS, Python3

Any thoughts on how to workaround?

Hi,

i made it compile without errors being thrown by adding libgeos-dev, as well as libatlas-base-dev described here:

https://numpy.org/devdocs/user/troubleshooting-importerror.html

It still keeps nagging about
"eccodes 1.3.3 has requirement numpy<1.20, but you'll have numpy 1.21.2 which is incompatible." but that doesn't stop it from running.

Given the GRIB filter appears to be going down every other week, we need to switch to something more reliable...