Closed joshua-laughner closed 1 year ago
Looking into this more thoroughly, I see that the .mod
and .vmr
files actually have different names: the .mod
files use "000W" and the .vmr
files use "000E". The .mod
files generate the file name with mod_utils.mod_file_name
and the .vmr
files do so with mod_utils.vmr_file_name
.
The issue is that we know the longitude when creating the .mod
file, so it can correctly compute east vs. west. The .vmr
file needs to use just the coordinate in the file name, which since it is 0, it assumes that the direction should be "east". The .mod
files correctly match gsetup
in that the coordinate should be west if the obs. longitude is at all < 0.
My fix for this was to change how ginput
determines the output path for .vmr
files; if it's given paths to .mod
files, it uses the .mod
file name directly, rather than recalculating it from the lat/lon. It falls back to that approach if given a data dictionary instead. This is implemented in commit 14b6fbe for v1.0.6 and 8aa57ff for the main branch.
Some jobs submitted with the following inputs failed (email redacted):
The error was:
Looking at the files produced for
uc
(the slightly west site):the
.vmr
files are named "000E" instead of the "000W" that the automation is looking for, so there is a mismatch in how rounding around 0 is handled.