Issues with generated monitor files
Summary
I'm trying to generate a monitor file following the steps in the online doc (section 6.2.3 Generate a monitor file compatible with the CIF). There were some issues (#2, #4), but I managed to get an output file. There are some issues with that file however.
Steps to reproduce
Follow the steps in the online doc. Update yaml file as needed, inspect generated monitor file (see also the table in section 4.1.1 of the docs).
What is the current bug behavior?
- All datasets are 64 bit floats, including the ones that should be integers in my opinion: dtstep, i, index, j, level, tstep, and tstep_glo.
- There are datasets that should be character according to the table in the doc, but are also 64 bit floats: network, station
- The i and j datasets are filled with NaNs. According to the table in the docs, they are indices into the model grid. So they can be calculated since the model grid definition is part of the yaml file.
- Something similar is the case for dtstep, tstep, and tstep_glo.
- The index dataset is filled with values, inlcuding negative entries. If this dataset is really the "ID number of each observation point", that should not be possible.
- The obs and obserror datasets are both entirely filled with NaNs. That can of course be a value intended for missing data, but currently, there is no data at all if you generate a monitor file.