Charlie Zender

Greatt! I'm glad the time we invested in this thread "paid off". I've modified the manual to clarify some steps based on your feedback. Unfortunately I don't use CMake often enough to debug your build problem. It has worked fine for the Conda-Forge builds, but maybe it will break with 5.2.9. I will try to prevent more regressions from creeping into this feature in the future. cz

p.s. I could not access the CAMS links in your original post, so perhaps you could attach the original data file

Hi Eric, Please read the new info about ECMWF/IFS grids at http://nco.sf.net/nco.html#vrt_hyb_ifs I think the previous documentation was unclear. In any case there was a regression sometime in the last few years that broke the IFS functionality. The latest NCO snapshot, 5.2.9-alpha01 works for me. I expect to release the 5.2.9-final in October. If you can build it, please LMK if it works for you. Otherwise, I suggest trying NCO 4.9.0. I think that version does not have the bug (and you may not need...

Hi Eric, This is an involved question, but you've written it clearly and my curiosity is picqued so I'll try to help. Give me a day or two to see if I can reproduce your results.

Yes, that would solve the version churn that I mentioned due to different OS's producing different ANTLR files. However, it would be a significant modification to the build system. And would need to be rock-solid in CMake (and preferably Autoconf too) before being adopted. Volunteers welcome :)

Hi Brett, Your interpretation of the role of antlr2 is correct. I am not a fan of the Antlr dependency either. I suppose NCO could bundle the Antlr files instead of building them from source. I think we did so 10+ years ago. It's also possible for you to fork NCO and experiment with generating a useful distribution with the Antlr files included. If you succeed and it seems portable we could merge your changes in from downstream. One problem I seem to remember from the old days is that different OS's...

That command will all the values in the snowfall variable. If the units of snowfall are absolute amounts per day (and not, e.g., rates per second each day) then the sum will be the total snowfall during the period.

First, Is the file that you are modifying a netCDF4 file or a netCDF3 file? I think the size increase is due to the underlying format. In netCDF4 it could be due to an inept chunksize. In netCDF3 it could be due to expanding the size of the record dimension. If time is the record dimension, then increasing it's size will allocate enough space to hold those extra records for every variable that includes the time dimension. Second, the "c" mode in your ncatted command requests "create" mode. This will...