Open matteodefelice opened 6 years ago
Dear @matteodefelice,
Yes, the rJava package installation it's usually an issue. Basically because the dependency with the JVM version installed on the OS.
The ncdf4 could be a solution but the reason to use the rJava it's because we are using the netcdf-java library which provides a IOSP to access many other formats and not only netcdf files as ncdf4.
The other big advantage it's the Unidata's Common Data Model which provides a Scientific data layer for accessing Grid/Coverage datasets interpreting horizontal (including 2D coordinates in projections), temporal ( including forecast time and runtime), vertical and ensemble/realization discrete coordinates.
AFAIK this scientific layer doesn't have an alternative implementation beyond the netcdf-library and which it would be very useful to have.
I know that I'm not providing the solution to the struggling with the installation of rJava, but to mitigate it we have made strong effort to provide instructions to have a working rJava. Right now we are researching how to create conda recipe to deploy rJava and in particular the right version of JVM without admin privilegies.
I know that I'm not providing the solution to the struggling with the installation of rJava, but to mitigate it we have made strong effort to provide instructions to have a working rJava. Right now we are researching how to create conda recipe to deploy rJava and in particular the right version of JVM without admin privilegies.
Great! That could be an optimal solution. I know that netcdf-java is fundamental to provide reliable & powerful & standard-compliant features but however the overhead to deal with rJava (and rJava + RStudio) is often significant. Looking forward to the conda recipe!
Dear developers, unfortunately in the last months I struggled a lot to have the climate4R working on several machines, mostly due to rJava on machines where I do not have admin rights. In those cases I had to switch to
ncdf4
and then to find a workaround to use the very useful tools in downscaleR and transformeR. I was wondering...is it possible to imagine a loadGridData function with ncdf4 as engine? Obviously without all the features of the actual function, but I think it would be very useful for all the people struggling to have a working rJava on their machine. I would be very happy to contribute to it, if you think that it would be feasible and useful.