This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.
David, We're trying to learn enough about chunking to improve the documentation in that respect and give some good guidance, but as you've discovered, we aren't there yet. Your definition for chunksize is exactly right for the Fortran90 interface, chunksizes= (/ 32,64,64,1 /) but in looking at this we just discovered that the F90 API doesn't reverse the chunksizes when calling the corresponding C function. Ed just fixed that, and the fix will be in the snapshot release created tonight. But until then, you could test that this really is the problem by reversing your chunksizes to use the C order, and let us know if things work with that change: chunksizes= (/ 1,64,64,32 /) If that works, you'll have to reverse them again to run against the bug-fixed release after tonight. I don't know what the non-helpful "HDF error" means, other than that we are not translating HDF5 errors into helpful error messages at the netCDF layer. It seems like the chunksizes you gave, while non-optimal if the order is reversed, should still have worked ... --Russ Russ Rew UCAR Unidata Program address@hidden http://www.unidata.ucar.edu Ticket Details =================== Ticket ID: NNS-961385 Department: Support netCDF Priority: Normal Status: Closed