Benefits of using this iterator API include:

• treats variables of all types, sizes, and shapes uniformly, whether they fit in memory or not, even if a single row won't fit in memory
• makes recursion unnecessary in handling arrays of arbitrary dimensionality
• accesses data efficiently, using largest blocks that will fit in memory buffer of designated size
• takes care of pesky leftovers, when smaller pieces are needed at the end of a loop to finish what's left
• works for either netCDF-3 or netCDF-4 enabled libraries
• takes advantage of chunking when present, reading a chunk-at-a-time rather than along dimension axes
• supports iterating over multiple variables concurrently, because each variable's iteration state is maintained in its own iterator structure

Other iterator functions used in nccopy iterate through all groups in an arbitrary group hierachy for netCDF-4 files, eliminating need for recursion to traverse groups.

These iterator APIs are available now in C source, but will be exposed and documented in a future version of the APIs, including Fortran support.

2012 Unidata NetCDF Workshop > Developing generic netCDF software