• Object-oriented interface
• Fairly well documented: Users Guide is complete and up-to-date.
• Type safety: better than the C interface.
• Less code than an equivalent C program, and it's object-oriented.
• Short argument lists, because there are no IDs for variables, dimensions, or attributes.
• Easier memory management, because space for blocks of data is allocated by the library.
• Hides "define mode", so no calls are needed to enter or leave define mode for changing the schema of a netCDF file (adding variables, dimensions, or attributes).

• Implemented as a thin layer on top of C library
• Showing its age: written before C++ standards for templates, exceptions, and namespaces
• Student updating the C++ interface for netCDF-4 this summer
• No garbage collection: you have to delete blocks of data and NcAtt objects (attributes) allocated by the library when done with them.
• Less tested than C, Fortran, and Java interfaces.
• No ncgen support: ncgen cannot (yet) generate any C++ code.
• Not used in other Unidata libraries or applications.
• Old-fashioned error handling: no C++ Exceptions are used, so need to check return status from method calls and use an is_valid() method to determine if constructors failed.
• Because "define mode" is hidden by the interface, you may not be aware of inefficiencies when you are changing the schema.

Annotated examples of complete C++ netCDF programs are available from the program examples page.

2007 Unidata NetCDF Workshop for Developers and Data Providers > NetCDF APIs