subclassing Scientific.IO.NetCDFFile

[Jon Beezley]

Hi all,

I have come across a problem that I am unsure how to get around.  What
I want to do is create a subclass of Scientific.IO.NetCDFFile, but
despite what the docstrings say isn't really a class at all, but a
function that returns some sort of data structure from the netcdf C
api.  I am aware of pure python implementations such as
scipy.io.netcdf_file, but these seem to have internal limitations
causing overflows when opening large (e.g. several gigabytes) files.

I have tried creating a class with an open NetCDFFile as a member
(f).  File attributes are easy enough to deal with
(self.__dict__=self.f.__dict__), and I can even create callable
methods such as self.dimensions which return self.f.dimensions.
However, when it comes to more esoteric aspects such as multi-
dimensional slicing of the variables (derived somehow from
numpy.array), I am somewhat at a loss.

So my question is: Is there a simple method of containing NetCDFFile
in a class?  If not, how can I design a class that will behave like
NetCDFFile from the user perspective?

Thanks,
Jon