Creating a custom variable¶

Suppose one wants to perform an operation on data represented by some pygeode variable that cannot be done by existing pygeode variables. One can always load in the data directly, access the underlying numpy array, and perform the operation directly; often this is the fastest and most straightforward approach.

Alternatively one can defined a new pygeode variable type in order to perform the operation within the pygeode context. This can involve more work, certainly if you are not familiar with the inner workings of PyGeode. This has several advantages; metadata consistency, tools for manipulating and interpreting the grids defined by an arbitrary pygeode variable, and often robustness. Once you get the hang of it, pygeode implementations can often be nearly as simple as a direct implemention, and are far more versatile (the former is _not_ always true :)).

variable construction, subclassing an appropriate var object views and selecting axes accessing data reducing memory footprints, rule of thumb