Matt
(You are presumably speaking of the grid structure stored in HDF5 output files, including in checkpoint files.) What do you intend to do with the grid structure you so obtain?
If you are interested in the exact grid structure, then yes, this is the correct way to do so. For example, when Carpet reads a grid structure e.g. for recovery, it uses this information.
However, if you e.g. want to visualise data, then there is an alternative. Each dataset (HDF5 component) contains attributes describing the coordinate of its upper left corner as well as the coordinate distance between grid points (grid spacing). This should suffice to correctly combine the datasets you read in, and may be much simpler than parsing the grid structure.
Note that the VisIt visualisation toolkit knows how to read Carpet output. You could look at its reader for details.
If you are really interested in parsing the grid structure, then it may be easier to discuss this in person. Are you available on Monday morning 11:00 EST for the Einstein Toolkit telecon? I can explain things there.
-erik