Hi Werner, On 11-04-13 16:40, Werner Benger wrote: > Hi, > > h5dump dumps just everything, so it remembers how it found each component. > > Scanning an entire file for a particular information, whether it be a > dataset > or a named datatype, is of course inefficient. > > Do you have any control about how the file is written or any knowledge > about the file layout that may help you to predict in advance where > the named data type might be? What are you planning to do with the > named data type?
I plan to put additional information about e.g. a compound datatype into an attribute attached to the datatype. This information will give additional information about the members. It is an option to put all datatypes into one directory and thus reduce the number of entries that I would have to scan to find the right committed datatype. That, however, makes the solution specific to the group layout chosen, which I want to do only if there is no better way of getting this information. Best, Bernd > On Wed, 13 Apr 2011 07:40:59 -0500, Bernd Rinn <[email protected]> > wrote: > >> Hi, >> >> Consider I have a dataset that uses a committed (named) datatype, what >> is the most efficient way to find the path of this datatype in an HDF5 >> file? Looking at the the API calls in H5T, the only way I can see is to >> scan all groups for committed datasets and calling H5Tequal() on each of >> them to find out whether this one equals the datatype of the dataset. >> The scanning might be very inefficient if the HDF5 file contains many >> groups with many members. Is there a more efficient way of doing that? >> How does h5dump get the information that it prints in the "DATATYPE" >> line for a dataset? >> >> Bernd >> >> _______________________________________________ >> Hdf-forum is for HDF software users discussion. >> [email protected] >> http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org _______________________________________________ Hdf-forum is for HDF software users discussion. [email protected] http://mail.hdfgroup.org/mailman/listinfo/hdf-forum_hdfgroup.org
