I was trying to write a HDF5 file using the H5Py package. What I wanted to create was a HDF5 dataset whose dataspace is
NULL. This is what I wanted (when dumped by h5dump):
I could not find any input parameters to the
create_dataset call that could achieve this. The closest was by passing an empty tuple to the
However, that created a dataset with this structure:
It turns out that the H5Py had no support to create a NULL dataspace! Thankfully, support for this had just been added to v2.7 as mentioned here.
I used the latest version of H5Py and was able to create a dataset with NULL dataspace. This is done by passing
None to the shape parameter.
Tried with: Ubuntu 14.04
HDF5 is the most popular file format for storing binary data structures. The HDFView tool makes it convenient to view and browse the contents of a HDF5 file in a Java GUI. But what if you are at the commandline and need a way to quickly view or dump the contents of a HDF5 file? The h5dump does exactly this.
- It can be installed easily:
$ sudo apt install hdf5-tools
- To dump the contents of a file:
$ h5dump foobar.h5
- By default, you will notice that some data like string arrays are shown in a matrix format. To display all contents in one column format:
$ h5dump --width=1 foobar.h5
$ h5dump -w 1 foobar.h5
Tried with: HDF5 Tools 1.8.16 and Ubuntu 16.04
If you are working with people who use MATLAB, they might sometimes pass you data as a HDF5 file. HDFView is a tool written in Java that can be used to view the contents of such files.
Installing it is easy:
$ sudo apt install hdfview
.h5 file using HDFView and it will display the hierarchy of the contents as a tree on the left. Click on the tree items to view its properties in the bottom. Double-click on the tree items to view its content in the main window.
Tried with: HDFView 2.9 and Ubuntu 14.04