howto_HDF5

This is a HOWTO for HDF5 presented at the IFM of the University of Michoacan (MEXICO) on July 12.

The official HDF5 documentation can be found here.

Requirements

• gcc (with gfortran)
• h5py
• matplotlib
• numpy

Installing HDF5

Download the latest version (or the version suitable for your system) of HDF5 from here

Go to the directory where the source code was uncompressed and compile the library

$ ./configure --prefix=/path/to/lib/dir --enable-fortran --enable-cxx
$ make -j6
$ make install

The argument in the prefix flag is the full address of the directory where we want our library to be installed. This path must be declared in the environment variable PATH in your init file (e.g., .bashrc, .profile, .zbashrc, .zprofile, etc.). If you are admin or have root permission, the recommended address is /usr/local/bin

Installing h5py

$ pip install h5py

How to work with this totorial

Clone this repository

git clone https://github.com/altjerue/howto_HDF5.git

Go to howto_HDF5 directory and you will have access to the example files in this repository.

Saving data in FORTRAN

In the file firsth5.F90 I have placed a simple code in which I show how to save scalars, vectors and matrices in a HDF5 formatted file.

For the case of 2D matrices I have written two blocks saving two different matrices and with different data set label: y(t = 1) and y(t = 2). This to exemplify how a N-dimensional array can be saved for, e.g., different times.

Working with HDF5 files in python

To manipulate HDF5 files in python in has to import the module

import h5py

In the file ploth5.py I show a simple script of how to extract data from an HDF5 file and plot it. Tu run this script one simply enters to the python interpreter. I recomend installing IPython. So we go to the terminal and type

ipython

Once inside the interpreter we just type:

In [1]: run ploth5.py

The above instruction is the same as writing the content (line by line) of ploth5.py in the interpreter.