===== SfePy =====
~~NOTOC~~
<block right third box round>
== Website ==
[[http://sfepy.org/doc-devel/installation.html]]

== Requirements ==
<code bash>
module load mpi
</code>

== Result ==
appdirs==1.4.0\\
apptools==4.4.0\\
configobj==5.0.6\\
cycler==0.10.0\\
Cython==0.25.2\\
ez-setup==0.9\\
FiPy==3.1.3\\
functools32==3.2.3.post2\\
h5py==2.6.0\\
matplotlib==2.0.0\\
mpi4py==2.0.0\\
mpmath==0.19\\
numexpr==2.6.1\\
numpy==1.12.0\\
packaging==16.8\\
petsc==3.7.2\\
petsc4py==3.7.0\\
pyface==5.1.0\\
Pygments==2.2.0\\
PyMetis==2016.2\\
pyparsing==2.1.10\\
python-dateutil==2.6.0\\
pytz==2016.10\\
scikit-umfpack==0.2.3\\
scipy==0.18.1\\
sfepy===2016.4-git-064c86ad4a95d7c3c86e3414bc286003f3e41c0e\\
six==1.10.0\\
subprocess32==3.2.7\\
sympy==1.0\\
tables==3.3.0\\
traits==4.6.0\\
traitsui==5.1.0\\

</block>

==== Usage / Virtualenv ====

SfePy is a very tricky piece of software. It requires multiple other libraries for its diverent modules. Experience has shown it is best used in its own python virtualenvironment.

==== Preparation of the Virtualenv ====


== Packages to install by pip ==

=== simple pip packages ===
cython\\
numpy\\
scipy\\
sympy\\
matplotlib\\
mpi4py - requires **''module load mpi''**\\
h5py      \\
pymetis\\
pyparsing\\
scikit-umfpack\\
tables\\
scikit-umfpack\\
ez-setup (breaking build of SfePy shows no sign that this package is neeeded, but it is)

=== not that simple packages ===

== petsc4py ==
Petsc4Py can be installed from pip. The installation shows some errors. Petsc missing, but it is build. And the resulting python packages are working.


== FiPy ==
FiPy can be installed with pip without any problems after installing all the packages above.


== SfePy at least ==
SfePy is installed from source.

<code bash>
git clone git://github.com/sfepy/sfepy.git
</code>

In case you wish to use a specific release instead of the latest master version, use: **''git tag -l''**

<code bash>
python setup.py build_ext --inplace
</code>

<code bash>
python setup.py install
</code>



=== not used packages ===
== PySparse ==
PySparse can be installed from pip or from source, neither is working easily with newer glibc versions.
It is possible from source with some CFLAGS. Sadly to say, working parameters are not documented (only known to Mr. Wu)

<code bash>
git clone git://pysparse.git.sourceforge.net/gitroot/pysparse/pysparse
</code>

== Mayavi ==
If Mayavi is needed, the Virtualenv has to be build with ''** --system-site-packages**''. I've found no way (dependencies) to install it in an system independent virtualenv.





== Result / Tests ==

To test the instalation SfePy provides a test-script.

<code bash>
python run_tests.py
</code>

98 test file(s) executed in 195.82 s, 3 failure(s) of 165 test(s)