Python_PIC_Gui will run on all major platforms that support python, however Windows is not currently officially supported. Mac OS X or Linux are recommended.
There are four required packages to run the software, and the version is very important. Numpy greater than version 1.11 is incompatible.
- numpy 1.7.1 to 1.11 (1.13 or greater does not work)
- matplotlib 2.0.2
- wxPython 3.0.0 or 4.0.0b1 (anything larger does not work)
- f90nml
- ffmpeg (optional)
To compile the Fortran backend a fortran compiler such as gfortran and a c compiler are recommended. We use gcc or clang on apple.
Python_PIC_Gui should run well on OS X, either the native python or Anaconda. We provide install instructions for both (the native python version looks slightly nicer). To avoid conflicts with any versions python packages currently installed, we recommend installing in a virtual environment.
Check if virtualenv is installed by typing in Terminal
virtualenv
If the command was found, move on. Otherwise install virtualenv. We recommend using easy_install
sudo easy_install virtualenv
The native install instructions should work even if you have Anaconda installed. However, if Anaconda installed a version of virtualenv, you cannot use it for the native install! The work around is to replace virtualenv with /usr/local/bin/virtualenv everywhere in this section of the install guide. You can check the version by typing which virtualenv.
The above command should be the only time root access is required, however virtualenv can be installed by other means without root (we do not decribe them here).
Now download the software:
git clone https://github.com/starside/Python_PIC_Gui
cd Python_PIC_Gui
Create the and activate the virtualenv
virtualenv ENV --python=/usr/bin/python
source ENV/bin/activate
The --python=/usr/bin/python tells the virtualenv to use Apple's python even if other version of python are installed. Now install the dependencies.
pip install --upgrade pip
pip install -r requirements.txt
Now build the Fortran. Note you must perform the steps above first, as the Fortran build process requires numpy.
make python
Almost everything is in place. On Apple, a special version of python is required to access the GUI. Copying Apple's pythonw to our virtualenv works.
cp /usr/bin/pythonw ENV/bin/python
To run the application (for example the Electrostatic code) type
python gui_mbeps1.py
If all went well, the application should appear.
To exit the virtual environment type deactivate. Later, if you want run the PIC gui simple cd to the Python_PIC_Gui directory and run source ENV/bin/activate. Below is the whole sequence of commands, that activates the virtual environment, runs the application, and then deactivates virtualenv.
cd Python_PIC_Gui
source ENV/bin/activate
python gui_mbeps1.py
deactivate
Installation on Anaconda should work for both Mac OS X and Linux. It will probably work for Windows if you can figure out how to compile the Fortran.
In the file mbeps1.source/Makefile disable OpenMP. For whatever reason, Anaconda's numpy does not correctly compile with OpenMP support.
Comment out lines 78 and 79 in mbeps1.source/Makefile then comment in 81 and 82.
Install the requirements listed in the Requirements section. You will probably have to downgrade numpy to 1.11 or lower (If you do not want to downgrade, read the section below). Anaconda has a version of wxPython (3.0.0). On Apple, wxPython should also install python.app, if not you should install it. Once everything is installed type
cd Python_PIC_Gui
make python
On Apple, run the application by typing
python.app gui_mbeps1.py
Using python.app is very important, otherwise the GUI will fail to start.
On Linux you can simply type python gui_mbeps1.py
Anaconda supports its own virtual environment. This section describes how to install in an Anaconda virtual environment so you do not have to downgrade numpy or other packages.
Download the code then create and activate the virtual environment
git clone https://github.com/starside/Python_PIC_Gui
cd Python_PIC_Gui
conda create -n ENV python=2.7
source activate ENV
Install required packages
conda install -n ENV "numpy<=1.11"
conda install -n ENV wxpython
conda install -n ENV matplotlib
pip install f90nml
Disable OpenMP in the mbeps1.source/Makefile and run
make python
To run the application type
python.app gui_mbeps1.py
The installation process on Linux is the similar to the direction on OS X. Using Anacondia the procedure should be identical to OS X, except you do not need to do the step of copying pythonw, or running with python.app.
For non-Anaconda installation, the procedure should follow the Apple Native install instructions (read it!), except Linux does not use easy_install. Instead use your distro's package manager. On Ubuntu, installing virtualenv would look like
sudo apt-get install virtualenv
The only tricky part is getting wxPython. Pip will likely try to build it from source (which will almost certainly fail), so remove wxPython from requirements.txt and see if you can find a binary install of wxPython for your distribution. wxPython provides binary packages for some platforms
https://extras.wxpython.org/wxPython4/extras/linux/
You can also use wxPython 3.0, however it will not install in a virtualenv, so you will have to install everything else outside of a virtualenv.
For Ubuntu, and other Linux distros you can try installing the dependencies with
pip install -r requirements_ununtu1604.txt
The file requirements_ununtu1604.txt can be modified to install any binary version of wxPython in https://extras.wxpython.org/wxPython4/extras/linux/ to suit your distro.