Currently Online Lab is available only from its git repository
hosted at GitHub. To get a copy of this repository, issue:
$ git clone git://github.com/hpfem/femhub-online-lab.git
or, alternatively:
$ git clone http://git.hpfem.org/git/femhub-online-lab.git
to clone via HTTP protocol (in case of e.g. firewall issues).
Online Lab runs under any Unix-like operating system that implements
non-blocking polling functionality, e.g. epoll (Linux), kqueue
(BSD) or select (universal, worst case scenario). To obtain best
performance, Linux operating system distribution with epoll support
should be used (kernel 2.6 or better is required).
The required packages for Online Lab are:
- Python >= 2.6 (http://www.python.org)
- Tornado >= 1.1 (http://www.tornadoweb.org)
- Django >= 1.1 (http://www.djangoproject.com)
and smaller, but not less important, are:
- pyinotify (http://github.com/seb-m/pyinotify)
- argparse (http://code.google.com/p/argparse)
- lockfile (http://pypi.python.org/pypi/lockfile)
- daemon (http://pypi.python.org/pypi/python-daemon)
- psutil (http://code.google.com/p/psutil)
- pycurl (http://pycurl.sourceforge.net)
- docutils (http://docutils.sourceforge.net)
For example, in Ubuntu Lucid issue:
$ sudo apt-get install python python-django python-pyinotify python-argparse python-lockfile python-daemon python-psutil python-pycurl python-docutils python-pygments
to get those packages installed. Note that Tornado didn't manage to get into software package management systems yet (e.g. apt-get or portage), so you have to install it manually, either by downloading its source code tarball from Tornado's website:
$ wget http://github.com/downloads/facebook/tornado/tornado-1.1.tar.gz
or by cloning its git repository that is hosted at GitHub. Make sure
that all required packages are available on PYTHONPATH before running
Online Lab.
If you use Python 2.7 or better, then you don't need to install argparse module, because it is included in Python standard library since 2.7 (see PEP 389 for details).
You can also run Online Lab in FEMhub numerical software distribution, where all necessary packages were included in appropriate versions, together with Online Lab it self.
Suppose Online Lab's repository was cloned into /home/lab:
$ cd /home/lab/femhub-online-lab
We have to create two work environments, one for Online Lab core server and the other for Online Lab services:
$ bin/onlinelab core init --home=../core-home --ui-path=ui $ bin/onlinelab service init --home=../service-home
onlinelab script automatically adds current directory to PYTHONPATH
so you don't have to worry about module visibility issues. The directories
with work environments will contain some configuration files and additional
subdirectories for storing runtime data (logs, PID files, blobs, etc.). In
the case of the core server, a database is also created and user interface
(UI) is set up.
Now open two terminals in parallel and run the following commands:
$ bin/onlinelab core start --no-daemon --home=../core-home $ bin/onlinelab service start --no-daemon --home=../service-home
These will start two servers listening on localhost (8000, 9000). Now go to your browser (preferably Firefox or Chrome) and redirect to:
http://localhost:8000
Login screen will appear, where you can create an account and finally proceed to Online Lab's desktop. Click 'Help' icon to show a tutorial about main features of the system.
If you are not interested in watching the output from Online Lab, the
you may consider running both core server and services as daemons
(just remove --no-daemon from the commands above). In this case,
you still will be able to read the logs that are stored in Online Lab
home directories.
To stop Online Lab simply press Ctrl+C in terminals in which
core and services are running. In daemon mode use the following
commands:
$ bin/onlinelab core stop --home=../core-home $ bin/onlinelab service stop --home=../service-home
Note that the order of running servers is relevant and core server must be started before services are started. However, stopping can done in any order.
If you have auxiliary Python modules that you would like to expose in
Online Lab (e.g. SymPy) and those modules aren't available on system-wide
PYTHONPATH for some reason (e.g. you would like to expose a certain
branch of a development repository), then add paths to those modules via
--python-path command-line option, e.g.:
$ bin/onlinelab service start --python-path=/devel/sympy
assuming that SymPy's module is located in /devel/sympy. You can also
use colon-syntax to add multiple paths:
$ bin/onlinelab service start --python-path=/devel/sympy:../numpy
You can also add multiple --python-path options and/or store them in
services' configuration files.
To install Mesh Editor (Flex) in Online Lab, clone its repository:
http://github.com/hpfem/mesheditor-flex
and follow its build instructions (see README). Next copy MeshEditor.swf
into static/external in core server's home directory, e.g.:
$ cp MeshEditor.swf /home/lab/core-home/static/external
and reload Online Lab user interface in your web browser. Next time you double-click Mesh Editor icon on the desktop, the plug-in will be loaded.
Go to http://localhost:8000, open Browser and click 'Import'. Copy plain text from Sage worksheet, e.g.:
{{{id=0|
some code
///
output
}}}
and click 'OK'. A new window will appear with all cells imported.
Online Lab's JavaScript engine is based on Google's V8 JavaScript engine, so additional dependencies have to be installed:
Detailed installation instructions can be found at:
There is a lot of reading to be found there, but it's sufficient to issue the following commands to install both libraries, assuming 64-bit architecture:
svn checkout http://v8.googlecode.com/svn/trunk/ v8 svn checkout http://pyv8.googlecode.com/svn/trunk/ pyv8 cd v8 CCFLAGS=-fPIC scons arch=x64 library=static cd ../pyv8 patch -p0 -i <path-to-online-lab-repo>/contrib/pyv8-sigint.diff V8_HOME=../v8 python setup.py build
Just make sure you don't use multi-lib with V8 (arch=x64) and that V8 is
compiled as a static library, otherwise PyV8 won't compile. After successful
compilation, put PyV8.py and _PyV8.so on PYTHONPATH, e.g.:
ln -s `readlink -f .`/build/lib.linux-x86_64-2.6/* ~/.local/lib/python-2.6/site-packages/
If you get engine-died error when using JavaScript engine via Online Lab
services, then it's most likely the case that PyV8 modules aren't visible in
Online Lab.
JavaScript engine's entry is automatically added to the database during initialization of Online Lab's core, however, if you want to use an existing installation then you have to add this entry manually, e.g.:
$ onlinelab core run --home=<path-to-core-home> --code=<<EOF > from onlinelab.core.models import Engine > Engine.objects.create(name="JavaScript") > EOF