1. Mature Python language programmers would easily install our soft in 1-2 minutes (usually they first of all prefer installation from subversion repository, at 2nd - installation from PYPI, at 3rd - Linux apt-get, and download source code otherwise). But if you have any problems with either OpenOpt Suite or optional solvers (e.g. it's not that easy to install IPOPT or ALGENCAN on non-Linux OS) you can easily write, edit and run your code on-line via our Sage server (unfortunately, it often hangs due to high load and low available RAM).
Pay attention: To use FuncDesigner with OpenOpt you should have their versions of exactly same date (stable releases or svn snapshots). The same for using SpaceFuncs, DerApproximator with FuncDesigner and OpenOpt.
2. Using Unladen Swallow instead of CPython sometimes can speedup calculations in several times (example), but building/installing US requires lots of time and efforts (especially I won't recommend it for unexperienced Linux users who could encounter problems with CPython path conflicts). Also, Unladen Swallow now is dead end (superseded by PyPy). For installation with PyPy see PyPy webpage.
4. Ensure you have
- Python v. 2.6, 2.7 or 3.x ( only OOSuite v >= 0.34 is compatible with Python 3). Most Linux distributives already have Python installed.
- Some Python distributions have no setuptools included, then you need install it by yourself (for Linux users using software channels is recommended, for example "[sudo] apt-get install python-setuptools" or "[sudo] yum install python-setuptools"
- NumPy ("python-numpy" package is available in Linux software channels, for example use "apt-get install python-numpy" for Debian or *BUNTU Linux). Also, you can try to install it via [sudo] easy_install numpy. since v. 0.39+ SciPy is included into FuncDesigner installation dependencies, but if you have problems with scipy instalaltion, you can just remove it from /FuncDesigner/setup.py file (however, some functionality, e.g. using spars matrices, some solvers, some Stochastic Programming features will be missing).
- Linux users, using software update channels to get NumPy, as well as OpenOpt-related software (like CVXOPT, setuptools, matplotlib, SciPy, with all required and optional dependencies) can yield a little bit obsolete version, but could greatly economy your time and efforts.
- (This item is LEAST recommended, because EPD is free only for educational purposes and 32bit platforms; also, it includes quite primitive Python IDEs, almost like Microsoft Notepad vs Microsoft Visual Studio). Windows and RHEL users may be interested in EPD ("Enthought Python Distribution") - it already includes NumPy, SciPy, matplotlib, etc.
- Linux, Windows, Mac users may be interested in SAGE (license: GPL) - "viable free open source alternative to Magma, Maple, Mathematica, and Matlab" - it contains Python + NumPy + SciPy + cvxopt + much more.
- Python(x,y) is new Eclipse-based scientific Python distribution for Windows and Linux, check the site, mb Mac version is already available.
5. Chose one of the following:
- You can download latest stable release 0.53 from Python Package Index (PYPI) via easy_install (provided you already have python-setuptools installed, just type in command prompt "easy_install openopt", "easy_install FuncDesigner", "easy_install DerApproximator", "easy_install SpaceFuncs"; AFAIK most downloads of our software are performed in this way) or directly:
- but you should take into account
|OpenOpt DFP doesn't work w/o derivative||Not yet, you could use FuncDesigner with NLP|
and automatic differentiation instead
|scipy.sparse.coo.py ValueError: row, column, |
and data array must all be the same length
|in v. 0.5301|
|interalg bug "NoneType is not indexable"||in v. 0.5304|
|interalg bug "ODE instance has no "timeVariable" attribute"||in v. 0.5305|
|Uploaded to PYPI|
|Some changes for PyPy compatibility||in 0.5301|
|FuncDesigner translator can handle sparse derivatives from AD||in 0.5301|
|New interalg parameter rTol (relative tolerance, default 10-8)||in 0.5305|
- You can download zip-files generated by Trac automatically from latest svn repository version:
- You can use subversion (if you have svn client installed, e.g. Tortoise svn for Windows):
- svn co svn://openopt.org/PythonPackages OOSuite
- or separate packages:
- svn co svn://openopt.org/PythonPackages/OpenOpt OpenOpt
- svn co svn://openopt.org/PythonPackages/DerApproximator DerApproximator
- svn co svn://openopt.org/PythonPackages/FuncDesigner FuncDesigner
- svn co svn://openopt.org/PythonPackages/SpaceFuncs SpaceFuncs
6. Run installation:
- [sudo] python setup.py install ; or
- [sudo] python setup.py develop
- Windows: you should run setup.py and choose "install" option in menu.
- You could also run setup.py from a filemanager (or python setup.py from terminal) and then chose option 2 (install) and (optional) change installation path to another (however, that one should be present in PYTHONPATH)
7. To ensure OpenOpt has been installed successfully you could go to directory "examples" and run "python glp_1.py" or "python nlp_1.py"
8. Found our soft useful? Please read our Appeal.