Installing modules (SciPy) without root permission.

[Pearu Peterson]

On 5 Aug 2002, Jacek Generowicz wrote:

> I think that SciPy would be useful at this point, but it is not
> available on the machines where we might be working ...
> this brings me to observe two points:
> 
> 
> 1) I often find myself in an environment where the local python
>    installation does not have a module I wish to use (usually
>    Numeric), and where I do not have write access to the python
>    installation directories. 
> 
>    Is there a way to add such modules without re-installing python in
>    my own space ? 

Yes. I used to use PYTHONPATH environment variable for that. For example,
set

  PYTHONPATH=/path/to/my/lib/python2.2/site-packages/

and when installing a third party Python module, use

  python setup.py install --prefix=/path/to/my

> 2) I have been meaning to check-out SciPy for a while now, but the
>    installation process seems almost fractal; each required package
>    itself requires some packages, each of which requires ...  :-) As
>    SciPy is not essential to me, I have not been sufficiently
>    motivated to coplete this process (my strongest effort ground to a
>    halt with gtk incompatibiliy problems, IIRC).
> 
>    Is installing SciPy really as painful as it seems ?  (If so, is
>    there any hope that it will become easier ?)

It depends. What is your platform? On Linux, for example, the installation
has become quite smooth. 

Minimal requirements for building and installing SciPy are:
1) Python, Numeric, and f2py -- they all are easy to install.
2) BLAS (preferable ATLAS) and LAPACK libraries. For most Linux
systems these libraries are readily available.
3) C and Fortran compilers.

Note that there are also SciPy binaries available for Win32, cygwin, and
linux2 systems. See

  http://www.scipy.org/site_content/download_list

In case of any problems, you can contact SciPy users list for advice. See

  http://www.scipy.org/site_content/MailList

> If I am to convince my (reluctant computer user) friend that using
> SciPy might be a good idea, the whole process can't start by wasting a
> whole afternoon (day, week ...) on merely installing it (something
> which he would have to do "back home").

Well, if later SciPy would turn out to be useful then in the worst case,
that is, building ATLAS and LAPACK libraries from their sources (that can
take few hours), the time spent may not be a complete waste...

HTH,
	Pearu