Hi,
This issue has been talked about many times here, but I would like to
bring it up again. That is the change from Numeric to numpy, as well
as the dependence on Scientific python. For discussions about numpy,
see the discussion threads starting at:
numpy discussions:
https://mail.gna.org/public/relax-devel/2006-08/msg00005.html
Numpy v. Numeric; Scipy V Scientific:
https://mail.gna.org/public/relax-users/2006-09/msg00014.html
[bug #9732] Python 2.5 - failure in constructing the gradient of the
global model: http://gna.org/bugs/?9732 or
https://mail.gna.org/public/relax-devel/2007-08/msg00005.html
ScientificPython and PDBs:
https://mail.gna.org/public/relax-users/2006-09/msg00041.html
Numeric is no longer maintained (and I think possibly deliberately
broken, well at least not fixed, to encourage people to change). We
had previously considered to move to numpy but decided not to mainly
because ScientificPython also depends on Numeric. But now if you go
to the ScientificPython web page,
http://starship.python.net/~hinsen/ScientificPython/, there is nothing
there. So my proposal is as follows. We migrate all of relax to
numpy 1.0.4 and higher as the linear algebra code, from my limited
testing, now appears to work. Numeric is completely removed from the
picture. Then we make ScientificPython an option and allow other PDB
readers into relax, again as options. The user function
structure.read_pdb(), which will have the PDB reader as a function
argument, will then throw a RelaxError if the package is not
installed. And finally, the minimisation code should not be migrated
to numpy because of the new project https://gna.org/projects/minfx/.
Hence the minimise/ directory will soon be removed from the 1.3 line.
Does anyone see any problems with these ideas?
Cheers,
Edward
relax dependencies and minfx: Migration of the 1.3 line from Numeric to numpy, alternative PDB readers, and a new project.