relax SVN archive

A overhaul of the handling of the tensor PDB atomic data and how atoms are added and connected.

A new data structure called 'self.atomic_data' has been created to store all the atomic data. The
data structure is a dictionary of arrays. The key is a string corresponding to the 'atom_id'. The
elements of the array are:
0: Atom number.
1: Element type.
2: The x coordinate of the atom.
3: The y coordinate of the atom.
4: The z coordinate of the atom.
5+: The bonded atom numbers.

Two new functions have been added to maintain the 'self.atomic_data' structure. The first is
'self.atom_add()' which creates a key-value pair within the data structure for the given atom adding
the atom number, element type, and coordinates of the atom.

The second function is 'self.atom_connect()'. Two arguments are expected, the 'atom_id' and
'bonded_id' strings. The atom number corresponding to 'bonded_id' is added to the 'atom_id' array
and vice versa.

The centre of mass atom is now added using the 'self.atom_add()' function. The atom corresponding
to the unique axis of the spheroid ('Dpar_vect') is also added using this function. These two atoms
are connected using 'self.atom_connect()'.

The 'self.write_pdb_file()' function has been modified accordingly.

The ability for different atoms of the PDB tensor representation has been added.

The hetatm array elements are now 2 element arrays consisting of the atom position and the element
type. An algorithm has been added to generate the chemical formula for the FORMUL PDB record.

The length of the vector representing the unique axis of the spheroid is now proportional to Dpar.

To calculate the new vector, the Dpar unit vector is multiplied by the eigenvalue Dpar. It is also
multiplied by the scaling factor 'scale'. This scaling factor comes from the prompt interface user
function 'pdb.create_tensor_pdb()' which defaults to 1.8e-6. The result is that the correlation
times translate to:

3 ns -> 100 Angstrom
10 ns -> 30 Angstrom
30 ns -> 10 Angstrom

The docstring of the 'pdb.create_tensor_pdb()' user function has been expanded to include a
description of the scaling factor and how the size of the tensor geometric object is larger for
smaller objects.

The unique axis of the spheroid is now represented by the Dpar unit vector in the tensor PDB.

As the Dpar unit vector has been added to the centre of mass to create a HETATM record, the distance
between the atom representing the centre of mass and the unique axis is 1 Angstrom.

The diffusion tensor data structure now returns unit vectors parallel to the tensor axes.

The 'self.__getattr__()' function has been expanded. The docstring now describes all the equations
used in the calculation of the parameters Diso, Dper, Dpar, Dratio, Dx, Dy, and Dz. It also
describes how the unit vectors parallel to Dpar, Dx, Dy, and Dz are related to the spherical or
Euler angles.

Dper, Dpar, and Dratio are now only calculated if the diffusion tensor is spheroidal and Dx, Dy, and
Dz are only calculated if the diffusion is ellipsoidal.

The Dpar, Dx, Dy, and Dz unit vectors are calculated and a Numeric array is returned.

Fix for bug #7540 (https://gna.org/bugs/index.php?7540).

This bug was reported by Alex Hansen (viochemist at gmail dot com). The problem was that the
function 'self.relax.generic.pdb.load()' had been renamed to 'self.relax.generic.pdb.read()'. The
correct function is now being called and the heteronucleus and proton name arguments, as well as the
'calc_vectors' argument have been removed.

Ported r2674 from the 1.2 line.

The command used was:
$ svn merge -r2673:2674 svn+ssh://bugman@.../svn/relax/1.2

This includes a number of improvements to the optimisation tests of the test suite. See the post at
https://mail.gna.org/public/relax-devel/2006-10/msg00145.html (Message-id:
<7f080ed10610260854p3d7d0a45w32c731345ed244d3@...>) for the exact details.

Improvements to optimisation tests of the test suite.

These improvements include all of the suggestions by Chris MacRaild at
https://mail.gna.org/public/relax-devel/2006-10/msg00141.html (Message-id:
<1161860754.3525.27.camel@...>).

The exact changes are detailed in the follow up post located at
https://mail.gna.org/public/relax-devel/2006-10/msg00145.html (Message-id:
<7f080ed10610260854p3d7d0a45w32c731345ed244d3@...>).

Ported r2671 and r2672 from the 1.2 line.

The command used was:
$ svn merge -r2670:2672 svn+ssh://bugman@.../svn/relax/1.2

This improves the optimisation tests and makes them a little more lenient to allow for machine
precision and round-off error artifacts on different architectures, operating systems, etc.

Fixed a bug in the test suite introduced by the previous revision.

The bug was that the optimisation warning flag for the steepest descent and coordinate descent
optimisation algorithms was set to 'Maximum number of iterations reached' as optimisation was
terminated early. This is now caught.