TODO

Here are some small hints which inprovements could be made.


1 Functions to be improved
===========================

1.1 In directory Zassen
=======================
   
   * the function reget_gen should use a stabilizer chain technique like
     applied in the function strong_generators.
     The technique use so far is an orbit algorithm which calculated
     in the point group.
     In any case the function can be improved by really calculating
     only in the point group, or at least do an orbit algorithm which
     only has as many point as the point group has elements.

   * the function identify should return the translational part of the
     transforming matrix as well.

1.2 In directory Voronoi
========================

   * the function normalizer should handle the cases of Bravais groups
     isomorphic to V_4 or C_2 (=<-I_n>) seperately via the normalizer
     of the homogenously decomposable group (which is a direct product
     of GL_n(Z) with possibly a twist for groups with Symbol 1;1
     1,1;1,1  1,1,1;1,1,1 and so on.
     The normalizer will then be calculated via gittstabneu.
     More generally it could handle all Bravais groups which are isormorphic
     to a direct product of C_2's.

   * the function normalizer should return I_n as normalizing element
     if the group is self normalizing.

   * the function normalizer should to more of the calculations in
     gmp-arithmetic.

1.3 In directory Orbit
======================

   * Write an all purpose orbit, which takes the operation,
     the standardisation, and equality via function pointers.

1.4 In directory Base
=====================

   * strong_generators could yield a presentation of the group in question
     if one is prepared to keep track of the words involved and makes a
     lot of tietze transformation.

   * the function normalizer_in_N should check whehter a returned generator
     of the normalizer is itself in the group it normalizes, and then skipp
     it. Tools are there, just adding an if!

1.5 In directory Idem
=====================

   * improve the function bravais_catalog via the following method:
     firstly transform the group to the next homogeneously decomposable
     Bravais group and find this in the catalog (NB: we got this group
     already in the function symbol).
     The when searching for the zclass, take the index (exactly elementary
     divisors) of the lattice in the smallest homogenously decomposable
     group into account. (Note that this has been cacluated in symbol as
     well).

   * the function idempotents just works for finite groups up to dimension
     7. This is a result of the algorithm which factorizes polynomials,
     and it migth give wrong results for higher degrees. Fix this, at
     least get a somewhat higher degree.
     This could be done by considering quadratic factors when factorizing.
     There are two places where changes have to be made (in get_idem2),
     and both of them are documented in the source code.

1.6 In directory ZZ
===================

    * QtoZ could easily return the matrices which conjugate the input
      group to the output one.

    * ZZprog should react on the level option -l in the way one would
      expect, and not calculate childrens of the first lattice in
      the last layer computed. Should be easy to fix.

    * QtoZ calls ZZ several times, and ZZ recomputes the same modular
      constituents every time. should be changed for speed up.

1.7 In directory Sort
=====================

    * we should employ a hash function which assigns each matrix
      an integer straight away, and store it in the matrix record.
      This should speed up searching emormously, because mat_comp
      is quite slow in some instances.

2 Programs to be developed
==========================

Here is a small compilation of programs which are desirable.

    * a program deciding equivalence of space groups in one go. This
      would involve the following strategy:
      - Standardize both given space groups, ie. claculate a presentation
        for the point group (in the natural generators) and transform
        the translation lattice to Z^n (this is now done in
        Extract/Standard_affine_form)
      - extract the Z-class of the point groups and decide Z-equivalence
        (now done in Z_equiv). The following only if they are in the same
        Z-class.
      - Transform the space groups to corresponding generators, which
        is now done in Same_generators.
      - Identify the resulting cozycles, Extract -i
      In all the process: keep track of the transforming matrices!

    * a program which splits a Q-class into affine classes and calculates
      the "graph of inclusion":
      a short description of the algorithm doing this:
      splitting Q-class into Z-classes: see QtoZ.
      splitting a Z-class into affine classes: see extensions
      now to the graph: for each group in the graph calculate
                        all minimal centerings, ie. all minimal
                        invariant superlattices of Z^n (see ZZprog).
                        Now "throw" theses lattices to the group
                        in question and see to which of the groups it
                        becomes isomorphic. This yield the edges of the
                        directed graph, and the vertices are the affines
                        in that Q-class.

    * a program which effectively decides Q-equivalence of groups.
      There is an algorithm outlined in [OPS97].
      A workaround for the moment (which is not good in any consideration):
      Given G,H, and decide whether they are Q-equivalent by calculation
      the Q-class of G (by QtoZ), and see if one of the resulting
      groups is Z-equivalent to H (cf. Z_equiv).