[Help-glpk] status of close-to-zero output rounding API

[Robbie Morrison]

Hello Andrew, API users

Just wondering about the future of the following API
call.  It was removed from the API manual following
version 4.28 or thereabouts.  But it does compile under
4.41.  And it worked fine under about 4.39 when I last
checked.  The API is:

  lpx_set_int_parm(d_prob, LPX_K_ROUND, 1)

  LPX_K_ROUND type: integer, default: 0
  Solution rounding option
    0 = report all primal and dual values "as is"
    1 = replace tiny primal and dual values by exact zero

Perhaps some kind of new hierarchical problem-wide
parameter setting is needed.  Something like:

  glp_probcp parm;             // problem-wide control parameters
  glp_init_probcp(&parm);      // initialize with defaults
  parm.zero_rnd = -10;         // trip to zero at 1.0e-10
  parm.msg_lev  = GLP_MSG_ERR; // [1]
  glp_load_probcp(P, &parm);   // load the control parameters

  // [1] initially problem-wide but able to be locally
  // overwritten by 'glp_iocp' for instance

One might also consider doing this at the GLPK
environment level, although I am not sure this is
particularly necessary.  In any case, that would be:

  glp_envcp parm;     // environment-wide control parameters
  // and so on ..

For the record, I also recover and report the following:

  LPX_K_ITCNT
  Simplex iteration count

Indeed, the two LPX_ parameters discussed here and
their associated set and get calls are the only
old-style APIs left in my code.

Just to reinforce the comments from a poster a while
back, I really appreciate the clean interface that GLPK
offers and fully support API changes that lead to
better consistency.

with best wishes
Robbie
---
Robbie Morrison
PhD student -- policy-oriented energy system simulation
Technical University of Berlin (TU-Berlin), Germany
University email (redirected) :
address@hidden Webmail (preferred) :
address@hidden [from IMAP client]