[Help-glpk] Linear expressions in the integrands of iterated expressions

[Barry Rountree]

Hello,

Short version:
I'd like to implement O_MAXIMUM using A_FORMULA, but I need 
some help finding where to put the patch.

Long version:
I'm hitting the following error message using glpsol v4.11:

   bpp.mod:76: integrand following max{...} has invalid type
   Context: ...j ] = 6 ; s.t. blrX : max { i in I , j in J } x [ i , j ] <=
   Model processing error

This can be caused, for example, by adding the following line to 
the bpp.mod example:

   s.t. blrMax:max{i in I, j in J} x[i,j] = 1;

Grepping the error message led to this comment in glpmpl1.c:

   /* now the integrand must be of numeric type or linear form
      (the latter is only allowed for the sum operator) */ 

Well, ok, O_MAXIMUM works for A_NUMERIC, and O_SUM works for A_NUMERIC 
and A_FORMULA, so I'll just find where in the pseudocode translation
of the summation takes place and implement a similar method to find the
maximum.  

And I can't seem to find where that happens.  reduce_terms() looked 
like the right choice, but the fp_add() doesn't do what I expect.
I've grepped all of the occurances of O_SUM, but I'm just not seeing
the place where the actual summation occurs.

Any pointers would be most appreciated.

Thanks,

Barry Rountree