Re: [Bug-glpk] Conflict graph issue

[Andrew Makhorin]

> I have been playing for a while with the conflict graph (CG) library
> "cfg.h" and I think I found a minor detail that can possibly be
> improved. I was extracting the CG of the instance gt2.mps (attached
> file) from the MIPLIB3.0 library,
> 
> 
> The conflict graph of this instance is empty, meaning that the only
> edges correspond to the trivial conflicts, i.e, (x_i=1,x_i=0). Before
> running the code, I didn't know that the CG was empty, otherwise I
> wouldn't have done it. 
> 
> 
> this instance has 188 integer variables, 24 which are binary. I found
> that glpk generated 24 vertices as well as the following info.
> 
> 
>  
>         G.ref
>        G.neg
>        G.pos
> [1]
>                33
>                 0
>                 0
> [2]
>                34
>                 0
>                 0
> [3]
>                35
>                 0
>                 0
> [4]
>                36
>                 0
>                 0
> [5]
>                37
>                 0
>                 0
> [6]
>                38
>                 0
>                 0
> [7]
>                39
>                 0
>                 0
> [8]
>                40
>                 0
>                 0
> [9]
>                41
>                 0
>                 0
> [10]
>                42
>                 0
>                 0
> [11]
>                43
>                 0
>                 0
> [12]
>                44
>                 0
>                 0
> [13]
>               100
>                 0
>                 0
> [14]
>               108
>                 0
>                 0
> [15]
>               116
>                 0
>                 0
> [16]
>               124
>                 0
>                 0
> [17]
>               132
>                 0
>                 0
> [18]
>               140
>                 0
>                 0
> [19]
>               148
>                 0
>                 0
> [20]
>               156
>                 0
>                 0
> [21]
>               164
>                 0
>                 0
> [22]
>               172
>                 0
>                 0
> [23]
>               180
>                 0
>                 0
> [24]
>               188
>                 0
>                 0
>  
> This makes sense because the CG is empty. However, something that was
> strange to me is that, when I was extracting the adjacency lists of
> the vertices (using degree = cfg_get_adjacent(G, i, adj);  ), degree
> was = 11 and the adj array was not empty. 
> 
> 
> I think in this case degree should be equal to 1 and adj  a singleton
> [0] (taking into account that glpk vectors start at 1).
> 

>From your printout it follows that no binary variable is included in the
conflict graph (since all G.neg and G.pos are zero); therefore, G.nv,
the number of vertices in CG, also should be zero. In this case any call
to cfg_get_adjacent would cause an error, because on entry it checks
that 1 <= v && v <= nv. Most likely you inspect CG at some wrong point,
i.e. before it has been completely built.