A step of the miprgr_native implementation. No other function in the package calls it. See the file header for the source the module follows.
Examples
cat(miprgr_cheatsheet())
#> miprgr: the LP relaxation is easy and usually FRACTIONAL, and ROUNDING is not a fix -- the rounded point is often infeasible and, when feasible, often strictly worse. BOUND: the relaxation's value bounds every integer point below that node, so once an incumbent exists any node no better than it can be discarded UNEXPLORED. BRANCH: Land and Doig enumerated values; DAKIN's dichotomy x_j <= floor(v) OR x_j >= ceil(v) excludes the fractional point, keeps every integer point, and leaves each node an LP -- so the whole search is a binary tree of LPs. Pruning must not change the optimum: run without it and compare.