Starting with HiGHS

Question

I have been messing about with HiGHS and trying to get a hang of its capability.

Here is an example of a simple model from their github repo https://github.com/ERGO-Code/HiGHS/blob/master/examples/call_highs_from_python.py

In this example, adding variables and constraints is done through lp.col and lp.row respectively.

I'm getting a handle on how to do that, but here is another example from the repo https://github.com/ERGO-Code/HiGHS/blob/master/examples/minimal.py#L8

where they use addVar and addConstr instead. This of course simpler and more straightforward. Anyway, when I try to replicate that addVar works and all the variables are added. On the other hand, addConstr results in an attribute error AttributeError: 'Highs' object has no attribute 'addConstr'

I have been looking at the documentation with no luck to fix this issue. I have the latest version of the library. I'm wondering if this is something they haven't released yet maybe? I tried also to find the source code to see what function and attribute the model object has but couldn't find that either.

Answer 1

With version 1.5.3 you should add constraints using the addRow(...) method in the format lhs <= constraint <= rhs

m = highspy.Highs()
m.addRow(
  lhs,
  rhs,
  quantity of non-zero variables
  indexes of variables
  coefficients of the variables
)

For instance, if you have a knapscack constraint like $\sum_{i} w_{i} * x_{ij} <= c * y_{j}, \forall j$, then the code should be:

m = highspy.Highs()
for j in B:
  m.addRow(
    -highspy.kHighsInf,
    0,
    N+1,
    [idx_x[i,j] for i in range(N)]+[idx_y[j]],
    [w[i] for i in range(N)]+[-c]
  )

Answer 2

The latest release @ github is 1.6.0

The latest release available @ pypi is 1.5.3. This is the one you obtain when installing through pip install highspy.

While the feature you are trying to use was added (to some branch; see MR) in April 2023 and 1.5.3 was released in May 2023, it seems it's NOT part of 1.5.3 as indicated in this git-compare of the two releases

Well... at least it looks like it. Imho the git-history/branching looks kind of "wild". Your observation of having access to one function but not the other is strange but i think it's a result of some overload (which looks a bit scary).