Alternatives to the use_strategic argument for solvers #720
Description
Requested enahancement or feature
In solvers like lp_solve and lcp_solve, at the moment use_strategic defaults to False.
When the game does not have a tree representation, we must use the strategic representation, and so use_strategic has no effect.
It is not ideal that when a user has a non-tree game and passes in use_strategic as False they get no feedback to say that we must use the strategic form in this case. This is not a bug, the docs are clear, it just does not seem idea.
If the default was True (not suggesting that we should do this), then at least one could give a warning if one ever tries to set it as False for a non-tree, which is more consistent in some sense with always being able to use the strategic representation but not always having a tree. I note that having a use_tree argument that defaults to True, has the symmetric issue of when a user tries to set it explicitly to True when calling on a non-tree -- it makes not sense for them to include that argument in that case, but there is also no way to give feedback since that is the default.
Maybe the current setup is fine, but this did all lead me to wonder if we simply want different functions for the solvers on different representations, though I know that would move us away from the approach taken in many modern scientific computing packages.