Move the soft inequalities directly into the solver.

- Clean up.

Fix BALMLayout archiving issues.

SharedSolver was archiving too many constraints, partly because of
multiple typos, also because it archived some which were just artifacts
of the layout process. These extra constraints are created when the
layout calls SetRange() on the left/top/right/bottom tabs during layout.

* LinearSpec/ActiveSetSolver had to be adjusted to get access to the
constraints added by the SetRange() calls.
* BALM::TabBase was adjusted to avoid a segfault during unarchiving,
caused by an unitialized member.
* ALMFriendLayoutTest was adjusted to include a more obvious custom
constraint for testing.

Refactor ActiveSetSolver::Find{Min|Max}s().

Introduce SharedSolver class to improve BALMLayout's friend feature.

Cleanup the constraint solver a bit and fix a bug when a variable is removed.

Add support for soft inequality constraints for quadratic solvers. A slack variable is used in a normal hard inequality and the slack variable is minimized with a high penalty. In this way the constraint is only violated when necessary. Some smaller clean up.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42225 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fix min/ max calculation by removing all soft constraints before doing so. Smaller fixes.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@40297 a95241bf-73f2-0310-859d-f6bbb57e9c96

Add an alternative solver to lp_solve. The solver based on Ingo's active set solver but is able to handle arbitrary hard and soft constraints. The advantage to lp_solve is that the active set solver can optimize variable in respect to a quadratic objective function. This makes it possible to minimise the quadratic derivation to a desired value e.g. \Sum_i(x_i - x_{i,pref})^2 -> min.

The solver part has been refactored in this way that both solver can be used with the same layout specifications. The active set solver is default now; the performance is not as good as lp_solve, though.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@40285 a95241bf-73f2-0310-859d-f6bbb57e9c96

Move the soft inequalities directly into the solver.

- Clean up.

Fix BALMLayout archiving issues.

SharedSolver was archiving too many constraints, partly because of
multiple typos, also because it archived some which were just artifacts
of the layout process. These extra constraints are created when the
layout calls SetRange() on the left/top/right/bottom tabs during layout.

* LinearSpec/ActiveSetSolver had to be adjusted to get access to the
constraints added by the SetRange() calls.
* BALM::TabBase was adjusted to avoid a segfault during unarchiving,
caused by an unitialized member.
* ALMFriendLayoutTest was adjusted to include a more obvious custom
constraint for testing.

Refactor ActiveSetSolver::Find{Min|Max}s().

Introduce SharedSolver class to improve BALMLayout's friend feature.

Cleanup the constraint solver a bit and fix a bug when a variable is removed.

Add support for soft inequality constraints for quadratic solvers. A slack variable is used in a normal hard inequality and the slack variable is minimized with a high penalty. In this way the constraint is only violated when necessary. Some smaller clean up.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@42225 a95241bf-73f2-0310-859d-f6bbb57e9c96

Fix min/ max calculation by removing all soft constraints before doing so. Smaller fixes.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@40297 a95241bf-73f2-0310-859d-f6bbb57e9c96

Add an alternative solver to lp_solve. The solver based on Ingo's active set solver but is able to handle arbitrary hard and soft constraints. The advantage to lp_solve is that the active set solver can optimize variable in respect to a quadratic objective function. This makes it possible to minimise the quadratic derivation to a desired value e.g. \Sum_i(x_i - x_{i,pref})^2 -> min.

The solver part has been refactored in this way that both solver can be used with the same layout specifications. The active set solver is default now; the performance is not as good as lp_solve, though.



git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@40285 a95241bf-73f2-0310-859d-f6bbb57e9c96