When I use a fixed constraint, the solve result ends up being within tolerance of the fixed value. But I would like it to be the exact value. This is for ergonomics so that callers don't need to use tolerance. They can know that if a fixed value goes in, it will come out exactly in the result.
A few ideas how this could be implemented:
- Set the initial value of the variable to the fixed value
- Somehow weigh the fixed constraint higher
A concern is that for over-constrained systems, ideally we wouldn't treat some constraints preferentially.