[m-rev.] the grade library is ready for review

[Zoltan Somogyi]

On Tue, 5 Apr 2016 14:43:25 +1000, Paul Bone <paul at bone.id.au> wrote:
> If the constraint problem is setup with information about the installed
> grades, then the first solution is fine and the solver only needs to return
> one solution.  If instead the solver does not know about the installed
> grades, its first solution may not be one of the installed grades.  It may
> need to return all the solutions.

The question is not "does the solver know the set of installed grades?".
The question is "what task is the solver asked to do?".

The two tasks that the current grade solver can be given are:

(1) best relative match: given a list of the installed grades, return the
grade among them that best matches the given partial grade specification.

(2) best absolute match: return the grade that best match the given
partial grade specification.

The second task is needed for the first, because working on a relative match
*starts* by figuring out the *exact* grade that the name of an installed grade
refers to. For example, given an installed grade named "none.gc", does it
use stack segments?

Returning the list of all matches to a partial grade specification
is *not* a task you can give the current solver.

> The first option seems simpler.

Exactly. If you want to know which installed grade best matches your
partial grade specification, why would you want to do that task *without*
knowing what grades are installed?

Zoltan.