| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
This also adds copyright headers (MIT license) and renames the module.
|
|
|
|
|
|
| |
This adds a few doctests for the algortihm. Sadly a few of the tests are
implementation-specific because for most boards there are a large number
of possible equivalent solutions.
|
|
|
|
|
| |
Scoring only improves the performance of finding the first solution by
about 2%.
|
|
|
| |
This results in the best (shortest) solution to be found first and is much quicker in finding all solutions.
|
| |
|
|
|
|
|
|
| |
This results in many more solutions (15k vs around 60 before) and taking
around a hundred times longer but it does include the shortest solutions
(albeit around the very end of the 15k solutions).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This saves about 5% in the process time, finds a few less solutions and
the average solution length goes down by around 6%.
Still still returns a sub-optiomal solution as the shortest. The reason
is that non-efficient paths to a certain board configuration may be
found first via a long path and will never be re-visited with a shorter
path.
A better way to do this would be to do a breadth-first search instead of
a depth-first one.
|
|
|
|
|
| |
This increases processing time by about 30% but reduces average solution
length by also about 30%.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
This is just the basic implementation of a backtracking search for a
solution and is still lacking the real functionality.
|