| KDE Games |
Support for solving Golf using bh-solver.
See https://github.com/shlomif/black-hole-solitaire and
https://www.shlomifish.org/open-source/projects/black-hole-solitaire-solver/
.
This is a preliminary version but seems to work. There is an
optional build dep with a fallback on the existing algo.
test that the golf variant is working as before and the new solver
is working.
| Automatic diff as part of commit; lint not applicable. |
| Automatic diff as part of commit; unit tests not applicable. |
Fixes and cleanups.
@aacid : for deal no. 5 (starting position) the old solver reports that "This game cannot be won", while my solver solves it quickly and to completion. I can try collecting statistics on a range of deals wrt the performance of the two solvers.
Hi!
I found out that kpat has a command line range solving mode. Here are the results of the first 5,000 deals:
Old New Count --- --- ----- lost lost 2291 lost unknown 462 lost won 930 unknown lost 9 unknown unknown 56 unknown won 38 won unknown 13 won won 1201
They were generated using:
#! /bin/bash # # analyze.bash # Copyright (C) 2019 Shlomi Fish <shlomif@cpan.org> # # Distributed under terms of the MIT license. # filt2() { grep -E '^[1-9]' | head -5000 } filt() { filt2 | perl -lanE 'say $F[1]' } echo $'Old\t\tNew\t\tCount' echo $'Old\t\tNew\t\tCount' | perl -lpE 's/\w/-/g' paste <(< old-golfs.txt filt) <(< new-golfs.txt filt) | sort | uniq -c | perl -lanE 'say "$F[1]\t\t$F[2]\t\t$F[0]"'
The new solver (bh-solver) solves more deals including ones that were reported as impossible. It seems to be slower in bulk mode though, but it is fast enough for solving a single deal.
For the record, the kpat cmd line I used was ./kpat --start 1 --end 1000000 --solve 12 2>&1 | tee -a new-golfs.txt
Better formatted table:
Old New Count --- --- ----- lost lost 2291 lost unknown 462 lost won 930 unknown lost 9 unknown unknown 56 unknown won 38 won unknown 13 won won 1201
and the new script:
#! /bin/bash # # analyze.bash # Copyright (C) 2019 Shlomi Fish <shlomif@cpan.org> # # Distributed under terms of the MIT license. # filt2() { grep -E '^[1-9]' | head -5000 } filt() { filt2 | perl -lanE 'say $F[1]' } head_() { echo $'Old\tNew\tCount' } ( head_ head_ | perl -lpE 's/\w/-/g' paste <(< old-golfs.txt filt) <(< new-golfs.txt filt) | sort | uniq -c | perl -lanE 'say "$F[1]\t$F[2]\t$F[0]"' ) | perl -lanE 'printf"%-20s%-20s%s\n",@F'
First 20,000 deals:
Old New Count --- --- ----- lost lost 9038 lost unknown 1971 lost won 3674 unknown lost 47 unknown unknown 254 unknown won 149 won unknown 56 won won 4811
Those numbers look good.
My only concern is that when we get bugs for "solver was wrong" now we have to magically guess which solver was used, but oh well.
I don't really think we have a maintainre for kpat, so i guess you can go ahead commit it and become the maintainer :D