This page is outdated, please visit front page www.nebol.se and find your way from there!

Development news & status

2005-07-13: Have made som big improvements in the database cgi programs. The solutions database graphical design is now much nicer and more interesting.

2005-06-28: Deleted solutions from older runs. (run_id < 15).

2005-06-27: Sometimes I could get duplicate entries in the solutions-table because of a thread problem. Gandalf Hernandez helped me fix it.
Implemented better LB, run_id 20.

2005-06-25: Stopped adding undo's to the total number of moves. It was pointless, just made the number unnessecarily large. run_id 19.

2005-06-24: I have improved the LB algorithm, in my head at least, going to implement it and try it out tomorrow.

2005-06-18: finally fixed a good Lower Bound-algorithm for the blocks. run_id 18 running now.

2005-06-12: working on a correct LB routine for the blocks

2005-06-04: run_id 14 was possibly broken, creating too long solutions. Fixed that. Also improved lower bound.

2005-05-30: implemented support for running on multiple computers.

2005-05-28: run_id 9 failed because of a faulty algorithm. Scrapped that idea.
Made a small improvement (on small levels anyway) by not allowing a block to be pushed to a position where it has been been before, unless another block has been pushed in between. run_id 14. Gonna let 14 run all night to see the results on some of the bigger levels.
Also made jeSokobanSolver recognize an unsolvable level and ignore it.
Added Matze's test-levels to the database.

2005-05-24: run_id 9 failed, have to figure out why.

2005-05-21: Fixed a bug in the LB algorithm that caused 1 out of 150 levels to fail. (internal: run_id 8)
Implemented a way of using LB to, when a level fails, make a better estimate of the true LB for the level. It will save a lot of time by eliminating a large number of unnessecary tries. (internal: run_id 9)

2005-05-20: Improved use of LB. (run_id 7). Doubles the time needed for easy levels, but there are huuuge savings on more complex ones.. (as usual! At this stage, I don't bother about optimizing the code, and I don't care much about the time needed. Instead I concentrate on the algorithms, and minimizing the number of moves the program must try. Every move saved means, because of the exponential nature of sokoban, incredible savings later on as the levels grow in size and complexity.

2005-05-19: Implemented better finding of deadlocks and better lower bound (internal: run_id 6)
I chatted a bit with Matthias Meger, the author of JSokoApplet that has a push-oriented solver. He gave me some advice regarding lower bound algorithms.

2005-05-18: Implemented AdjacentToPath-algorithm. (internal: run_id 2)
Implemented basic Static Lower Bound algorithm (mark I). (internal: run_id 3)
Fix for Erim's levels that start already solved. jeSokobanSolver will not try to solve them.
run_id 3 was broken, run_id 4 is working nicely.
Implemented Static DeadEnds algorithm. (internal: run_id 5).
Gaining some speed but still have a long way to go.

2005-05-17: I now have the basic version of the solver up & running for the first time since the rewrite. It is very simple, but the basic algorithm is in place and the machinery is working perfectly. Now I just have to add building block after block, adding algorithms that will improve the solution finding time. And I have to finish some work on the solutions webpages and open them up for the public.

2005-05-11: Haven't done much work yet, but things are coming along. Hope to have some results to show soon.

2004-07-04: NEWSFLASH! I'm scrapping the old solver and starting from scratch on a completely new and improved version! This new version will feature a grid solution, ie multiple solving-clients working together in a network. More info later.

Click here for the information on the old and obsolete version of the software

Page Updated : måndag, april 07, 2008 02:18:54
Created by Johan Eliasson. Copyright © Valheru Productions & Nebol Software 1995 - 2008.
All Rights Reserved.