Sudoku solver for humans / News: Recent posts

-> user interface
-> graphics with sdl
-> translation is possible as well for many expressions
-> only available for linux users by now

* framework improved (initialization, memory manager, signals, scripting (cvars, cmds, completion), files management system)
* -o/-w/-W parameters for yass.
* bugs fixed
* yass.script

* framework improved.
* The size of the program has decreased. (The buffer size is much smaller by default, and it's now a double linked list).
* new stats field about memory, PLUS it's now displayed with a suitable unit.
* bug fixed in memory management when freeing null pointers.
* bug fixed when displaying forcing chain.

-> framework updated.
-> few fixes.
-> minor changes due to sudokuLib.Copy2, ListAllocate, CandStatsDuplicate...

-> Sudokus are managed in a more efficient way.
-> framework: no buffer overrun.

Oops, 2 bugs fix... Sorry. :)

* Nishio and Guessing run much faster, and less memory consuming.
* statistics are displayed on stdout (memory stats).
* New sudoku procedures: CutAndPasteLog, Copy, Reset.
* sudokutries abandonned. New: tries.c/h.
* Every calculus about columns, rows... given the offset are hardcoded.
* The difficulty calculation procedure has changed. Wide range now.

bugs fixed in displaying methods.
bug fixed when managing sudoku string.
memory management improved, so as the sudoku check-up.

* Fix in displaying method, the last node was left out.
* Fix in displaying help.
* New memory allocator. Much time saved. (no smp)
* New review node management. Much clearer and effective.
* new parameter: "-m" to display memory consumption.
* sniffing procedures to debug.
* major changes in region and cell structures.
* the review log is displayed differently, seems clearer to me.
* new functions to handle strings (for windows for sure!)

many improvements:
* time saved due to stack management rewritten from scratch; findcombos.c, PRECALCULATE_ARRAYS, part of candidatesStats structure hardcoded...
* timer to display elapsed time to solve the grid.
But now, SDL lib's needed.
SVN repository updated (trunk and MemCheck branch).
For windows users, there may be something annoying, the standard output is a file due to sdl.dll (version 1.2.12). So, to watch at the results, just think to have a look at stdout.txt

bugs fixed in candidates management.

few improvements regarding the region management...

trunk directory's now up-to-date

changes in sudoku structure and in the source code now cleaner.

version 0.41...

Yass is now its name. It can be found here:
http://sourceforge.net/project/showfiles.php?group_id=182924&package_id=231600&release_id=507788

many changes such as: a new structure to tidy data up; bugs fixed in naked, hidden multiples and forcing chain; many English comments added; the tries of all cell candidates are done in a more efficient way; and [-V] parameter which shows the option used for compiling.

-> smp : bugs fixed for forcing chain which is now smp.
-> smp version limited to two threads at a time, but can be easily modified (NB_MAX_THREADS)
-> sudokutries.c/listofint.c : many changes and bug(s) fixed.
-> source code easier to read. (typedefs specified)
-> perfo slightly on the rise.

-> sudoku parameter can be either "1000070900300[...]" or "1....7.9..3..[...]".
-> help and README updated.
-> a try of SMP version for linux only. ('pthread' must be installed). BUT not very happy about it.
-> few bugs fixed...
-> sudokutries.h shared by nishio.c/guessing.c/forcingchain.c (can be SMP or not).

-> a "vicious" memory leak now fixed.
-> singlecandidate has been split into two methods, one really easy (just to fill the last candidate in a region), and the other one, harder (the former one in fact). The difficulty is bound to have changed.
-> Much time saved on very hard sudokus. Nishio and Guessing now detect impossible sudokus. sudokusolver updated consequently to handle that.

-> the order in parameters doesn't matter anymore (caution, a single sudoku at a time).
-> more hints in Nishio & Guessing.
-> piValueStats, allocation on demand (SudokuGetAllRemainingCandidates updated)

-> "Single Candidate" has now hints.
-> Memory allocation for stacks on demand.
-> ReviewLinkCreate does everything, no other procedure. Simpler that way.

Especially if something turned out to be a bug, or if something started to run out.

-> Oops. Sorry for the inconvenience. A memory leak has been spoted in SudokuAddLinkToQueue.

-> Bug fixed in forcingchain, nishio, guessing. Creation of sudoku structure now simplified.
-> Much time saved after having written from scratch ExtractSmallestBit and ExtractBiggestBit.
-> Bug fixed in DeleteCandidateFromMarkedCellsAndAddToReview
-> solver.c/h has become sudokusolve.c/h