The New Sudoku Players' Forum

[ronk]

::: comment :::

That's an interesting comment. Why should a rating program "care" whether or not a puzzle is minimal?

it doesn't

champagne's comment -- "to revise my process for non minimal puzzles" -- refers to his other activities, not to the rating software

That would be a logical assumption, but for me that's not the way it reads ... not even after a re-read.

[dobrichev]

Here are 5 puzzles (from my post with wrongly minimized puzzles in Hardest thread)

SE original version rated them

Code: Select all

060000000003070020204300100000031400100400802000200000000000300000090000308100045 9.2/1.2/1.2
060000000003070020204300100002031400100400800000200000000000300000090000308100045 9.2/1.2/1.2
060000000000070020204300100000001400130400802000200003000000300000090000308100045 9.2/1.2/1.2
060000000000070020204300100002001400130400800000200003000000300000090000308100045 9.2/1.2/1.2
060000030000070000204300100000001400130400802000200003000000000000090200308100045 9.2/1.2/1.2


skfr rated them

Code: Select all

.6.........3.7..2.2.43..1......314..1..4..8.2...2...........3......9....3.81...45 ED=9.3/1.2/1.2
.6.........3.7..2.2.43..1....2.314..1..4..8.....2...........3......9....3.81...45 ED=9.2/1.2/1.2
.6...........7..2.2.43..1.......14..13.4..8.2...2....3......3......9....3.81...45 ED=9.3/1.2/1.2
.6...........7..2.2.43..1....2..14..13.4..8.....2....3......3......9....3.81...45 ED=9.2/1.2/1.2
.6.....3.....7....2.43..1.......14..13.4..8.2...2....3.............9.2..3.81...45 ED=9.3/1.2/1.2


Differences are in puzzles 1,3,5. Which result is correct?

Cheers,
MD

[champagne]

Here are 5 puzzles (from my post with wrongly minimized puzzles in Hardest thread)

Differences are in puzzles 1,3,5. Which result is correct?

Cheers,
MD

I'll check what happened here but let me answer in a general way.

After the test run, skfr should very often rate in that area .1 or .2 below SE.
The reason is that SE fails very often in looking for the shortest path toward it's own rule.

I have however the following problem that could lead sometimes to a +0.1 in skfr rating.

In the dynamic process, you have many ways to establish a false for a candidate.
Depending on the order in which you develop the chains, it can be that the best possibility toward SE final count is not seen when I start the reverse search.

The test run in the "dynamic" area has moreover 4 goals:

a) check that the process works properly
b) check that the count is done strictly following SE rules
c) check that lower rating in skfr are justified
d) explain higher rating in skfr

I think a) is ok now up to level 2 and I start test for level 3.
All other items are still requiring comparative tests that need a lot of hand work.
At the end, d) should be the exception.


champagne

[champagne]

May be it's time to tell more about the ongoing process.

I am currently working with the release v 1.2.0 that anybody can download from the google project.

That release does not process the level 4 of serate
and some more validation has to be done at lower levels.

Level 4 will be the last level implemented in skfr.
The double nested level in serate has never been used so far.

The last revision (r148) has the drafted version for all the code (including level 4 of serate), but is in validation testing phase.

In the upper ratings, thanks to mladen dobritchev contribution, the speed ratio compared to serate is somewhere in between 1:100 and 1:1000.

I used intensively skfr in the pattern games 155 and 156 in quite different ways

I played game 155 in a "standard way" and got pretty good results.
For game 156, I did not really play the game, but I looked specifically for high ratings
again, the result was very good and it would not be so difficult, where we are, to win the game with a total of less than 10 submissions.

I think definitely that a player using skfr as a strong advantage.

The first release of could be available soon

champagne

[ronk]

I am currently working with the release v 1.2.0 that anybody can download from the google project.

I tried to find a link to that ... without success.

[dobrichev]

I am currently working with the release v 1.2.0 that anybody can download from the google project.

I tried to find a link to that ... without success.

http://code.google.com/p/skfr-sudoku-fast-rating/

[ronk]

I tried to find a link to that ... without success.

http://code.google.com/p/skfr-sudoku-fast-rating/

I was there earlier. Now see a bunch of source code under "updates", but where's the executable? The logical location, under "downloads", is empty. Logging on to the site made no difference.

[champagne]

I tried to find a link to that ... without success.

http://code.google.com/p/skfr-sudoku-fast-rating/

I see a bunch of source code under "updates", but where's the executable? The logical location, under "downloads", is empty.

This is basically an open source code for a project in development.

The overall organisation follows the rules of such projects controlled by an SVN manager.

The releases are controlled by the "tag" branch that can deliver the sources at the appropriate stage.
If you go in the browser you'll find the V_1_2_0 release source package.

That package is about one month old.
Later revisions are not stable, sometime not valid at all, just check points in the ongoing process.

I guess we can also prepare some equivalent executables files to download.
So far, we only share the source code, each member compiling the program in it's own platform.
If you are only interested in getting the win32 executable file, I'll see with mladen how do do that.

( I can also upload the exec file on my web site, but i think the idea is to keep all that stuff in the google project structure).

champagne

[ronk]

The releases are controlled by the "tag" branch that can deliver the sources at the appropriate stage.
If you go in the browser you'll find the V_1_2_0 release source package.

I feel like I'm pulling hen's teeth here. Does that mean ... one-by-one download the 24 files at this link? Or is there a zip-file somewhere?

[champagne]

The releases are controlled by the "tag" branch that can deliver the sources at the appropriate stage.
If you go in the browser you'll find the V_1_2_0 release source package.

I feel like I'm pulling hen's teeth here. Does that mean ... one-by-one download the 24 files at this link? Or is there a zip-file somewhere?

You got the right link, but it appears clearly through that discussion that the google project organisation require some skills not so common.

Up to now, nobody showed interest in getting the code without being deeply involved in the project. So all the team member learned first to work with a SVN infrastructure.

The right way to download sources in such an environment is to use a SVN manager.I am using tortoiseSVN, a free program.

But I agree that this is more designed for contributors to the project or "active followers".

Let us one day to face that new situation and to do the necessary work to make it easy for visitors.

The minimum would be to have an exec file to download in win32 mode (easy)

It should be also possible to create the zip file corresponding to a release but if anybody wants to go that way, it could be better for him to activate a SVN soft and to be free to extract what he wants. In that case, we should prepare a post explaining the process to apply.

what is your exact need and what is the limit you don't want to pass

champagne

[ronk]

It should be also possible to create the zip file corresponding to a release but if anybody wants to go that way, it could be better for him to activate a SVN soft and to be free to extract what he wants. In that case, we should prepare a post explaining the process to apply.

what is your exact need and what is the limit you don't want to pass

champagne, are those questions for me? If so, it should be obvious by now that I would like something easy to use. Downloading files 1-by-1 or using a SVN utility does not count as easy in my book. A need to compile is already bad enough. BTW if there are compile-time options, will there be a representative makefile showing the proper way to use them?

And why are you yelling ... again?

[champagne]

champagne, are those questions for me? If so, it should be obvious by now that I would like something easy to use. Downloading files 1-by-1 or using a SVN utility does not count as easy in my book. A need to compile is already bad enough. BTW if there are compile-time options, will there be a representative makefile showing the proper way to use them?

And why are you yelling ... again?

let me just evacuate the last sentence.

I put in bold and larger characters the point on which i wanted more precisions, that's it.

We (the team) discussed that issue.

Although we don't want as a general policy to release before we have a stable code, we will make an exception for the executable file corresponding to the tag v_1_2_0 you identified.

That exec is the program I have in use.

We (the team) see the current code as a construction project, not yet releasable.
As we work openly, anybody can see it, study it .....
Anybody willing to contribute is welcome in the team.

you'll find, loaded by mladen the zip file you are waiting for.

that file contains

- the program, a win32 console executable
- an important readme file with a summary of the command line working for that program.

Any lack of documentation, comment on errors, discussion on how to get the maximum from that program...
can be discussed here;

the input file is very similar to serate's one, but the output keep only the 81 first positions as you will see.

To anybody willing to do more:

the projects files include projects definitions (in microsoft VC++ mode) to buil a dll, that executable ...



champagne


PS for new visitors, here is the link to the downlooads of the projects

code.google.com/p/skfr-sudoku-fast-rating/downloads/list

[eleven]

SKFR is written in C++ and doesn't use any specific API from Windows.

This not quite true, some string and time functions are windows specific as well as the union definition of t_128. But after 2 hours i could compile it under linux (apart from the dll of course) and it seems to work.

So some "#ifdef WINDOWS ... else #ifdef UNIX" lines would be nice, as well as a Makefile for linux/unix, in the stable version then. (Please dont ask me for that, i never show my Makefiles to others ).

[champagne]

SKFR is written in C++ and doesn't use any specific API from Windows.

This not quite true, some string and time functions are windows specific as well as the union definition of t_128. But after 2 hours i could compile it under linux (apart from the dll of course) and it seems to work.

So some "#ifdef WINDOWS ... else #ifdef UNIX" lines would be nice, as well as a Makefile for linux/unix, in the stable version then. (Please dont ask me for that, i never show my Makefiles to others ).

Hi eleven,

I suspect that what you call "Windows specific" is more "microsoft C++ specific"

Could you give a list of instructions you had to change

In principle, I see no problem

. either to try not to use such instructions (but microsoft C++ does not like some instructions I used before with the Borland IDE")
. or to include in the code some specific lines for linux if
this does no disturb compilation
it is not spread over the code

I guess the makefile is in microsoft IDE automatically produced out of the project definition
the project definitions are included in skfr

champagne

[dobrichev]

It is fact that the current code is not portable. The compiler/OS specifics are not so much, they are isolated, and it is matter of few hours to port the code to other modern compiler.

Is this a priority now? I guess no.