YZF_Sudoku

Programs which generate, solve, and analyze Sudoku puzzles

Re: YZF_Sudoku

Postby denis_berthier » Tue Dec 15, 2020 3:33 pm

SpAce wrote:z-chains are nothing but forcing nets

So, now they are not only "mapped to", but they "are nothing but"....
Still more stupid.
You should read my post here: pattern-based-constraint-satisfaction-2nd-edition-t32567-7.html
denis_berthier
2010 Supporter
 
Posts: 2767
Joined: 19 June 2007
Location: Paris

Re: YZF_Sudoku

Postby StrmCkr » Wed Dec 16, 2020 12:23 am

how about we redirect this solvers thread back to its intended purpose instead of cluttering yzf's good work up?
- thanks
Some do, some teach, the rest look it up.
User avatar
StrmCkr
 
Posts: 1213
Joined: 05 September 2006

Re: YZF_Sudoku

Postby SpAce » Wed Dec 16, 2020 2:56 am

StrmCkr wrote:how about we redirect this solvers thread back to its intended purpose instead of cluttering yzf's good work up?

I fully agree. The only possible relevance of this side track would be if yzfwsf were somehow interested in providing a mapping of his techniques to Denis' names and possibly notations. In most cases it's straight-forward, and I could help with that.

To me the most useful feature of Denis' system is the fine-grained classification of various types of nets, which is clearly missing on the other side (even though the patterns themselves obviously aren't). The whip/braid distinction, the z-, t-, and zt-prefixes, the g- and S- extensions, and the unambiguous notion of length, are all useful differentiators. They allow describing common network patterns much more accurately and uniformly than anything else I've seen, and those classifications are valid regardless of any notations, theories, and finding methods being used.

In an ideal world Denis would decouple those terms from his notation and other baggage that comes attached. That would make them much more useful, and probably more popular too. I could still write my beloved AICs and krakens etc any way I like, but I could also describe their underlying patterns and complexities in comparable terms using Denis' classification. That'd be a win-win for me.

Of course that won't happen because Denis obviously can't stand if people even look at his chains and terms from an unapproved perspective, much less use them in a wrong context. Buying one part implies buying the whole system, which is one reason why few people are willing to make that investment. Personally I like modular designs that allow picking and combining the best parts from different systems. (I make an exception with Apple, though.)
User avatar
SpAce
 
Posts: 2671
Joined: 22 May 2017

Re: YZF_Sudoku

Postby yzfwsf » Thu Dec 17, 2020 3:29 am

Fix a bug in GSP sticks symmetry type 1 and add support for sticks symmetry type 2. In order to facilitate players to learn GSP and add puzzle morph function, you can use this function in the tool menu.

BTW:
After the puzzle is position morph, you only need to re-mark the numbers to recover, which is suitable for Gurth's placement technique. If the position of some units can remain unchanged during the position transition, the effect of using this technique will be more obvious.
sticks symmetry type 1:permutation rows to 123789456 and cols to 132465798
Code: Select all
00   01   02   03   04   05   06   07   08            00   02   01   03   05   04   06   08   07   
09   10   11   12   13   14   15   16   17            09   11   10   12   14   13   15   17   16
18   19   20   21   22   23   24   25   26            18   20   19   21   23   22   24   26   25
27   28   29   30   31   32   33   34   35            54   56   55   57   59   58   60   62   61
36   37   38   39   40   41   42   43   44            63   65   64   66   68   67   69   71   70
45   46   47   48   49   50   51   52   53      ==>   72   74   73   75   77   76   78   80   79
54   55   56   57   58   59   60   61   62            27   29   28   30   32   31   33   35   34
63   64   65   66   67   68   69   70   71            36   38   37   39   41   40   42   44   43
72   73   74   75   76   77   78   79   80            45   47   46   48   50   49   51   53   52


sticks symmetry type 2:permutation rows to 32146798 and cols to 897456312
Code: Select all
00   01   02   03   04   05   06   07   08            25   26   24   21   22   23   20   18   19
09   10   11   12   13   14   15   16   17            16   17   15   12   13   14   11   09   10
18   19   20   21   22   23   24   25   26            07   08   06   02   04   05   00   01   03
27   28   29   30   31   32   33   34   35            34   35   33   30   31   32   29   27   28
36   37   38   39   40   41   42   43   44            43   44   42   39   40   41   38   36   37
45   46   47   48   49   50   51   52   53      ==>   52   53   51   48   49   50   47   45   46
54   55   56   57   58   59   60   61   62            61   62   60   57   58   59   56   54   55
63   64   65   66   67   68   69   70   71            79   80   78   75   76   77   74   72   73
72   73   74   75   76   77   78   79   80            70   71   69   66   67   68   65   63   64

Other variants of GSP are well known to players.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby ghfick » Thu Dec 17, 2020 10:15 pm

The morphing tool is a valuable addition for human solvers and it is easy to use.
btw I notice in the 'Color Cell' tab, a button called 'Chain'. Can you explain how to use this Chain button? If you can draw in some chains manually, that would be an interesting add for human solvers as well.
ghfick
 
Posts: 149
Joined: 06 April 2016

Re: YZF_Sudoku

Postby yzfwsf » Sat Dec 19, 2020 9:47 am

release 618v4; fix bug in custom puzzle morph
ghfick wrote:btw I notice in the 'Color Cell' tab, a button called 'Chain'. Can you explain how to use this Chain button? If you can draw in some chains manually, that would be an interesting add for human solvers as well.

In fact, I haven’t finished this function yet. I am hesitant to add this function and how to implement this function.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby yzfwsf » Sat Dec 26, 2020 1:40 pm

Release version 619:
Add GSP puzzle generation function, add hand animation chain function. Press the left or right mouse button on the candidate number to determine the starting point, and press the left or right mouse button on the other candidates again to confirm the end point, but pressing the left button will draw a solid line, otherwise draw a dotted line.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby jco » Sat Dec 26, 2020 9:25 pm

Hello, yzfwsf,

yzfwsf wrote:Release version 619:
Add GSP puzzle generation function, add hand animation chain function. Press the left or right mouse button on the candidate number to determine the starting point, and press the left or right mouse button on the other candidates again to confirm the end point, but pressing the left button will draw a solid line, otherwise draw a dotted line.


Thank you for providing this new feature for drawing chains.
This feature together with the available colouring tool will be very helpful for manual solvers.

Best regards,
jco.
JCO
jco
 
Posts: 311
Joined: 09 June 2020

Re: YZF_Sudoku

Postby ghfick » Sat Jan 09, 2021 6:53 pm

I am using the new chain tool. Looks very fine.
Is there a way to back up just one drawn line? You can remove the entire displayed chain but going back one step would be helpful.
Is there a way to copy the entire displayed solution path? Maybe this would be OS specific. Not sure.
Sometimes, with a JE2 step, the compatibility check excludes a digit that is not in the JE band and not in the 'S' cells. The JE2 in U434 is an example. Such exclusions are often seen with JE4 but I am unsure how they are seen with JE2.
ghfick
 
Posts: 149
Joined: 06 April 2016

Re: YZF_Sudoku

Postby yzfwsf » Sat Jan 09, 2021 11:32 pm

ghfick wrote:I am using the new chain tool. Looks very fine.
Is there a way to back up just one drawn line? You can remove the entire displayed chain but going back one step would be helpful.
Is there a way to copy the entire displayed solution path? Maybe this would be OS specific. Not sure.
Sometimes, with a JE2 step, the compatibility check excludes a digit that is not in the JE band and not in the 'S' cells. The JE2 in U434 is an example. Such exclusions are often seen with JE4 but I am unsure how they are seen with JE2.

The method of erase one line is same to draw line.

Compatibility checking is the operation of enumerating pairs of true base digit.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby ghfick » Sun Jan 10, 2021 12:52 am

Please expand your explanations.
Why would a compatibility check exclude a digit that is in neither the JE band nor the 'S' cells?
ghfick
 
Posts: 149
Joined: 06 April 2016

Re: YZF_Sudoku

Postby yzfwsf » Sun Jan 10, 2021 1:21 am

Enumerate base pairs and apply JE rules to find out all reasonable number pairs. If these reasonable number pairs can delete a certain number, then it will become the final elimination.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby ghfick » Sat Jan 23, 2021 5:47 pm

A few minor edits and some possible next steps :
1) Death Blossom -5r1c1 should be r1c1 <>5
2) AIC _ Type _ 1 _ add spaces between Type, 1 and the start of the chain
3) AIC _ Type _ 2 _ add spaces between Type, 2 and the start of the chain
4) I am unsure how chains are currently sorted for display in the default solution path. XY Chains are the easiest. Type I may give more exclusions than Type II but this is not clear.
Some might say the chain length comes next. I could argue for Type I's listed in length order and then Type IIs listed in length order.

Your solver now has very thorough Exocets that are Junior or Senior. Next would be Exocets that are neither Junior nor Senior. These Exocets could be checked [ as champagne notes ] by a digit by digit check. A contradiction approach looks like the best way. Remove the digit from the Target cells and then show there is no valid solution. To be instructive for human solvers, the explanation window would need to contain all of the checking needed. It might be rather lengthy but no matter. Sudoku Explainer often has really long explanations.
ghfick
 
Posts: 149
Joined: 06 April 2016

Re: YZF_Sudoku

Postby yzfwsf » Tue Jan 26, 2021 3:01 am

ghfick wrote:A few minor edits and some possible next steps :
1) Death Blossom -5r1c1 should be r1c1 <>5
2) AIC _ Type _ 1 _ add spaces between Type, 1 and the start of the chain
3) AIC _ Type _ 2 _ add spaces between Type, 2 and the start of the chain
4) I am unsure how chains are currently sorted for display in the default solution path. XY Chains are the easiest. Type I may give more exclusions than Type II but this is not clear.
Some might say the chain length comes next. I could argue for Type I's listed in length order and then Type IIs listed in length order.

Your solver now has very thorough Exocets that are Junior or Senior. Next would be Exocets that are neither Junior nor Senior. These Exocets could be checked [ as champagne notes ] by a digit by digit check. A contradiction approach looks like the best way. Remove the digit from the Target cells and then show there is no valid solution. To be instructive for human solvers, the explanation window would need to contain all of the checking needed. It might be rather lengthy but no matter. Sudoku Explainer often has really long explanations.

In fact, the solver searches the chain sequentially from cells 0 to 80, with candidate numbers 1 to 9 as the starting point. If it finds the output immediately, without finding all the chains, then find the shortest one.
BTW: I think for the manual solver that uses the coloring method to find the chain, the chain length should not be a troublesome thing.

For Senior Exocet, the deep check cross-line will output the cover house for each base digit.

Now when constructing custom chains, it is more convenient to color the candidates of node at the same time.

Now it supports the Grouped AIC.

Version 620 release.
yzfwsf
 
Posts: 422
Joined: 16 April 2019

Re: YZF_Sudoku

Postby ghfick » Tue Jan 26, 2021 7:17 pm

Hi yzfwsf,

Excellent to have Grouped AICs added.

Have a look at :

.1.5...9.2...3...5..9.......4.6....8.3.....1...1.8.76....2....16....7.2..8..6.4..

If you move to just before the first Grouped AIC Type I and then 'Show Next Step' , you get a step labelled ALS but the visual is not correct [showing part of the Grouped AIC?] . 'Show Next Step' should show the first Grouped AIC as listed in the solution path.

I should have more for you later today or tomorrow.

Gordon
ghfick
 
Posts: 149
Joined: 06 April 2016

PreviousNext

Return to Software