The New Sudoku Players' Forum

[ronk]

Off-topic

you must be aware of the fact that Ron has a very small screen.

That's because I have very small eyes. Seriously, I use a 1024x768 pixel setting and "Medium" text size in MS Internet Explorer. Is that really considered "small?"?

[fermat]

Off-topic

you must be aware of the fact that Ron has a very small screen.

That's because I have very small eyes. Seriously, I use a 1024x768 pixel setting and "Medium" text size in MS Internet Explorer. Is that really considered "small?"?

I am ditto, mine did not wrap, although the post was wider than I usually see for "code".

[Pat]

you must be aware that ronk has a very small screen.

I use 1024×768 pixels and "Medium" text-size in MS Internet Explorer.
Is that really considered "small?"?

i suggest you try SuDoku.com/forums
( i.e. in a separate window rather than part of SuDoku.com
— saves you the wasted space at left for the navigation-bar )

Pat

[gsf]

Was curious about the pencilmark grids (which techniques were applied for eliminations, when are the grids printed, etc), so I checked a few details. First: seems that you have a way of specifying "print the pm-grids before an x-wing is applied", or "every crucial step", or "every step", which is a nice option.

the selection can happen 2 ways

the first way uses options to restrict the constraint methods applied
along with an option that prints (via a separate printf-like format option) just before
a guess is required (i.e., no active constraint method advances the solution)
I used this for the x-wing collection

the second way adds a numeric label to the method step trace:
run once listing the labels, then run again with an option to print and
exit when the labelled step is hit -- this is handy for mining individual puzzles

Second, for choice of elimination techniques, I checked especially #6: Here are six possible eliminations by naked pairs (@34), but these are not sufficient (and not necessary). The x-wing can be applied directly, the np-eleminations are not needed.

although my solver allows individual constraint methods to be enabled/disabled (except naked singles),
the application order is fixed from a subjective easiest to hardest:

naked/hidden singles
box claims (locked candidates)
naked/hidden n-tuple, order-n xwing, 2<=n<=4 (order n checked before order n+1)
an in-progress-on-the-weekends-since-december cycle based method

the last is close to being a simple generalization of the non-guessing non-uniqueness methods
just need a few more weekends

back to your second question, this command

Code: Select all

sudoku -qFNW -e G ocean-xwing.dat

listed no puzzles, so all of your examples solve with just naked/hidden singles and xwings

each method is identified by a single letter and an order (e.g., W3==swordfish, G==guess)
this command shows that, using ony naked/hidden singles and order 2 xwing,
only puzzle 4 requires more than one xwing

Code: Select all

sudoku -S -qFNW2 -e 'W>1' -f'puzzle %n xwings %(W)x' ocean-xwing.dat
output: puzzle 4 xwings 3

where -S stops the current constraint iteration and rewinds to the easiest
when any constraint advances the solution
-S is still subject to algorithm implementation bias (row/col order, try 1 .. 9, etc.)
so its possible that applying the xwings in a different order could result
in a different number of required xwings

(updated post of the solver with these newer options real soon now)

[ronk]

Off-topic

i suggest you try SuDoku.com/forums
( i.e. in a separate window rather than part of SuDoku.com
— saves you the wasted space at left for the navigation-bar )

Thanks, but I haven't seen that nav bar for quite a while.

I am ditto [edit: using the same settings], mine did not wrap

Strange, as I would have thought the view had to be the same.

And I can hear Ocean thinking ... "enough of the off-topic already.":)

[Ocean]

the selection can happen 2 ways (...)

Thanks for a thorough explanation. Your options seem to be quite flexible. Useful for detailed analysis of how selected puzzles can be solved.

[Ocean]

[#3:]
Here is my contribution to "interesting..."
... "Continuous" Almost Locked Sets XY

... candidate grids are really hard to read when lines wrap.

I have a hard time reading them even when lines don't wrap. (Twelve cells marked up with five letters a-b-c-d-e: difficult to see the logic even when told. Hopefully I learn it some day...) But for a solver program it's certainly an "interesting" test case. Maybe I should try to implement ALS... Anyway, thanks for pointing at the possibility, Havard.

I would call the xy-chain pointed at by tso easier to spot:
2-[r1c4]-1-[r1c2]-9-[r9c2]-1-[r9c7]-8-[r7c9]-1-[r2c9]-2 =>r1c7<>2. Before and after it's singles only.

[Havard]

[#3:]
Here is my contribution to "interesting..."
... "Continuous" Almost Locked Sets XY

... candidate grids are really hard to read when lines wrap.

I have a hard time reading them even when lines don't wrap. (Twelve cells marked up with five letters a-b-c-d-e: difficult to see the logic even when told. Hopefully I learn it some day...) But for a solver program it's certainly an "interesting" test case. Maybe I should try to implement ALS... Anyway, thanks for pointing at the possibility, Havard.

I would call the xy-chain pointed at by tso easier to spot:
2-[r1c4]-1-[r1c2]-9-[r9c2]-1-[r9c7]-8-[r7c9]-1-[r2c9]-2 =>r1c7<>2. Before and after it's singles only.

Hi! It is actually not that hard when you know how to interpret it! Each set (ALS) is marked with a letter. So if you have more than one "a", they all belong to the same set. As you know, n cells with a total of n+1 different candidates makes out an ALS.
Then you have to find the "X" that links each set. By "X" I mean a candidate that exists in both sets (say, a and b or b and c), and make sure that ALL occurences of that canididate in both sets "see" eachother! Or said differently: if occurence of that candidate in either set was TRUE, then all the other occurences would be FALSE... Then you can just work your way from set to set, noting what the "x" have been. Note you can not have the same "x" twice in a row... Finally you will see that the last set connects back to the first one, and when this happens you can eliminate:
* all candidates that see ALL the "x" candidates in the sets
* all candidates that can see all occurences of ANY of the other candidates(not X's) in any set...

Hope that helped a bit!

Havard

[ronk]

Finally you will see that the last set connects back to the first one, and when this happens you can eliminate:
* all candidates that see ALL the "x" candidates in the sets
* all candidates that can see all occurences of ANY of the other candidates(not X's) in any set...

Hope that helped a bit!

Brief notation in the original posting would've helped as much.

Code: Select all

5      19#c    169   | 12      8       7       | 126#a#e   4     3
123#c  7       13#c  | 9       4       6       | 5         8     12#d
126#c  8       4     | 1235    125     123     | 7         9     126#d
---------------------+-------------------------+------------------------
369    2       3679  | 34568   5679    3489    | 468#a#e   1     689
8      4       1369  | 1236    1269    1239    | 26#a#e    7     5
169    5       1679  | 12468   12679   12489   | 2468#a#e  3     2689-2
---------------------+-------------------------+------------------------
7      3       2     | 1468    16      148     | 9         5     18
19     6       8     | 7       19      5       | 3         2     4
4      19#b    5     | 128-1   3       1289-1  | 18#b      6     7

"Continuous" Almost Locked Sets XY

Maybe something along the lines of ...

- a(1|246|8) - b(8|1|9) - c(9|123|6) - d(6|2|1) -

Then one can easily see the links between sets are digits 8, 9, 6, and 1 for set pairs a-b, b-c, c-d, and d-a, respectively.

And it's also easy to see, when the loop is continuous, that digits 246, 1, 123, and 2 are locked in sets a, b, c, and d, respectively. (Don't forget the exclusion r1c3<>1.)

[Ocean]

[...] Finally you will see that the last set connects back to the first one, and when this happens you can eliminate:
* all candidates that see ALL the "x" candidates in the sets
* all candidates that can see all occurences of ANY of the other candidates(not X's) in any set...

Brief notation [...] Maybe something along the lines of ...

- a(1|246|8) - b(8|1|9) - c(9|123|6) - d(6|2|1) -

Then one can easily see the links between sets are digits 8, 9, 6, and 1 for set pairs a-b, b-c, c-d, and d-a, respectively.

And it's also easy to see, when the loop is continuous, that digits 246, 1, 123, and 2 are locked in sets a, b, c, and d, respectively. (Don't forget the exclusion r1c3<>1.)

Håvard and Ron: With your explanations combined, I finally see the whole picture. Thanks!
(Also observing: "Continuous" Almost Locked Sets XY degenerate to simple xy-chains if every set consists of one cell.)

[Havard]

Maybe something along the lines of ...

- a(1|246|8) - b(8|1|9) - c(9|123|6) - d(6|2|1) -

Then one can easily see the links between sets are digits 8, 9, 6, and 1 for set pairs a-b, b-c, c-d, and d-a, respectively.

And it's also easy to see, when the loop is continuous, that digits 246, 1, 123, and 2 are locked in sets a, b, c, and d, respectively. (Don't forget the exclusion r1c3<>1.)

mmm... Me like! Makes for a very clear presentation of what is going on... Also, what about:

a(246)-8-b(1)-9-c(123)-6-d(2)-1-a

Here you don't have to present the "x" twice, and that might be clearer once you remove the colors (the colors make it very very clear, but they are not always available)

Thoughts?

Havard

[Havard]

I know we are drifting off topic here again, but here is another fun one for a bit of ALS excersise:

Code: Select all

Almost Locked Sets XY rule:
9     135-1 7     | 6     18#a  358   | 2     38    4
45#h  8     34#f  | 2     35#g  7     | 1     9     6
6     13#r  2     | 138-1 49#b  49    | 5     38    7
------------------+-------------------+------------------
14#i  7     89    | 358   6     348   | 39    25#k  12#j
25#o  25#p  34#e  | 9     47#c  1     | 37#d  6     8
13#m  6     89    | 578   2     358   | 379   4     15#l
------------------+-------------------+------------------
78    4     5     | 178   89#a  29    | 6     12    3
23#n  9     1     | 4     35    6     | 8     7     25
78    23#q  6     | 1357  178   2358  | 4     125   9


The challenge is to write up the "chain" for this one, pointing out what the "x" is between each set!

Havard

[ronk]

- a(1|246|8) - b(8|1|9) - c(9|123|6) - d(6|2|1) -

Then one can easily see the links between sets are digits 8, 9, 6, and 1 for set pairs a-b, b-c, c-d, and d-a, respectively.

And it's also easy to see, when the loop is continuous, that digits 246, 1, 123, and 2 are locked in sets a, b, c, and d, respectively.

... VERSUS ...

Also, what about:

a(246)-8-b(1)-9-c(123)-6-d(2)-1-a

Here you don't have to present the "x" twice ...

I like both styles, the first because all set members are within the parentheses, and the second because it's more in keeping with nice loop notation. I'll probably try both for a while and decide later ... after seeing some cases where naked N-tuples are "created" within the ALS. (And there's also Myth Jellies' notation here to learn yet.)

... and that might be clearer once you remove the colors (the colors make it very very clear, but they are not always available) (...) Thoughts?

The colors were one-time highlighting, not a suggestion that colors be part of any "standard" notation. It's waaay too much work.

[Pat]

filtered ~2M generated symmetric puzzles from last March --

[ 28 clues ]

Code: Select all

 . 4 7 | 3 . . | 8 . .
 . . . | . 1 . | . 5 .
 . 1 3 | 2 . 8 | . . .
-------+-------+------
 . . . | 1 . . | . . 5
 . . 1 | 5 . 2 | 6 . .
 7 . . | . . 3 | . . .
-------+-------+------
 . . . | 4 . 5 | 1 6 .
 . 9 . | . 6 . | . . .
 . . 4 | . . 1 | 2 7 .



[Pat]

? have you seen
Extreme Su Doku
by Wayne Gould

here's a nice example --

[ 28 clues ]

Code: Select all

 . 9 . | . . . | . 6 .
 5 . . | . 2 . | . . 4
 . . 1 | 7 . 4 | 9 . .
-------+-------+------
 9 . . | . 5 . | . . 3
 . 5 . | 4 . 2 | . 9 .
 4 . . | . 1 . | . . 8
-------+-------+------
 . . 6 | 2 . 5 | 3 . .
 3 . . | . 9 . | . . 6
 . 7 . | . . . | . 5 .