marek stefanik wrote:There are no CSP variables for uniqueness contradictions, which makes it impossible to write it as a whip.
However, much like in a whip[1].
Whips are not a hotpot where anything can be put in. Thats' what allows to develop some theory about them. The obvious counterpart is, not everything can be written as a whip. So, you're perfectly right that this cannot be written as a whip.
I've defined some extension of whips such as S-whips and W-whips (accepting resp. Subsets and other whips as inner right-linking objects), but I've never defined (and will never define) anything including uniqueness patterns. In my view, uniqueness has to be proven along resolution. I know this is a personal choice that no one has to accept.
marek stefanik wrote:it's a candidate causing an immediate conflict, ... I said what candidate it was and then described the contradiction.
I don't know what else I should have done...
You first said "otherwise 35 would be unresolvable in the rest of the grid". At that point, it was only T&E(XX).
Then, in a new post, you added "It looks like an extended UR"- which I didn't comment, because I know of no name for this extended pattern.
[Edit]: For fun, starting from the RS after Singles and whips[1]
- Code: Select all
Resolution state after Singles and whips[1]:
+----------------+----------------+----------------+
! 5 7 6 ! 8 2 1 ! 9 3 4 !
! 9 1 2 ! 7 3 4 ! 8 56 56 !
! 3 8 4 ! 69 5 69 ! 2 7 1 !
+----------------+----------------+----------------+
! 8 356 9 ! 356 4 2 ! 7 1 356 !
! 4 356 7 ! 1 8 356 ! 35 2 9 !
! 2 356 1 ! 3569 7 3569 ! 4 56 8 !
+----------------+----------------+----------------+
! 6 4 8 ! 35 1 7 ! 35 9 2 !
! 1 2 35 ! 4 9 35 ! 6 8 7 !
! 7 9 35 ! 2 6 8 ! 1 4 35 !
+----------------+----------------+----------------+
I tried if n5r2c8 could be eliminated directly* by any of the chain rules in CSP-Rules (whips, braids, g-whips, g-braids...). The answer is no.
(*) this can be done with command : (try-to-eliminate-candidates (nrc-to-label 5 2 8))