Thanks, JC! That was very instructive again!
JC Van Hay wrote:In general :
1. The less the number of native constraints in a step, the best.
I think I see what you mean but I also think "the best" is somewhat subjective. I personally prefer clarity over brevity. I do agree, however, that my first chain may be unnecessarily long, which is also bad for clarity. I just wrote it the way I found it without too much thinking. Its one benefit is that its logic should be pretty easy to follow even for an ALS-novice (because one wrote it). Still, I'd probably do it this way now (because the elimination has the same effect with a shorter chain):
(1=4=5)r46c8-(5=6)r9c9-(6=1)r9c2 => -1 r6c2
Of course it can also be seen as an ALS-XZ (or several) but I'm not very good at thinking in those terms. A chain is a chain, and it's much easier for me to see it as such instead of thinking of patterns and restricted common candidates and so on. But, if I do choose to view it as an ALS-XZ (or simply as a chain of two ALSs, which is more natural for me) isn't there at least two ways we can do it:
1) ANS-1: r46c8{145} and ANS-2: r9c28{156}, where the RCC is 5 and the chain is: (1=4=5)r46c8-(5=6=1)r9c82 (like you wrote it).
2) ANS-1: r469c8{1456} and ANS-2: r9c2{16}, where the RCC is 6 and the chain is: (1=45=6)r469c8-(6=1)r9c2
Have I understood that correctly? The problem with many ALS nodes and especially big ones is that it gets harder and harder to see what's going on just by reading the chain without seeing the grid (and even then it's sometimes hard). Simpler chains can be followed more easily in one's head, even if they have more nodes, because one can see the effects on cell level. That's why I tend to prefer chains with as few and small ALS nodes as possible, even if they're a bit longer. I bet they're more intuitive for a novice than ALS patterns anyway, but it may just be my own bias.
I haven't yet digested what you said about step 2.
