denis_berthier wrote:blue wrote:But the reverse isn't always the case: that the extension of a template[k-1] by a compatible template[1] for a different number, is (still) a member of the relevant template[k] list. That was the point. I think you're being argumentative for no good reason.
I don't understand what you mean by "relevant list". If you mean that they are compatible with the grid data (i.e. the current resolution state), all the template[k] instantiations I ever consider are a priori compatible with the grid data - until they are eventually proven to be impossible.
For an example, if the template[k-1] item is in the list for digits {1,2}, and the "compatable template[1] item" is in the list for {9}, then the "relevant list" for the extension, is just the template[3] list for digits {1,2,9}.
denis_berthier wrote:blue wrote:"Too many eliminations" ... you would think so, but the "no solution" problem hasn't come up, even for your list of 100 T3 puzzles.
So, considering our common results below, this suggests:
- either you have a bug and it is localised in your templates[4];
- or I miss some eliminations in T4 (I've already checked my code several times, but I'll do it again).
It's the other way around: my code never makes it to T4.
Either it deletes too much in T3, and by luck, it never deletes anything crucial; or your T3 "code" isn't deleting as much as it could/should.
