77topaz wrote:Hmm, maybe a function could be implemented so that, if a soup takes more than a certain amount of time to process or goes above a certain size without stabilising, that soup is flagged as abnormal and the searching process continues. That would make it easier to search rules that explode, but quite rarely; currently, in certain rules go through several thousand soups only to have the whole searching process suddenly freeze, preventing me from submitting a haul.
That sounds quite doable -- simply label the whole thing PATHOLOGICAL, and continue with the next soup.
What would a good cut-off value be? 5 minutes? 10? 30? You'd want to strike a balance between not accidentally labelling difficult-to-classify patterns (such as high-period yl's) as PATHOLOGICAL on one hand, and ensuring searchability on the other hand. (Mind, we're talking about rules here where there's an
occasional problematic soup; if all (or most, or many) soups blow up, you still won't be able to search the rule, but that's more due to the rule's own dynamics than to apgsearch's optimism re: eventual stabilization.)
Something else that would be nice to have -- occasional status messages on the console if a problematic soup is encountered.
I don't know if Calcyman's going to read any of this (much less in this thread), or (if he does) whether he's going to have the time and inclination to implement anything, but if you cook up something yourself, let me know.