BTW, yet another way to do it (more compact than the bitwise and prime options) is the Cantor pairing function https://en.wikipedia.org/wiki/Pairing_function

... z = (x+y+1)(x+y)/2 + y - but you have to sort x,y first to get the order independence you wanted. This function is famously used in the argument that the set of integers and the set of rationals have the same cardinality.

▲

noduerme 4 days ago | parent [-]

mm. I did see this when I was figuring it out. The sorting first was the specific thing I wanted to avoid, because it would've been by far the most expensive part of the operation when looking at a million poker hands and trying to target several players for potential collusion.

▲

bazzargh 3 days ago | parent [-]

you're only sorting players within a single hand. so a list of under 10 items? thats trivial

	▲	noduerme 3 days ago \| parent [-]
		So the goal was to generate signatures for 2, 3 or more players and then be able to reference anything in the history table that had that combination of players without doing a full scan and cross-joining the same table multiple times. Specifically to avoid having ten index columns in the history table for each seat's player. This was also prior to JSON querying in mysql. I needed a way to either bake in the combinations at write time, or to generate a unique id at read time in a way that wouldn't require me to query whether playerIDs were [1201,1803,2903] or [1803,1201,2903] etc. Just a one-shot unique signature for that combination of players that could always evaluate the same regardless of the order. If that makes sense. There were other considerations and this was not exactly how it worked, since only certain players were flagged and I was looking for patterns when those particular players were on the same table. It wasn't like every combination of players had a unique id, just a few combinations where I needed to be able to search over a large space to find when they were in the same room together, but disregarding the order they were listed in.