I had a similar problem while displaying memory pool blocks from the output of some virtual DOM h-functions composition. Basically I faced to the same problem as sorting multi-criteria data like scoring results from players around the world.
I have noticed that multi-criteria sorting is:
- sort by the first column
- if equal, sort by the second
- if equal, sort by the third
- etc... nesting and nesting if-else
And if you don't care, you could fail quickly in a if-else nesting hell... like callback hell of promises...
What about if we write a "predicate" function to decide if which part of alternative using ? The predicate is simply :
// useful for chaining test
const decide = (test, other) => test === 0 ? other : test
Now after having written your classifying tests (byCountrySize, byAge, byGameType, byScore, byLevel...) whatever who need, you can weight your tests (1 = asc, -1 = desc, 0 = disable), put them in an array, and apply a reducing 'decide' function like this:
const multisort = (s1, s2) => {
const bcs = -1 * byCountrySize(s1, s2) // -1 = desc
const ba = 1 *byAge(s1, s2)
const bgt = 0 * byGameType(s1, s2) // 0 = doesn't matter
const bs = 1 * byScore(s1, s2)
const bl = -1 * byLevel(s1, s2) // -1 = desc
// ... other weights and criterias
// array order matters !
return [bcs, ba, bgt, bs, bl].reduce((acc, val) => decide(val, acc), 0)
}
// invoke [].sort with custom sort...
scores.sort(multisort)
And voila ! It's up to you to define your own criterias / weights / orders... but you get the idea. Hope this helps !
EDIT: * ensure that there is a total sorting order on each column * be aware of not having dependencies between columns orders, and no circular dependencies
if, not, sorting can be unstable !