I think that the best way to proceed is to use a modified Knuth shuffle. The code below is a complete example. Just run it in a shell with swiftc -o customshuffle customshuffle.swift && ./customshuffle
after saving the content in customshuffle.swift
.
import Foundation
let decksize = 52
let rankStrings = ["A","2","3","4","5","6","7","8","9","10","J","Q","K"]
let suitStrings = ["♠", "♥","♦","♣"]
struct card : Hashable, Equatable, CustomStringConvertible {
var rank: Int //1,2...,11,12,13
var suit: Int // 1,2,3,4
var hashValue: Int {
return rank + suit
}
static func == (lhs: card, rhs: card) -> Bool {
return lhs.rank == rhs.rank && lhs.suit == rhs.suit
}
var description: String {
return rankStrings[self.rank - 1] + suitStrings[self.suit - 1]
}
}
// seems like Swift still lacks a portable random number generator
func portablerand(_ max: Int)->Int {
#if os(Linux)
return Int(random() % (max + 1)) // biased but I am in a hurry
#else
return Int(arc4random_uniform(UInt32(max)))
#endif
}
// we populate a data structure where the
// cards are partitioned by rank and then suit (this is not essential)
var deck = [[card]]()
for i in 1...13 {
var thisset = [card]()
for j in 1...4 {
thisset.append(card(rank:i,suit:j))
}
deck.append(thisset)
}
// we write answer in "answer"
var answer = [card]()
// we pick a card at random, first card is special
var rnd = portablerand(decksize)
answer.append(deck[rnd / 4].remove(at: rnd % 4))
while answer.count < decksize {
// no matter what, we do not want to repeat this rank
let lastrank = answer.last!.rank
var myindices = [Int](deck.indices)
myindices.remove(at: lastrank - 1)
var totalchoice = 0
var maxbag = -1
for i in myindices {
totalchoice = totalchoice + deck[i].count
if maxbag == -1 || deck[i].count > deck[maxbag].count {
maxbag = i
}
}
if 2 * deck[maxbag].count >= totalchoice {
// we have to pick from maxbag
rnd = portablerand(deck[maxbag].count)
answer.append(deck[maxbag].remove(at: rnd))
} else {
// any bag will do
rnd = portablerand(totalchoice)
for i in myindices {
if rnd >= deck[i].count {
rnd = rnd - deck[i].count
} else {
answer.append(deck[i].remove(at: rnd))
break
}
}
}
}
for card in answer {
print(card)
}
This can be computed fast and it is reasonably fair, but not unbiased sadly. I suspect it might be hard to generate a "fair shuffle" with your constraint that runs fast.