What's the fastest way to get a random pattern match from a very long string in lua?

Question

I have a string with over 2 million characters, and I feel like my current way of finding a random match from a pattern isn't fast as it could be.

local function getRandomMatch(string, pattern)
    local occurenceCount = select(2, string.gsub(string, pattern, ""))
    local index, randomIndex = 0, math.random(1, occurenceCount)
    for match in string:gmatch(pattern) do
        index = index + 1
        if index == randomIndex then
            return match
        end
    end
end

Is there a way this could be any faster?

If your string is `1234` and your pattern is `%d%d`, you will never get `23` with your algorithm. — Egor Skriptunoff, Aug 07 '18 at 07:04

Egor Skriptunoff · Accepted Answer · 2018-08-08T07:30:42.230

local find, random, match = string.find, math.random, string.match

local function getRandomMatch(string, pattern)
   local pos, random_pos = 0, 0
   for cnt = 1, math.huge do
      pos = find(string, pattern, pos + 1)
      if not pos then
         return match(string, pattern, random_pos)
      elseif random(cnt) == 1 then
         random_pos = pos
      end
   end
end

for j = 1, 20 do
   print(getRandomMatch("1234", "%d%d"))
end

UPDATE:
Fast-and-Dirty solution:
("Dirty" means "matches are random but chosen with non-equal probabilities")

local random, match = math.random, string.match

local function getRandomMatchFastAndDirty(string, pattern)
   return match(string, pattern, random(#string)) or match(string, pattern)
end

This ran much much slower for me, but it did help address a problem I had. — Yuzu, Aug 07 '18 at 18:56

What's the fastest way to get a random pattern match from a very long string in lua?

1 Answers1