RANDOMSTRING.lua

-- Implementations
local function randstring_1(size)
  local c,r = string.char,math.random
  local str = {}
  for i = 1,size do
    str[i] = c(r(0,255))
  end
  return table.concat(str)
end

local MAX_UNPACK_SIZE = 50000
local function randstring_2(size)
  local c,r,u,m = string.char,math.random,(table.unpack or unpack),math.min
  local str,temp,n = {},{},1
  while size > 0 do
    local L = m(size,MAX_UNPACK_SIZE)
    for i = 1,L do
      temp[i] = r(0,255)
    end
    str[n] = c(u(temp,1,L))
    n = n + 1
    size = size - L
  end
  return table.concat(str)
end


-- Profile
local tests = {10^3, 10^4, 10^5, 10^6, 10^7, 10^8, 10^9}
for i,size in ipairs(tests) do
  
  collectgarbage('collect')
  collectgarbage('collect')
  local t = os.clock()
  randstring_1(size)
  print('method 1 ('..size..'):',(os.clock() - t))
  
  collectgarbage('collect')
  collectgarbage('collect')
  local t = os.clock()
  randstring_2(size)
  print('method 2 ('..size..'):',(os.clock() - t))
  
end


--[[ RESULTS on i5-2520M w/ 6 GiB 1333 MHz RAM on Windows 7 SP1 64-bit, via
     64-bit Lua 5.2.3 from http://joedf.users.sourceforge.net/luabuilds/:

  D:\Desktop
  λ lua test.lua
  method 1 (1000):        0
  method 2 (1000):        0
  method 1 (10000):       0.002
  method 2 (10000):       0.002
  method 1 (100000):      0.026
  method 2 (100000):      0.014
  method 1 (1000000):     0.269
  method 2 (1000000):     0.134
  method 1 (10000000):    2.828
  method 2 (10000000):    1.317
  method 1 (100000000):   27.333
  method 2 (100000000):   13.064
  lua: not enough memory

NOTES: On the second-to-last test (10^8), method 1 climbed to ~2.5 GiB of memory
to generate the random string, while method 2 used less than 150 MiB. Method 2's
speed (and memory consumption) can be improved furhter by using larger values
for MAX_UNPACK_SIZE. You need to see what is acceptable. It varies by platform
and build of Lua, it seems.

--]]