Re: question on the patch FastString

[Philippe Verdy <verdyp@...>]

Le mer. 27 mai 2020 à 04:39, 云风 Cloud Wu <cloudwu@gmail.com> a écrit :
>
> I wrote a script to check the collision of those hash functions by
> simulating lua hash tables [1].
> [1] : https://gist.github.com/cloudwu/61befc67d5ef6eee7ef8fda251612439


Note that there's a small bug in your simulation:

  local function hashstring(str, func)
    local h = seed
    str:gsub(".", function (c) h = h ~ ( func(h, c:byte()) & 0xffffffff ) end)
    return h
  end

The initialisation should be:

   local h = seed ~ #str   -- the string length (or at least some low-order bits) should also be hashed (xor'ed with the seed)