Quick profiling

p.lua

local md5 = require "main.md5"

local by_index = {}
local by_id = {}

local function time_it(fnc)
	local before = os.clock()
	fnc()
	print((os.clock() - before)*1000)
end


local function find_by_id(find_this)
	return by_id[find_this]
end

local function find_by_index(find_this)
	local id
	for i=1, #by_index do
		id = by_index[i]
		if find_this == id then
			return true
		end
	end
	return false
end

local function loop_all_index()
	local id
	local _trash = ""
	for i=1, #by_index do
		id = by_index[i]
		_trash = "well" .. "hi" .. "there"
	end
	return true
end

local function loop_all_id()
	local _trash = ""
	for id, _ in pairs(by_id) do
		_trash = "well" .. "hi" .. "there"
	end
	return true
end

function init(self)
	local before = os.clock()
	local total_elements = 1000000
	local m
	local id
	for i=1, total_elements do
		m = md5.new()
		m:update(tostring(i))
		id = md5.tohex(m:finish())
		by_id[id] = true
		by_index[#by_index + 1] = id
	end
	print("Setup", (os.clock() - before)*1000)
	
	local first = by_index[1]
	local mid = by_index[total_elements/2]
	local last = by_index[total_elements]

	print("find_by_id: first")
	time_it(function() find_by_id(first) end)
	print("find_by_id: mid")
	time_it(function() find_by_id(mid) end)
	print("find_by_id: last")
	time_it(function() find_by_id(last) end)

	print("find_by_index: first")
	time_it(function() find_by_index(first) end)
	print("find_by_index: mid")
	time_it(function() find_by_index(mid) end)
	print("find_by_index: last")
	time_it(function() find_by_index(last) end)

	print("loop: loop_all_id")
	time_it(loop_all_id)
	
	print("loop: loop_all_index")
	time_it(loop_all_index)
end

Did some more tests, had to keep it at 1m to even get some numbers on it. Anything lower and most stuff wouldn't show any thing. So first observation: The table need to have a lot of information in it to actually matter much.

Removing with an index (Pokemon number 123121 or something) was a lot faster in an indexed list but not by much at all. When removing by ID the hash table was a lot quicker, practically instant (showed 0 on the test) while it took a lot of time for the indexed table (24-35). Later test have also shown that that looping over an indexed table can also be slow.

I think we might be optimizing when there is no need either. But for now I will change it to use the "hashed" list.