The following code takes a string and converts it to it's number equivalent. A=1 ... Z=26, while keeping all other characters intact. It formats the string using two delimiters. One for between characters and one for between words. I want to optimize it, and have fixed a lot of redundancy, and slow parts. However, I can't seem to make it any faster. The longest amount of time is taken for decrypting an encrypted string. I would like to optimize both parts.

Suggestions on techniques or code with explanation would be most helpful. This is mostly for me to understand the bottlenecks in Lua logic.

function parseString(str,del1,del2)
    local delim1 = del1 or ' '
    local delim2 = del2 or '-'
    local tab = {}
    local sub = ''
    local index = 1
    local c = ''
    local i = 1
    while (i > #str and sub == '') == false do
        c = str:sub(i,i)
        if c == delim2 or c == delim1 or c == '' then
            if string.match(sub,'%d') == nil then
                tab[index] = sub
                index = index + 1
            else
                tab[index] = string.char(tonumber(sub) + 96)
                index = index + 1
            end
            if c == delim1 then
                tab[index] = ' '
                index = index + 1
            end
            sub = ''
        else
            sub = sub .. c
        end
        i = i + 1
    end
    return table.concat(tab)
end
function strToNums(str)
    local t = {string.byte(str,1,#str)}
    for i,v in ipairs(t) do
        if v < 97 and v >= 65then
            t[i] = v - 64
        else
            if v <= 172 and v >= 97then
                t[i] = v - 96
            else
                t[i] = v
            end
        end
    end
    return t
end
function formater(del1,del2,arg)
    local out = {}
    local index = 1
    local nDel1 = string.byte(del1)
    local nDel2 = string.byte(del2)
    for i,v in ipairs(arg) do
        if v <= 26 and v >= 1 then
            out[index] = tostring(v)
            index = index + 1
            if i < #arg and arg[i+1] ~= nDel1 then
                out[index] = del2
                index = index + 1
            end
        else
            if v == 32 then
                out[index]  = del1
                index = index + 1
            else
                out[index] = string.char(v)
                index = index + 1
                if i < #arg and arg[i+1] ~= nDel1 then
                    out[index] = del2
                    index = index + 1
                end
            end
        end

    end
    return table.concat(out)
end
function main()
    local line = ''
    local del1 = ' '
    local del2 = '-'
    local x = os.clock()
    local line2 = ''
    local str =''
    for i=1,1000 do
        str = strToNums("This string'll test my system!")
        line = formater(del1,del2,str)
        line2 = parseString(line,del1,del2)
    end
    print(line,line2)
    print(string.format("elapsed time: %.10f\n", os.clock() - x))
    --  while true do
    --      line = io.read()
    --      if string.match(line,'%a+') == nil then break end
    --      line = formater(del1,del2,strToNums(line))
    --      print(line)
    --      print(parseString(line,del1,del2))
    --  end
end
main()

This is the output that I would expect for the given input:

20-8-9-19 19-20-18-9-14-7-'-12-12 20-5-19-20 13-25 19-25-19-20-5-13-!   this string'll test my system!
elapsed time: 0.0880000000

strToNum takes a string and converts it to a table of char values.

Formatter takes two types of delimiters: one for between characters and one for between words. It also takes the table of values generated by strToNum. It takes these values and creates a string with each character separated by one delimiter and each word separated by another.

parseString takes a string and two delimiters. The delimiters are to be checked against. The string is iterated over, and between iterators another sub string is created and added to as long as there is not a delimiter. Each time it runs into a delimiter it adds it to a table. Which is then concatenated at the end. This function takes the most amount of time to run. I think it is because of the using substring, string.char and tonumber.