Re: Cant think of an elegant lua solution to this little problem

[Paul K <paulclinger@...>]

Hi Sergey,

> The improved version does still suffer from the bug caused by appending
> (99-#z): "012b" > "12a", whereas it should be "012b" > "12a".
> In case "#t == 1" you shouldn't append #r, because "0.11" < "0.2".

Thank you for pointing this out (I think you meant "012b" < "12a" in
the first comparison). Here is a better solution that fixes both of
these issues (although it's also not perfect in dealing with leading
zeroes):

function alphanumsort(o)
  local function padnum(d) local dec, n = string.match(d, "(%.?)0*(.+)")
    return #dec > 0 and ("%.12f"):format(d) or
("%s%03d%s"):format(dec, #n, n) end
  table.sort(o, function(a,b)
    return tostring(a):gsub("%.?%d+",padnum)..("%3d"):format(#b)
         < tostring(b):gsub("%.?%d+",padnum)..("%3d"):format(#a) end)
  return o
end

This also sorts decimal numbers:

Abc 0.100a
Abc 0.1b
Abc 0.11
Abc 0.2

I updated the page to capture various solutions we discussed (from
simplest to the best so far) as well as their results:
http://notebook.kulchenko.com/algorithms/alphanumeric-natural-sorting-for-humans-in-lua.

> "." < " "   -- you'll face the ("." > "/") bug instead

I didn't quite get this comment...

> I doubt the real numbers case has a use though :)

That I agree with ;).

Paul.

On Sat, Aug 25, 2012 at 6:46 PM, GrayFace <sergroj@mail.ru> wrote:
> On 26.08.2012 8:09, Paul K wrote:
>>
>> Agree; I just sent an improved version that shouldn't suffer from this
>> problem.
>
>
> I see we used the same idea of first comparing the length of numbers and
> then using lexicographical comparison of numbers.
> The improved version does still suffer from the bug caused by appending
> (99-#z): "012b" > "12a", whereas it should be "012b" > "12a".
> In case "#t == 1" you shouldn't append #r, because "0.11" < "0.2".
> You made me also think about the real numbers case. My attempt at
> implementing it has these bugs so far:
> "." < " "   -- you'll face the ("." > "/") bug instead
> "0.100a" > "0.1b"
> I doubt the real numbers case has a use though :)
>
>
> --
> Best regards,
> Sergey Rozhenko                 mailto:sergroj@mail.ru
>
>