Behavior of the Explorer's sort, and table.sort()

[DevBuckette]

table.sort
As a Roblox developer, it is currently inefficient to sort a table alphabetically (using table.sort(t) as intended). It results in an illogical manner. This is also occurs in the Explorer. It is intended to sort all objects in it alphabetically, including numbers.
However because of how it’s handled, things like this happen:

Which should be

It would make much more sense if this code:

local t = {"1","5","3","4","7","2","6","10","8","9"}
table.sort(t)
for _ , number in pairs (t) do
print(number)
end

Would result in:

>1
>2
>3
>4
>5
>6
>7
>8
>9
>10

[sleitnick]

But that’s how a proper alphabetical sort works. What you’re asking for is a more contextual-based sort. Windows started doing this in either Vista or Windows 7, but before had the same issue. (e.g. a file named “Image_10” would show before “Image_2”).

In an alphabetical sort, the algorithm looks at a character-per-character basis; it doesn’t consider the whole context (e.g. finding a multi-digit number).

Changing the default behavior of table.sort could result in game-breaking issues.

[Operatik]

Pro tip: Add a 0 before the number, if you plan them to have double digits, or tripple digits.

For some reason they always read from left to right order of characters.

[sleitnick]

I agree with this. If you’re generating things progamatically, you can automatically pad your numbers using this format:

string.format("%.3i", number)

Where the “3” represents the number of digits. For instance:

for i = 1,100 do print(string.format("Hello_%.3i", i)) end

Output:

Hello_001
Hello_002
Hello_003
...
Hello_098
Hello_099
Hello_100

[DevBuckette]

I did just test this, and worked as you showed. Can I ask what the two i 's represents in

string.format("Hello_%.3i", i)

[sleitnick]

In the format function, an i or d represent an integer. The i by itself as the 2nd parameter is from the for loop. Check out the Format String page for more info.

[sleitnick]

I did some more research on this. Apparently what you’re looking for is something called a “Natural Sort”. After some googling, I found a Lua implementation. I reformatted it just a bit:

[Code Source]

function NaturalSort(tbl)
	local function Convert(s)
		local res, dot = "", ""
		for n, m, c in tostring(s):gmatch("(0*(%d*))(.?)") do
			if (n == "") then
				dot, c = "", dot .. c
			else
				res = res .. (dot == "" and ("%03d%s"):format(#m, m) or "." .. n)
				dot, c = c:match("(%.?)(.*)")
			end
			res = res .. c:gsub(".", "\0%0")
		end
		return res
	end
	table.sort(tbl, function(a, b)
		local ca, cb = Convert(a), Convert(b)
		return (ca < cb or ca == cb and a < b)
	end)
	return tbl
end

Example:

local strs = {"Hi10", "Hi2", "Hi1", "Hi32", "Hi5", "Hi3", "Hi0"}
NaturalSort(strs)
for i,s in pairs(strs) do print(i,s) end

And here is the blog post that gave me lots of good info on this subject.

[Anaminus]

Here’s what I went with. It works by continually splitting each string by its first sequence of digits.

local function NaturalSort(a, b)
	-- Split string by first sequence of digits.
	local function findNum(s)
		local i, j, n = s:find("(%d+)")
		if not i then
			return s, 0, ""
		end
		-- Return prefix, number, suffix
		return s:sub(1, i-1), tonumber(n), s:sub(j+1)
	end

	local apfx, anum, bpfx, bnum
	while true do
		if b == "" then
			return false -- b < a
		end
		if a == "" then
			return true -- a < b
		end
		apfx, anum, a = findNum(a)
		bpfx, bnum, b = findNum(b)
		if apfx ~= bpfx then
			return apfx < bpfx
		end
		if anum ~= bnum then
			return anum < bnum
		end
	end
end

local t = {"foo10","bar29","foo11","bar2","foo1","foo7","bar25","foo12","foo9"}
table.sort(t, NaturalSort)
for i = 1,#t do print(t[i]) end

Also, the concept of sort order seems to be referred to as “collation”, if you want to look up more.

[DevBuckette]

@Anaminus & @Crazyman32
Thank you for the help in this. I’ll be looking further into this. I really appreciate it

[sleitnick]

Ah thanks Google lol

[DevBuckette]

It explains the term “collage”, as