Add WireLib.ParseEscapes

[Denneisk]

New WireLib function ParseEscapes. Takes a string input generated by the user (such as from a TextEntry panel) and parses escape characters out. The output is the escaped string. This is a complete implementation of Luajit's escape sequences.

Purpose:
Some user inputs have their own implementation of escaping characters. This is meant to be a unified and thorough alternative and supports the entire Lua escape sequence library.

Notes:
This differs from Lua's escaping only in that invalid escape sequences are skipped instead of erroring.

I've only implemented this in constant value more as a demonstration than anything. Any user input should be changed to use this if desired. E2's tokenizer, for example, is a noteworthy target.

[Denneisk]

Noticed E2 string literals could advantage from this. Not going to touch it until #2555 is done though.

[Denneisk]

Feels a bit uncomfortable adding any more passes. Also I added WireLib.HexToNum because I thought it could also be useful but let me know if that's overstepping.

[thegrb93]

HexToNum can just be tonumber(x,16)

[Denneisk]

Of course. I knew I was missing something.

[Denneisk]

The implementation of this keeps bugging me, obvious global functions or not. I'm going to draft it until it feels right.

[Denneisk]

That's not even the multi-pass gsub. I'm going to bed.

[thegrb93]

I'd like to see an implementation more like

local singleCharEscapes =
{
	n = "\n", r = "\r", t = "\t", ["\\"] = "\\",
	["'"] = "'", ["\""] = "\"", a = "\a",
	b = "\b", f = "\f", v = "\v"
}

local multiCharEscapes =
{
	x = function(arg)
		
	end
}

str = string.gsub(str, "(\\(.?)([^\\]?[^\\]?))", function(str, i, arg)
	return singleCharEscapes[i] or (multiCharEscapes[i] and multiCharEscapes[i](arg)) or str
end)

[Denneisk]

Not as simple against unicode but I'll see what I can do.

[Denneisk]

I can only feasibly get that idea to make sense by still having unicode parsed separately, and in that case it only parses two (or 10) multi conditions which is \x and \0-9.

[Denneisk]

I think I'll settle for this which derives from your idea but inlines each case, which looks almost like what I started with... huh.

[thegrb93]

You can't do two passes otherwise stuff like \\u will get parsed by the unicode despite it should parse to \u

[Denneisk]

Ah, you're right. Sorry, I'm not being thorough enough.

[Denneisk]

I tried your suggestion again and realized it was simpler than I had thought. The only thing I'd like to know is if you wouldn't mind having it inlined instead.

[thegrb93]

Looks pretty nice. I don't mind if its inlined.

[thegrb93]

One last thing that we might want to be done is to benchmark it and limit the string length depending on how much it can do. Like maybe feed it a string.rep("\ ", 100)

[Denneisk]

Hex should be specifically two characters long to match the Lua standard.

[thegrb93]

All right, sure

[Denneisk]

On my 2015 CPU in a server filled with sents, benchmark is between 0.001 to 0.004 seconds with 27075 semi-randomized characters and sequences with roughly 6290 matches. Even string.reping it to ~31440 matches is almost always less than a tick (0.009 seconds average). So, really inconsequential.
Using an escapeless string of 88064 chars is generally under 0.001 seconds.

[Denneisk]

And I was just about to include it in the tokenizer. Oh well, another time.