Can you propose a more elegant way to 'tokenize' c# code for html formatting?

Question

(This question about refactoring F# code got me one down vote, but also some interesting and useful answers. And 62 F# questions out of the 32,000+ on SO seems pitiful, so I'm going to take the risk of more disapproval!)

I was trying to post a bit of code on a blogger blog yesterday, and turned to this site, which I had found useful in the past. However, the blogger editor ate all the style declarations, so that turned out to be a dead end.

So (like any hacker), I thought "how hard can it be?" and rolled my own in <100 lines of F#.

Here is the 'meat' of the code, which turns an input string into a list of 'tokens'. Note that these tokens aren't to be confused with the lexing/parsing-style tokens. I did look at those briefly, and though I hardly understood anything, I did understand that they would give me only tokens, whereas I want to keep my original string.

The question is: is there a more elegant way of doing this? I don't like the n re-definitions of s required to remove each token string from the input string, but it's difficult to split the string into potential tokens in advance, because of things like comments, strings and the #region directive (which contains a non-word character).

//Types of tokens we are going to detect
type Token = 
    | Whitespace of string
    | Comment of string
    | Strng of string
    | Keyword of string
    | Text of string
    | EOF

//turn a string into a list of recognised tokens
let tokenize (s:String) = 
    //this is the 'parser' - should we look at compiling the regexs in advance?
    let nexttoken (st:String) = 
        match st with
        | st when Regex.IsMatch(st, "^\s+") -> Whitespace(Regex.Match(st, "^\s+").Value)
        | st when Regex.IsMatch(st, "^//.*?\r?\n") -> Comment(Regex.Match(st, "^//.*?\r?\n").Value) //this is double slash-style comments
        | st when Regex.IsMatch(st, "^/\*(.|[\r?\n])*?\*/") -> Comment(Regex.Match(st, "^/\*(.|[\r?\n])*?\*/").Value) // /* */ style comments http://ostermiller.org/findcomment.html
        | st when Regex.IsMatch(st, @"^""([^""\\]|\\.|"""")*""") -> Strng(Regex.Match(st, @"^""([^""\\]|\\.|"""")*""").Value) // unescaped = "([^"\\]|\\.|"")*" http://wordaligned.org/articles/string-literals-and-regular-expressions
        | st when Regex.IsMatch(st, "^#(end)?region") -> Keyword(Regex.Match(st, "^#(end)?region").Value)
        | st when st <> "" -> 
                match Regex.Match(st, @"^[^""\s]*").Value with //all text until next whitespace or quote (this may be wrong)
                | x when iskeyword x -> Keyword(x)  //iskeyword uses Microsoft.CSharp.CSharpCodeProvider.IsValidIdentifier - a bit fragile...
                | x -> Text(x)
        | _ -> EOF

    //tail-recursive use of next token to transform string into token list
    let tokeneater s = 
        let rec loop s acc = 
            let t = nexttoken s
            match t with
            | EOF -> List.rev acc //return accumulator (have to reverse it because built backwards with tail recursion)
            | Whitespace(x) | Comment(x) 
            | Keyword(x) | Text(x) | Strng(x) -> 
                loop (s.Remove(0, x.Length)) (t::acc)  //tail recursive
        loop s []

    tokeneater s

(If anyone is really interested, I am happy to post the rest of the code)

EDIT Using the excellent suggestion of active patterns by kvb, the central bit looks like this, much better!

let nexttoken (st:String) = 
    match st with
    | Matches "^\s+" s -> Whitespace(s)
    | Matches "^//.*?\r?(\n|$)" s -> Comment(s) //this is double slash-style comments
    | Matches "^/\*(.|[\r?\n])*?\*/" s -> Comment(s)  // /* */ style comments http://ostermiller.org/findcomment.html
    | Matches @"^@?""([^""\\]|\\.|"""")*""" s -> Strng(s) // unescaped regexp = ^@?"([^"\\]|\\.|"")*" http://wordaligned.org/articles/string-literals-and-regular-expressions
    | Matches "^#(end)?region" s -> Keyword(s) 
    | Matches @"^[^""\s]+" s ->   //all text until next whitespace or quote (this may be wrong)
            match s with
            | IsKeyword x -> Keyword(s)
            | _ -> Text(s)
    | _ -> EOF

Answer 1

I'd use an active pattern to encapsulate the Regex.IsMatch and Regex.Match pairs, like so:

let (|Matches|_|) re s =
  let m = Regex(re).Match(s)
  if m.Success then
    Some(Matches (m.Value))
  else
    None

Then your nexttoken function can look like:

let nexttoken (st:String) =         
  match st with        
  | Matches "^s+" s -> Whitespace(s)        
  | Matches "^//.*?\r?\n" s -> Comment(s)
  ...