Lexer in D

[Namespace]

> This is a mess of conditionals is what a direct array of 256 
> entries would avoid:
>
> 	int idx;
> 	if (value[0] != '_') {
> 		idx = value[0] - 'a';
> 		if (idx == 26) return false;
> 	} else {
> 		idx = 26;
> 	}
> 	
> 	if (idx < 0 || idx > 26) {
> 		// debug writeln("kword: ", idx, ':', value[0]);
> 		return false;
> 	}
> 	
> 	if (keywords[idx] is null) return false;
> 	
> 	return keywords[idx].canFind(value);
>
> Gaining some speed in the process. Plus another layer of array 
> to discern keywords by length. You see why I suggested to 
> generate the code in the first place ? ;)
>
> BTW what's the reason to separate keywords and type keywords? 
> They are processed the same in lexer and only parser somewhere 
> up above knows what to do with these regardless. Just return 
> different token values for each.

I changed it and merged them together.
Also I use now a array of 256 entities, but I must keep the check 
if idx is < 0, because 'D' - 'a' is negative.
And yes I see what you meant.^^

Code: http://dpaste.1azy.net/317241c0
I reach still 215 - 230 msecs.