Anyone know regular expressions?

[Robert Willems]

Hi everybody,
i'm using regular expressions to filter thing from a String.
Can a regular expression be recursive?
I need a expression (that works in org.apache.regexp.RE) that parses something like a function. A function starts with '${' and ends with '}'. But in the text between there can also be new 'instances' of ${...}
Eg.
When parsing the string "bla bla${bla ${bla} ${bla}}bla ${bla} bla" the first thing the epxression needs to give me is "${bla ${bla} ${bla}}"
I'm not very good with expression and have been using \$\{.+?\} as an expression at the moment, but that wont give me the right results.
Does anybody have a good idea for a working expression? Or should i be writing my own parsing routine?
Thanks in advance,
Robert Willems

[Jessica Sant]

moving this to the Java In General (Intermediate) forum for two reasons:
1- this isn't a very advanced question.
2- Max Habibi is a moderator there and he wrote a whole book on Regular Expressions!

[Alan Moore]

If the "${...}" constructs can be nested more than one level deep, you'll have to write your own parser. It's the classic "nested delimiters" problem: regexes can't handle them. However, if you know they'll never go more than one level deep, you can fake it with something like this:

"\\$\\{(?:[^$]|\\$\\{[^${}]*\\})*\\}"

This assumes that, in well-formed text, the '$' character will never appear without the "{...}", and vice versa; if that isn't the case, a more complex regex will be needed.

[Robert Willems]

Thanks Alan,
too bad regexp's cannot be recursive... Oh well we'll just call it an undocumented restriction that those constructs cannot be nested ;-)))
Thanks for your reply...
Regards,
Robert Willems

[Max Habibi]

Hi Robert,
I took a crack @ this, and couldn't come up with anything purely regex: sorry if that busts any illusions you had about authors .
You might want to consider a recursive method call. The method could use regex internally.
All best,
M

[Robert Willems]

Hi Max,
thanks for your reply... my illusions are totally shattered
... but is it possible to 'count' occurences of a string in a regexp? So i can do something like:
- The selected string has to start with ${
- It has to end with }
Not so difficult until this point ;-), but can you do something like:
- It has to contains as much occurences of { as } ?
Because that was another of the main problems.. if i counted on the wellformedness of the string it would be either to reluctent on grabbing characters (returning ${bla {$bla}) or to greedy (returning ${bla ${bla} ${bla}}bla ${bla})
Regards,
Robert

[Michael Dunn]

If the format is always ${..}, not sometimes {..}, this might work (needs testing)

class Testing
{
  public Testing()
  {
    String str = "bla bla${bla ${bla} ${bla}}bla ${bla} bla";
    int count = 0, startPos = 0;
    boolean openBrace = false;
    for(int x = 0; x < str.length()-1; x++)
    {
      if(str.substring(x,x+2).equals("${"))
      {
        if(openBrace)count++;
        else
        {
          openBrace = true;
          startPos = x;
          count++;         
        }
      }
      else if(str.substring(x,x+1).equals("}")) count--;
      if(openBrace && count == 0)
      {
        System.out.println(str.substring(startPos,x+1));
        openBrace = false;
      }
    }        
  }
  public static void main(String[] args){new Testing();}
}