On 2008-02-03 23:26:58 -0500 "Mukul Gandhi" <gandhi.mukul@g...> 
wrote:
> String str = "<root><abc x='1'>text1</abc><pqr 
> y='1'>text2</pqr></root>";
> 
> Pattern pattern = Pattern.compile("<[^/]+>");  //anything from '<' to
> '>', and not having '/'
> Matcher matcher = pattern.matcher(str);
> 
> while (matcher.find()) {
>    String group = matcher.group();
>    System.out.println(group);
> }
> 
> 'str' is a String representation of a XML fragment.
> 
> I want to extract all pieces from the string (the tokens), which form
> a start tag (including attribute parts).
> 
> I am expecting output:
> <root>
> <abc x='1'>
> <pqr y='1'>

But that's not what you asked for.  You said "longest string starting 
with '<' and ending with '>' that doesn't contain '/'.

> But the output produced by the above program is:
> <root><abc x='1'>
> <pqr y='1'>

Yup.  Exactly matches the regex.  No / in either one, is there?  
Specifically, even though you think you asked for "just the start 
tag," you have <abc> nested inside <root>; there's no / anywhere 
around to prevent the regex from matching to the end of <abc>

The problem with using regular expressions to parse any grammar with 
paired tokens (XML for example, but also most programming languages 
with paired braces of any sort, or comments in a language that permits 
comment nesting) is that regular expressions can't handle parity.

You need something more powerful than regex.

If you're determined to find the next layer of problems associated 
with using a too-weak tool to do the job, you should find it shortly 
after making this change:

Pattern.compile("<[^/<]+>");

That prevents it from picking up a nested element tag.  Most of the 
time.

For giggles:

<root><?my-pi wotsit ?><abc x='1'><![CDATA[<?xml version="1.0?>
<root><abc x='1'>text1]]></abc>
</root>

HTH.

Amy!
-- 
Amelia A. Lewis                    amyzing {at} talsever.com
Confidence: a feeling peculiar to the stage just before full
comprehension of the problem.