Hi Steve,

The \x3[0-9] approach is interesting, it appears to solve the problem without
paying a higher cost I originally thought (looking back, for example).

The logic of initializing/setting/unsetting of "beginQuote" to true/false appears to be incorrect when there are multiple \Qn...\E in one pattern. Ln#1622 setting will
always be followed by Ln#1630, if I read the code correctly.

For example

Pattern pattern = Pattern.compile("\\011\\Q1sometext\\E\\011\\Q2sometext\\E");
        Matcher matcher = pattern.matcher("\t1sometext\t2sometext");
        System.out.printf("find=%b%n", matcher.find());

will still return false?

-Sherman

On 12/09/2011 10:05 PM, Stephen Flores wrote:
Please review the following webrev (includes new test to demonstrate the bug):

  http://cr.openjdk.java.net/~sflores/6990617/

for bug: 6990617 Regular expression doesn't match if unicode character next to a digit.

A DESCRIPTION OF THE PROBLEM :

  Unicode characters are represented as \\+number.
  For instance, one could write:
            Pattern p = Pattern.compile("\\011some text\\012");
            Matcher m = p.matcher("\tsome text\n");
            System.out.println(m.find()); // yields "true"

  However, if we want to match a string with a digit next to
  the unicode character, it doesn't match (whether we "quote"
  the regular expression or not). Note the "1" next to the tab
  character (unicode 011).
            Pattern p = Pattern.compile("\\011\\Q1some text\\E\\012");
            Matcher m = p.matcher("\t1some text\n");
            System.out.println(m.find());  // yields "false"

  This happens because Pattern accepts either \\0011 or \\011 for
  the same character. From the javadoc:

    \0nn  The character with octal value 0nn (0 <= n <= 7)
    \0mnn The character with octal value 0mnn (0 <= m <= 3, 0 <= n <= 7)

Evaluation:

  Pattern.RemoveQEQuoting() does NOT handle this situation correctly.
  The existing implementation now simply copies all ASCII.isAlnum()
  characters when handing a quote.

Description of fix:

  In the method Pattern.RemoveQEQuoting any ASCII digit at the
  beginning of a quote will now be prefixed by "\x3" (not just \
  because this would be a backref). 0x30 is the ASCII code for '0'.

Thanks,

  Steve.

Reply via email to