Matching Line Boundaries in a Regular Expression (Pattern.MULTILINE/(?m)) is broken in Java?

Question

The example on http://www.exampledepot.com/egs/java.util.regex/Line.html gives false for me twice but should'nt! Why?

CharSequence inputStr = "abc\ndef";
String patternStr = "abc$";

// Compile with multiline enabled
Pattern pattern = Pattern.compile(patternStr, Pattern.MULTILINE);
Matcher matcher = pattern.matcher(inputStr);
boolean matchFound = matcher.find();    // true

// Use an inline modifier to enable multiline mode
matchFound = pattern.matches(".*abc$.*", "abc\r\ndef");     // false
System.out.println(matchFound); // false
matchFound = pattern.matches("(?m).*abc$.*", "abc\r\ndef"); // true
System.out.println(matchFound);// false !!!!!

Well, if I put an additionally (?s) it works, but shouldn't it work without the (?s)? Did this behaviour change in the past or didn't the authors simply check there example?

Answer 1

First of all, exampledepot.com is an extremely bad site: never ever assume anything found there as being "the truth".

In regex, the $ never matches a character, it matches a position. In (?m) mode, it matches the "empty string" before a line break, or the end of the string. So given the string "abc\r\ndef" the regex ".*abc$.*" does not match since the \r\n isn't present in your regex. The $ matches the position between c and \r.

What you should do is this:

System.out.println("abc\r\ndef".matches(".*abc$\r\n.*"));     // false
System.out.println("abc\r\ndef".matches("(?m).*abc$\r\n.*")); // true

Answer 2

I'm not familiar with how community wiki works, but feel free to use this example if deemed useful.

    System.out.println(
        Pattern.matches("(?m)^abc$\n^def$", "abc\ndef")
    ); // prints "true"

    System.out.println(
        Pattern.matches("(?sm)^abc$.^def$", "abc\ndef")
    ); // prints "true"