Question

I’m learning regexp and thought I was starting to get a grip. but then…

I tried to split a string and I need help to understand such a simple thing as:

String input = "abcde";
System.out.println("[a-z] " + Arrays.toString(input.split("[a-z]")));
System.out.println("\\w " + Arrays.toString(input.split("\\w")));
System.out.println("\\w*? " + Arrays.toString(input.split("\\w*?")));
System.out.println("\\w+? " + Arrays.toString(input.split("\\w+?")));

The output is
[a-z] - []
\w    - []
\w*?  - [, a, b, c, d, e]
\w+?  - []

Why doesn’t any of the two first lines split the String on any character?
The third expression \w*?, (question mark prevents greediness) works as I expected, splitting the String on every character. The star, zero or more matches, returns an empty array.

I’ve tried the expression within NotePad++ and in a program and it shows 5 matches as in:

Scanner ls = new Scanner(input);
while(ls.hasNext())
    System.out.format("%s ", ls.findInLine("\\w");

Output is: a b c d e

This really puzzles me

Answer 1

If you split a string with a regex, you essentially tell where the string should be cut. This necessarily cuts away what you match with the regex. Which means if you split at \w, then every character is a split point and the substrings between them (all empty) are returned. Java automatically removes trailing empty strings, as described in the documentation.

This also explains why the lazy match \w*? will give you every character, because it will match every position between (and before and after) any character (zero-width). What’s left are the characters of the string themselves.

Let’s break it down:

1. [a-z], \w, \w+?

   Your string is

   abcde
   

   And the matches are as follows:

    a  b  c  d  e
   └─┘└─┘└─┘└─┘└─┘
   

   which leaves you with the substrings between the matches, all of which are empty.

   The above three regexes behave the same in this regard as they all will only match a single character. \w+? will do so because it lacks any other constraints that might make the +? try matching more than just the bare minimum (it’s lazy, after all).

2. \w*?

     a  b  c  d  e
   └┘ └┘ └┘ └┘ └┘ └┘
   

   In this case the matches are between the characters, leaving you with the following substrings:

   "", "a", "b", "c", "d", "e", ""
   

   Java throws the trailing empty one away, though.