Character class for Unicode digits

Question

I need to create a Pattern that will match all Unicode digits and alphabetic characters. So far I have "\\p{IsAlphabetic}|[0-9]".

The first part is working well for me, it's doing a good job of identifying non-Latin characters as alphabetic characters. The problem is the second half. Obviously it will only work for Arabic Numerals. The character classes \\d and \p{Digit} are also just [0-9]. The javadoc for Pattern does not seem to mention a character class for Unicode digits. Does anyone have a good solution for this problem?

For my purposes, I would accept a way to match the set of all characters for which Character.isDigit returns true.

To a certain degree, I'd say yes. The Character class has an isDigit method which returns true for non-Latin digits (I've been testing with Japanese numerals). — Aurand, Feb 18 '13 at 22:10
What happens if you use Pattern.UNICODE_CHARACTER_CLASS like in `Pattern p = Pattern.compile("(?=\\w+", Pattern.UNICODE_CHARACTER_CLASS)` ? — Cyrille Ka, Feb 18 '13 at 22:10
This might help: - http://stackoverflow.com/a/4307261/1679863 — Rohit Jain, Feb 18 '13 at 22:10
@Pillar I had a good look for a dupe, but didn't find one. Well done - it's a dupe. — Bohemian, Apr 25 '16 at 18:31

mgibsonbr · Accepted Answer · 2013-02-18T22:36:57.220

7

Quoting the Java docs about isDigit:

A character is a digit if its general category type, provided by getType(codePoint), is DECIMAL_DIGIT_NUMBER.

So, I believe the pattern to match digits should be \p{Nd}.

Here's a working example at ideone. As you can see, the results are consistent between Pattern.matches and Character.isDigit.

edited Feb 18 '13 at 22:36

answered Feb 18 '13 at 22:12

mgibsonbr

21,755
7
70
112

Simply `\p{N}` works: `System.out.println("3๓३".matches("\\p{N}+")) // true` – Bohemian Apr 25 '16 at 18:33
2

@Bohemian But `p{N}` also matches `Nl` and `No`, which `isDigit` does not match. [Example](http://ideone.com/1GHJ1P). Sometimes you *want* to match those, but since the OP asked for a behavior consistent with `isDigit`, I answered with just `Nd`. – mgibsonbr Apr 26 '16 at 00:12

score 5 · Answer 2 · edited May 20 '16 at 16:52

5

Use \d, but with the (?U) flag to enable the Unicode version of predefined character classes and POSIX character classes:

(?U)\d+

or in code:

System.out.println("3๓३".matches("(?U)\\d+")); // true

Using (?U) is equivalent to compiling the regex by calling Pattern.compile() with the UNICODE_CHARACTER_CLASS flag:

Pattern pattern = Pattern.compile("\\d", Pattern.UNICODE_CHARACTER_CLASS);

edited May 20 '16 at 16:52

Wiktor Stribiżew

607,720
39
448
563

answered Apr 25 '16 at 18:23

Bohemian

412,405
93
575
722

One of the rare occasions where I don't think a duplicate post deserves a downvote. New answer and all that. Don't you have merge powers, or does that not apply here? – Savior Apr 25 '16 at 19:23
1

@Pillar so merged. IMHO this answer is easier to remember and understand - who can remember all those funky posix classes? – Bohemian Apr 25 '16 at 19:29
What POSIX classes do you mean? POSIX character classes are `[:punct:]`, `[:digit:]`, etc. `\p{N}`, or `\p{L}` etc. are Unicode category classes (term used in .NET), or Unicode character properties (term used in PHP), and these are very handy, especially `\p{Ll}` and `\p{Lu}`. In Java, certainly `(?U)\d` looks preferable. – Wiktor Stribiżew Apr 28 '16 at 20:54

Character class for Unicode digits

2 Answers2

Related