java regex preferences when multiple matches occur

Question

Hi I'm trying to understand in particular how does the OR operator "|" work in java regex. I'm just starting to use it and most of the things are unclear to me.

Suppose I wish to match fractions and integers, that is to say things of the form 1/2, 12/25, and also things of the form 13, 235, etc.

I have tried these 2 patterns:

pattern1 = Pattern.compile("\\d+|\\d+/\\d+"))
pattern2 = Pattern.compile("\\d+/\\d+|\\d+"))

In English, pattern1 says "digits OR digits/digits", whereas pattern2 says "digits/digits OR digits".

Now consider this input string:

inputStr = "blah... 231/232 blah... 4 blah... 2"

For pattern1, I found these matches:

[junit] found 231
[junit] found 232
[junit] found 4
[junit] found 2

For pattern2, I found these matches:

[junit] found 231/232
[junit] found 4
[junit] found 2

Now the only difference between pattern1 and pattern2 is the orders of its matched elements. Of course pattern2 is the one I wanted, as it seems to "prefer" a real faction than to take them apart.

So the most important question for me is this: Is this behaviour reliable/predictable, or is it going to be different for different platforms?

But also just curious... this question too: I also find it confusing because the operator "OR" should be symmetric with regard to its arguments, like addition. You'd imagine people be worried when 1+2 and 2+1 carries different semantics... is there any reason for pattern1 and pattern2 here to be semantically different?

Check out the first section: http://www.regular-expressions.info/engine.html — nhahtdh, Oct 10 '12 at 15:04

score 7 · Accepted Answer · answered Oct 10 '12 at 15:02

7

| isn't just OR, it means "match the first thing, and if that fails, match the second thing".

Thus, you want to put the fraction first since it's the preferred form.

answered Oct 10 '12 at 15:02

nneonneo

171,345
36
312
383

For reference, some regexp matchers match `|` greedily, i. e. the longest match. In that case you would not have that problem. I think it's especially *regular expression* matchers (as in those that correspond exactly to the regular languages) that do so. – ReyCharles Oct 10 '12 at 21:01

Victor Mukherjee · Answer 2 · 2012-10-10T15:37:26.427

4

A more useful regex for your purpose would be \\d+(/\\d+)?which mandatorily checks for a group of digits and an optional group formed by a slash followed by digits.

edited Oct 10 '12 at 15:37

answered Oct 10 '12 at 15:07

Victor Mukherjee

10,487
16
54
97

score 2 · Answer 3 · edited May 23 '17 at 10:24

2

The alternation operator is like a lazy-OR in that it will match the first thing it can. There are other posts on the topic that help clarify its behavior:

Java regex alternation operator "|" behavior seems broken

Why order matters in this RegEx with alternation?

In general, all regex's work this way... Except POSIX. So, portability should not be a concern in Java.

edited May 23 '17 at 10:24

Community

1
1

answered Oct 10 '12 at 15:05

jheddings

26,717
8
52
65

score 0 · Answer 4 · answered Oct 10 '12 at 15:04

The | is called an alternation, and it gives the ability to list alternatives for a given match, and will stop on the first alternate pattern matched, from left to right. AFAIK, this is very consistent across all Java versions and the programming languages/tools I've used regexes with: Java, Perl, Python, PHP, sed.

java regex preferences when multiple matches occur

4 Answers4