java pattern matcher

Thread view:

Moiristo - 12 Jul 2006 16:35 GMT

I have a problem with the java Matcher.

Consider the following regex: <<<(.)+>>>
Now, consider the following String containing two occurrences of the
above regex:

SELECT TOP <<<Upper limit>>> FROM (SELECT TOP <<<Lower limit>>> FROM person)

However, when I use Matcher.find(), it does not what I expect it to do.
Instead of matching the two substrings, it finds only one, namely:

'Upper limit>>> FROM (SELECT TOP <<<Lower limit'

This is of course correct, but not what I intended. Can someone help me
to solve this?

Reply to this Message

Stefan Ram - 12 Jul 2006 16:42 GMT

>Consider the following regex: <<<(.)+>>>

Try

<<<(.)+?>>>

or

<<<([^>])+>>>

Reply to this Message

Stefan Ram - 12 Jul 2006 16:46 GMT

> Try
><<<(.)+?>>>
> or
><<<([^>])+>>>

But you actually might want

<<<(.+?)>>>

or

<<<([^>]+)>>>

, respectively.

Reply to this Message

Moiristo - 12 Jul 2006 18:24 GMT

>> Try
>> <<<(.)+?>>>
[quoted text clipped - 10 lines]
>
> , respectively.

Thank you! Could you please explain why this works and not the regex I used?

Reply to this Message

Oliver Wong - 12 Jul 2006 18:36 GMT

>>> Try
>>> <<<(.)+?>>>
[quoted text clipped - 11 lines]
> Thank you! Could you please explain why this works and not the regex I
> used?

I guess Java's regular expression engine is by default greedy (most RE
engines are greedy by default). I'm guessing the first alternative, (.+?),
the '?' acts as a modifer on '+', telling it not to be greedy. That is,
instead of matching the longest possible substring, it tries to match the
shortest possible substring.

In the second alternative, instead of accepting "<<<" followed by
anything, followed by ">>>", it accepts "<<<" followed by anything except
">", followed by ">>>".

- Oliver

Reply to this Message

lordy - 13 Jul 2006 00:29 GMT

>>>> Try
>>>> <<<(.)+?>>>
[quoted text clipped - 23 lines]
>
> - Oliver

And the latter is more efficient. The first will do a lot of
backtracking given the expected input strings.

Lordy

Reply to this Message

Stefan Ram - 12 Jul 2006 18:59 GMT

>> <<<(.+?)>>>
>> <<<([^>]+)>>>
>Thank you! Could you please explain why this works and not the regex I used?

The first suggestion should be used, because the second one
fails to match "<<<ab<c>def>>>".

Oliver by now has explained the expressions. See also:

http://download.java.net/jdk6/docs/api/java/util/regex/Pattern.html

Reply to this Message

Moiristo - 12 Jul 2006 20:50 GMT

>>> <<<(.+?)>>>
>>> <<<([^>]+)>>>
[quoted text clipped - 6 lines]
>
> http://download.java.net/jdk6/docs/api/java/util/regex/Pattern.html

Thank you both. I knew it was something like that, but I didn't know
about modifiers in regex's; I only knew that '?' stood for 'once or not
at all'.

Reply to this Message

java pattern matcher

Free Magazines