Hi Nazario,
That's a typical substring leak case. Matcher.group(n) eventually calls
String.substring() to return the result. Before removing the count and
offset fields in String, that'll return a String sharing the value char
array, which may cause memory leaks if you hold on to a small substring of
a large string.
It's not your HashSet leaking memory, just the substring.
- Kris
On Fri, Apr 12, 2013 at 9:50 PM, wangsheng.0376 <wangsheng.0...@163.com>wrote:
> hi, all,
>
> I agree with you to remove offset, today when I run following code in
> jdk7(sorry I forget the detail version), my code is like this:
>
>
>
>
> Pattern pattern = Pattern.compile(regex);
>
> Matcher matcher = pattern.match(content);
>
> while (matcher.find()) {
>
> String link = matcher.group(1);
>
> set.add(link);// set is a HashSet type
>
> }
>
>
>
>
> after I running for some time, I found that the 'set' object has memory
> leak, I found though 'link' is a part of 'content', but the value field in
> link variable is same as 'content' value field, difference is that the link
> offset field is not zero.
>
> So that mean the set(HashSet) is not store the 'link', it store the
> 'content' value field, it will cause the memory leak problem, So I am very
> happy to reomve the offset field.
>
>
>
>
> Best Regards
>
> =nazario.wang=
>
