[
https://issues.apache.org/jira/browse/GEODE-6410?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16777204#comment-16777204
]
Bill Burcham commented on GEODE-6410:
-------------------------------------
Some observations on the differences between {{computeIfAbsent()}} and
{{putIfAbsent()}}
Have a look at the two JDK tickets:
[https://bugs.openjdk.java.net/browse/JDK-8161372]
[https://bugs.openjdk.java.net/browse/JDK-6737839]
The first one (for {{computeIfAbsent()}}) is fixed in JDK11. The second one
(for {{putIfAbsent()}}) is not.
Initially I thought these two methods differed only in the way the value was
computed (or not). But the differences go deeper:
[https://stackoverflow.com/questions/48183999/what-is-the-difference-between-putifabsent-and-computeifabsent-in-java-8-map/48184207#48184207]
Difference #2 in that citation is:
!image-2019-02-25-10-38-24-493.png!
While {{computeIfAbsent}} can be viewed as a {{get()}} with a fallback value
(the result of applying the method reference), {{putIfAbsent()}} does not work
that way. {{putIfAbsent()}} is like a regular old {{get()}} that returns the
existing (old) value (or null if there was none) and then sets a fallback value
if there was no value before.
I suppose users might rightfully expect no write-locks in either case (if a
value was present before the call). But the implementations of the two are
pretty different.
This is what the naive fix to both methods would look like:
{code}
private static class MyMap<K,V> extends ConcurrentHashMap<K,V> {
@Override
public V computeIfAbsent(final K key, final Function<? super K, ? extends
V> mappingFunction) {
final V val = get(key);
if (null == val)
return super.computeIfAbsent(key, mappingFunction);
else
return val;
}
@Override
public V putIfAbsent(final K key, final V value) {
final V val = get(key);
if (null == val)
super.putIfAbsent(key, value);
return val;
}
}
{code}
> review use of putIfAbsent
> -------------------------
>
> Key: GEODE-6410
> URL: https://issues.apache.org/jira/browse/GEODE-6410
> Project: Geode
> Issue Type: Bug
> Components: general
> Reporter: Jason Huynh
> Assignee: Bill Burcham
> Priority: Major
> Attachments: image-2019-02-25-10-38-24-493.png
>
>
> Usages of putIfAbsent in Geode on ConcurrentHashMap may not have realized the
> actual synchronized/atomic nature of the putIfAbsent. See
> [https://bugs.openjdk.java.net/browse/JDK-6737839]
> This ticket is for someone to review or possibly change the putIfAbsent
> usages to be more performant.
> Not sure if putIfAbsent is called enough and under contention enough for this
> to matter...
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
