[
https://issues.apache.org/jira/browse/ARTEMIS-5121?focusedWorklogId=972470&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-972470
]
ASF GitHub Bot logged work on ARTEMIS-5121:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 10/Jun/25 18:37
Start Date: 10/Jun/25 18:37
Worklog Time Spent: 10m
Work Description: jbertram commented on code in PR #5757:
URL: https://github.com/apache/activemq-artemis/pull/5757#discussion_r2138551748
##########
artemis-protocols/artemis-mqtt-protocol/src/main/java/org/apache/activemq/artemis/core/protocol/mqtt/MQTTSessionState.java:
##########
@@ -587,8 +588,8 @@ public Integer getMatchingId(String topic) {
private void update(MqttTopicSubscription newSub, Integer newId) {
if (newId != null && !newId.equals(id)) {
- if (this.topicFilterPattern == null ||
!subscription.topicFilter().equals(newSub.topicFilter())) {
- topicFilterPattern = Match.createPattern(newSub.topicFilter(),
MQTTUtil.MQTT_WILDCARD, true);
+ if (this.address == null ||
!subscription.topicFilter().equals(newSub.topicFilter())) {
+ address = new
AddressImpl(SimpleString.of(newSub.topicFilter()), MQTTUtil.MQTT_WILDCARD);
Review Comment:
This particular `AddressImpl` _is_ re-used between multiple invocations of
`matches`. Were you referring to this line in `getMatchingId`:
```java
if (id != null && new AddressImpl(SimpleString.of(topic),
MQTTUtil.MQTT_WILDCARD).matches(address)) {
```
I wasn't keen on creating a new instance of `AddressImpl` here for every
invocation of `matches`, but the constructor performs a `split` that would be
necessary either way so I thought it made some sense to just re-use that. Is
your concern here GC pressure or something else?
In any case, I'll look at creating a `static` version of `matches` that will
fit with this as well as the existing use-cases.
Issue Time Tracking
-------------------
Worklog Id: (was: 972470)
Time Spent: 2h 50m (was: 2h 40m)
> Improve performance of MQTT subscription ID matching
> ----------------------------------------------------
>
> Key: ARTEMIS-5121
> URL: https://issues.apache.org/jira/browse/ARTEMIS-5121
> Project: ActiveMQ Artemis
> Issue Type: Improvement
> Components: MQTT
> Affects Versions: 2.37.0
> Reporter: Adam Zyzak
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.42.0
>
> Time Spent: 2h 50m
> Remaining Estimate: 0h
>
> Recently I performed stress test of a system using Artemis as a message
> broker for MQTT and STOMP protocols. I noticed huge difference in performance
> and CPU usage between these protocols.
> I made some jstacks and it shown that majority of time is spent in method
> {{o.a.a.a.c.p.m.MQTTSessionState#getMatchingSubscriptionIdentifiers}}
> compiling regular expressions.
> It would be helpful if performance of this method was improved, especially
> since it is executed on every publish, either by caching compiled patterns,
> adding special cases for simpler topic names (topic names without special
> characters may be compared using equality, {{#}} subscriptions can be always
> added and topic names having {{#}} at the end can be compared using
> {{startsWith()}}) or just adding option to configuration allowing to skip
> adding SUBSCRIPTION_IDENTIFIER to publish message.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact
