[
https://issues.apache.org/jira/browse/CASSANDRA-16901?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17412150#comment-17412150
]
Tatu Saloranta commented on CASSANDRA-16901:
--------------------------------------------
No, I don't have cluster benchmarks unfortunately. I doubt effect would be big
enough to show on latencies; might be visible on heap dumps wrt garbage
(assuming `QualifiedName.to[CQL]String()` is heavily used, maybe it isn't).
My thinking is more along the lines of this being incremental improvement which
would not add complexity (a simple helper method). But I agree that ideally it
would show some measurable improvement.
> Optimize heavily-used `String.format()`-with-constant formatter use cases
> -------------------------------------------------------------------------
>
> Key: CASSANDRA-16901
> URL: https://issues.apache.org/jira/browse/CASSANDRA-16901
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Tatu Saloranta
> Priority: Normal
>
> For background, I did some benchmarking to compare performance of two simple
> ways to concatenate 2 Strings with a separator (used in f.ex
> `QualifiedName.java` for "keyspace:name" concatenation):
> # String.format("%s.%s", keyspace, name);
> # new StringBuilder().append(keyspace).append('.').append(name).toString()
> Difference is somewhat significant (for me, factor of 40x), as per:
> https://cowtowncoder.medium.com/measuring-performance-of-java-string-format-or-lack-thereof-2e1c6a13362c
> and so it seems worthwhile replacing at least some of usage from Cassandra.
> With a quick look it seems that classes like:
> * QualifiedName.java, FieldSelector.java, Selectable.java (in cql3)
> could benefit from this. And the second implementation can be packaged as a
> simple utility method (concatenateStringsWithChar?) so that code should be as
> readable as before.
> I can provide a patch, or if anyone else wants to have a go, feel free to.
>
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
