RFR(s): 8176894 Provide specialized implementation for default methods putIfAbsent, computeIfAbsent, computeIfPresent, compute, merge in TreeMap

Tagir Valeev amaembo at gmail.com
Sun Oct 13 09:51:44 UTC 2019


Please review the updated patch (no sponsorship is necessary; review only):

The difference from the previous version [1] is minimal: I just
noticed that the specification for computeIfAbsent should say
"mapping" instead of "remapping", so I fixed the spec. No changes in

Also please review the associated CSR:
I updated it, according to Joe Darcy's comments.

An additional explanation and background is copied below from my
original e-mail [2] for your convenience:

The patch was originally submitted by Sergey Kuksenko in March 2017 and
reviewed by some people:
The latest patch submitted by Sergey is here:

I asked Sergey why it was abandoned. Seems there were no particular reason
and Sergey asked if I could pick up this work. I will be happy to finish it.

Here's the list of differences between the latest Sergey patch and my patch:
- A bug is fixed in putIfAbsent implementation. TreeMap.java, lines 576 and
596:  `|| oldValue == null` condition added (the null value should be
overwritten no matter what)
- A testcase is added to cover this problem (InPlaceOpsCollisions.java,
also checks HashMap and LinkedHashMap). Not sure if it's the best place for
such test though. Probably a separate file would be better?
- TreeMap.merge() implementation is added.
- TreeMap is added to the FunctionalCMEs.java test (btw this file copyright
says that it's a subject to Classpath exception which is probably wrong?)
- Simple microbenchmark is added: TreeMapUpdate.java

My quick benchmarking shows that optimized version is indeed faster for the
most of the tests and no regression is observed for put() method. Here's
raw results of jdk13-ea+26 and jdk13-ea+26+patch if anybody is interested.

With best regards,
Tagir Valeev.

[1] https://cr.openjdk.java.net/~tvaleev/webrev/8176894/r2/
[2] https://mail.openjdk.java.net/pipermail/core-libs-dev/2019-July/061309.html

More information about the core-libs-dev mailing list