On a side note, profiling shows similar issues w/ expensive growth on the
ArrayList class. The same datastructure would optimize that growth… but it’s a
far more complex implementation thanks to the random access and in-place
insertion/removal. In short, the opportunity is there but the tradeoffs are
worse.
John
From: core-libs-dev <core-libs-dev-r...@openjdk.org> On Behalf Of Markus KARG
Sent: Saturday, March 29, 2025 3:56 AM
To: core-libs-dev@openjdk.org
Subject: RE: [EXTERNAL] JDK-8352891 Performance improvements to
ByteArrayOutputStream
CAUTION: This email originated from outside of the organization. Do not click
links or open attachments unless you can confirm the sender and know the
content is safe.
Supporting what Alan said. Would love to see a static factory for a
non-synchronized byte-array backed OutputStream, like
`OutputStream::newByteArrayOutputStream`. John, if you like we can team up for
authoring this, I have free capacity.
Am 29.03.2025 um 09:33 schrieb Alan Bateman:
On 28/03/2025 13:05, Engebretson, John wrote:
Hi all! This message is to discuss the proposal for a public class that is
faster/cheaper than ByteArrayOutputStream. Details are on the ticket [1] so I
will only summarize here:
- ByteArrayOutputStream is slower than the provided alternative, and wastes
memory bandwidth and allocation.
- The new alternative cannot replace ByteArrayOutputStream because BAOS exposes
two implementation-specific fields.
- The problem is broadly present, and different solutions exist in Spring,
Tomcat, multiple applications inside my company, and undoubtedly elsewhere.
- There are places within the JDK that will benefit from the improved
performance.
There are many possible directions and APIs that could be explored here, lots
of fun.
The JBS issue seems to all about BAOS performance, esp. due to it maintaining a
contiguous array and needing to resize. Rather than introduce a subclass into
the API then maybe you instead explore adding a static factory that returns a
BAOS with a different implementation. For a long time we've been mulling over
adding a factory method to get a BAOS that isn't thread safe (the undocumented
synchronization dates from JDK 1.0 and too risky to change after 30 years of
usage). It would also be feasible to have BOAS use a different implementation
when used directly vs. when subclassed although that wouldn't be pretty.
-Alan
