Use Unsafe memory buffer instead of byte buffer - https://github.com/imprint-serde/imprint-java/issues/21 by expanded-for-real · Pull Request #24 · imprint-serde/imprint-java

expanded-for-real · 2025-06-27T16:43:32Z

This was both easier and harder than I originally expected but finally got their. Uses the Unsafe memory buffer largely borrowed from Agrona/Fury to avoid things like boundary checking. This seems standard in other serialization libraries so was ultimately worth it.

Some to-dos - really need to consolidate some of the code better. There's a lot of duplication especially in how we parse and serialize headers.

Also this should pretty much be the end branch that resulted from constant rebasing so future MRs should look a little tidier.

…mance tracking; add comprehensive String benchmark

…to remove a bunch of stuff

…ing deserialization a bit faster

Try to enhance string deserialization

A full list of enhancements can be found here - #3

…ome micro-optimizations added that were found along the way

* Full comprehensive comparison tests with a lot of other libraries + some micro-optimizations added that were found along the way * replace deprecated gradle methods with latest --------- Co-authored-by: expand3d <>

# Conflicts: # src/jmh/java/com/imprint/benchmark/ComparisonBenchmark.java # src/main/java/com/imprint/core/ImprintRecord.java # src/main/java/com/imprint/types/TypeHandler.java # src/main/java/com/imprint/types/Value.java

…patch overhead

# Conflicts: # build.gradle # src/jmh/java/com/imprint/benchmark/ComparisonBenchmark.java # src/jmh/java/com/imprint/benchmark/ImprintDetailedBenchmark.java # src/jmh/java/com/imprint/benchmark/serializers/AvroSerializingBenchmark.java # src/jmh/java/com/imprint/benchmark/serializers/ImprintSerializingBenchmark.java # src/main/java/com/imprint/core/ImprintFieldObjectMap.java # src/main/java/com/imprint/core/ImprintRecord.java # src/main/java/com/imprint/core/ImprintRecordBuilder.java # src/main/java/com/imprint/ops/ImprintOperations.java # src/main/java/com/imprint/types/ImprintDeserializers.java # src/main/java/com/imprint/types/ImprintSerializers.java # src/main/java/com/imprint/types/MapKey.java # src/main/java/com/imprint/types/TypeCode.java # src/main/java/com/imprint/util/VarInt.java # src/test/java/com/imprint/ops/ImprintOperationsTest.java # src/test/java/com/imprint/profile/ProfilerTest.java

expanded-for-real · 2025-06-27T16:45:06Z

build.gradle

    jmhImplementation 'org.msgpack:jackson-dataformat-msgpack:0.9.8'
    jmhImplementation 'org.apache.thrift:libthrift:0.19.0'
    jmhImplementation 'javax.annotation:javax.annotation-api:1.3.2'
+    jmhImplementation 'net.openhft:chronicle-wire:2.25ea5'


Got frustrated halfway through the effort and decided to add ChronicleWire as a comparison

expanded-for-real · 2025-06-27T16:45:19Z

src/jmh/java/com/imprint/benchmark/ComparisonBenchmark.java

-@Fork(value = 1, jvmArgs = {"-Xms4g", "-Xmx4g"})
+@Measurement(iterations = 10, time = 1)
+@Fork(value = 1, jvmArgs = {"-Xms4g", "-Xmx4g", 
+    "--illegal-access=permit",


ChronicleWire apparently needs all this

expanded-for-real · 2025-06-27T16:46:15Z

src/main/java/com/imprint/core/ImprintFieldObjectMap.java


 /**
 * Specialized short→object map optimized for ImprintRecordBuilder field IDs.
- * Basically a copy of EclipseCollections's primitive map:


Need to clean up javadocs everywhere still

expanded-for-real · 2025-06-27T16:46:55Z

src/main/java/com/imprint/core/ImprintRecord.java

 @ToString(of = {"header"})
 public class ImprintRecord {
-    ByteBuffer serializedBytes;
+    ImprintBuffer serializedBytes;


Fortunately I was able to get the unsafe buffer to just be a drop in replacement for ByteBuffer so there's no functional changes here in the ImprintRecord

expanded-for-real · 2025-06-27T16:48:50Z

src/main/java/com/imprint/core/ImprintRecordBuilder.java

        return this;
    }

-    // Builder utilities


expanded-for-real · 2025-06-27T16:49:41Z

src/main/java/com/imprint/core/ImprintRecordBuilder.java

-        // 2. Sort fields by ID for directory ordering (zero allocation)
+        // 2. Sort fields by ID for directory ordering
        var sortedFieldsResult = getSortedFieldsResult();
-        var sortedValues = sortedFieldsResult.values;


This whole block I removed was hacky anyways but also unneeded - the ImprintBuffer is auto-expandable now and handles capacity increases so we don't have to worry too much about it

expanded-for-real · 2025-06-27T16:51:50Z

src/main/java/com/imprint/ops/ImprintOperations.java

@@ -4,375 +4,393 @@
 import com.imprint.core.*;


This was difficult but algoritmically everything is the same. I also decided to make merge and project their own static sub-class and separate out some other stuff to Core mostly as an effort to try to organize things (I still need to clean this up eventually since there's a lot of duplication). As I said though - algorithmically nothing has changed here

expanded-for-real · 2025-06-27T16:53:52Z

src/main/java/com/imprint/types/ImprintSerializers.java

@@ -2,11 +2,14 @@



No change here really - just drop in replacement

expanded-for-real · 2025-06-27T16:54:33Z

src/main/java/com/imprint/types/ImprintSerializers.java


-    // Rough size estimate since actual takes time; might be able to accomodate this better with a growable buffer though
+
    public static int estimateSize(TypeCode typeCode, Object value) {


Debating if I even need this size estimate now since the ImprintBuffer is auto-growable. Can just remove it later if I feel like it

expanded-for-real · 2025-06-27T16:55:14Z

src/main/java/com/imprint/util/ImprintBuffer.java

@@ -0,0 +1,605 @@
+package com.imprint.util;


Mostly borrowed concepts from Agrona and Fury for this

expanded-for-real · 2025-06-27T18:26:54Z

@agavra latest benchmark numbers are here - #21

Serialization is all over the place, like +/- a whole microsecond sometimes. I'd have to run in a less noisy environment (like as in not my PC) to get more consistent numbers I think

agavra · 2025-07-08T17:18:32Z

@expanded-for-real sorry for the delay in reviews, pretty busy at work -- will take a look at this this week!

agavra

I looked at the production code (particularly the changes to ImprintRecordBuilder and the new ImprintBuffer) and the strategy looks good to me! Since the PR is a little big let me know if there's an area of the code you'd like me to focus a review on.

expanded-for-real and others added 30 commits June 1, 2025 13:23

Update .gitignore

5cb33c9

initial commit for imprint-java

63313e4

initial commit for imprint-java

dd4fdbc

Add GitHub Actions CI workflow for automated testing

bce1d13

Merge remote-tracking branch 'origin/dev' into dev

f5d90b5

Update GitHub Actions workflow to use upload-artifact@v4

72c468f

Add Gradle wrapper validation to CI workflow

468d682

Fix gitignore to include gradle-wrapper.jar for CI

cf05b13

Force add gradle-wrapper.jar to repository

d0d7983

Update wrapper validation action to v3

f2cdd1b

Fix Javadoc syntax errors and disable strict Javadoc checking

57c8249

Add JMH benchmark .bat and .sh for full suite benchmarking and perfor…

edb3057

…mance tracking; add comprehensive String benchmark

fix map serialization error in benchmark test and streamline ci file …

2853e3f

…to remove a bunch of stuff

Add execute permissions back for gradlew in CI

3a5a113

Add some more string based performance benchmarks and try to make str…

50a288b

…ing deserialization a bit faster

Merge pull request #2 from imprint-serde/faster-strings

ea1c4c4

Try to enhance string deserialization

second main commit to address initial commits

43cab28

A full list of enhancements can be found here - #3

additional cleanup to address concerns in #3

fdb8a56

minor style fixes

2e56688

minor style fixes again

9353388

minor style fixes on benchmark tests and supress unused

09d0377

minor reordering

6209bb1

Merge branch 'main' into dev

ace7c67

Full comprehensive comparison tests with a lot of other libraries + s…

4632e01

…ome micro-optimizations added that were found along the way

replace deprecated gradle methods with latest

3738861

Merge Comparisons into dev branch (#8)

12d2823

* Full comprehensive comparison tests with a lot of other libraries + some micro-optimizations added that were found along the way * replace deprecated gradle methods with latest --------- Co-authored-by: expand3d <>

Lazy load of directory and header data

f7a6e8e

Merge remote-tracking branch 'origin/main' into dev

2834dbb

# Conflicts: # src/jmh/java/com/imprint/benchmark/ComparisonBenchmark.java # src/main/java/com/imprint/core/ImprintRecord.java # src/main/java/com/imprint/types/TypeHandler.java # src/main/java/com/imprint/types/Value.java

minor cleanup

83ed961

minor cleanup

a605b65

expand3d added 12 commits June 13, 2025 12:35

make serilaizers static/final; begin to refactor to avoid virtual dis…

686a855

…patch overhead

add new Imprint specific benchmark

ffc7918

Remove Value and TypeHandler to significantly reduce dynamic dispatch

e562ac3

add static serializers

999f48c

Add Unsafe direct buffer wrapper class

3b0a0d1

Try to get merge to use the new direct/unsafe/growable imprint buffer

8c54405

update comments

1148079

convert operations to use expandable buffer

541eddd

finalize using unsafe buffer everywhere

0f7b52b

fix formatting

60aff65

fix comparison tests

8a8fbc5

expanded-for-real commented Jun 27, 2025

View reviewed changes

remove extra comments and debug outputs

d3369cf

expanded-for-real commented Jun 27, 2025

View reviewed changes

src/main/java/com/imprint/core/ImprintRecordBuilder.java

return this;

}

// Builder utilities

Copy link

Collaborator Author

expanded-for-real Jun 27, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

unused

expanded-for-real commented Jun 27, 2025

View reviewed changes

expanded-for-real requested a review from agavra June 27, 2025 18:25

expanded-for-real assigned agavra Jun 27, 2025

agavra approved these changes Jul 14, 2025

View reviewed changes

expanded-for-real merged commit bc6f6c4 into main Jul 14, 2025
3 checks passed

expanded-for-real deleted the custom-buffers branch July 14, 2025 23:35


		// Rough size estimate since actual takes time; might be able to accomodate this better with a growable buffer though

		public static int estimateSize(TypeCode typeCode, Object value) {

Conversation

expanded-for-real commented Jun 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

expanded-for-real Jun 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

expanded-for-real commented Jun 27, 2025

Uh oh!

agavra commented Jul 8, 2025

Uh oh!

agavra left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

expanded-for-real commented Jun 27, 2025 •

edited

Loading

expanded-for-real Jun 27, 2025 •

edited

Loading