Skip to content

compression.md

Latest commit

 

History

History
132 lines (94 loc) · 5.77 KB

compression.md

File metadata and controls

132 lines (94 loc) · 5.77 KB
Raw
title Compression
sidebar_position 10
id compression
license Licensed to the Apache Software Foundation (ASF) under one or more contributor license agreements. See the NOTICE file distributed with this work for additional information regarding copyright ownership. The ASF licenses this file to You under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0 Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

This page covers compression options for reducing serialized data size.

Integer Compression

ForyBuilder#withIntCompressed/ForyBuilder#withLongCompressed can be used to compress int/long for smaller size. Normally compressing int is enough.

Both compression options are enabled by default. If the serialized size is not important (for example, you use FlatBuffers for serialization before, which doesn't compress anything), then you should disable compression. If your data are all numbers, the compression may bring 80% performance regression.

Int Compression

For int compression, Fory uses 1~5 bytes for encoding. The first bit in every byte indicates whether there is a next byte. If the first bit is set, then the next byte will be read until the first bit of the next byte is unset.

Long Compression

For long compression, Fory supports two encodings:

SLI (Small Long as Int) Encoding (Default)

  • If long is in [-1073741824, 1073741823], encode as 4 bytes int: | little-endian: ((int) value) << 1 |
  • Otherwise write as 9 bytes: | 0b1 | little-endian 8bytes long |

PVL (Progressive Variable-length Long) Encoding

  • First bit in every byte indicates whether there is a next byte. If first bit is set, then next byte will be read until first bit of next byte is unset.
  • Negative numbers will be converted to positive numbers by (v << 1) ^ (v >> 63) to reduce cost of small negative numbers.

If a number is of long type but can't be represented by smaller bytes mostly, the compression won't get good enough results—not worthy compared to performance cost. Maybe you should try to disable long compression if you find it didn't bring much space savings.

Array Compression

Fory can compress primitive arrays (int[] and long[]) when every value fits in a narrower primitive type. JDK 8 through 15 use scalar range analysis. On JDK 16 and later, the multi-release fory-core JAR automatically selects the Vector API implementation.

How Array Compression Works

Array compression analyzes arrays to determine if values can be stored using fewer bytes:

  • int[] → byte[]: When all values are in range [-128, 127] (75% size reduction)
  • int[] → short[]: When all values are in range [-32768, 32767] (50% size reduction)
  • long[] → int[]: When all values fit in integer range (50% size reduction)

Configuration and Registration

To enable array compression, you must explicitly register the serializers:

Fory fory = Fory.builder()
  .withXlang(false)
  // Enable int array compression
  .withIntArrayCompressed(true)
  // Enable long array compression
  .withLongArrayCompressed(true)
  .build();

// You must explicitly register compressed array serializers
CompressedArraySerializers.registerSerializers(fory);

Compressed array serializers are included in fory-core. When running on JDK 16 or later, resolve the incubator Vector API module when starting the application:

java --add-modules=jdk.incubator.vector ...

No registration or configuration change is required when moving between the scalar and Vector implementations. They make the same compression decisions and use the same serialized format.

String Compression

String compression can be enabled via ForyBuilder#withStringCompressed(true). This is disabled by default.

Configuration Summary

Option Description Default
compressInt Enable int compression true
compressLong Enable long compression true
compressIntArray Enable int array width compression false
compressLongArray Enable long array width compression false
compressString Enable string compression false

Performance Considerations

  1. Disable compression for numeric-heavy data: If your data is mostly numbers, compression overhead may not be worth it
  2. Array compression implementation is JDK-specific: JDK 8 through 15 use scalar range analysis; JDK 16 and later automatically select the Vector API implementation
  3. Long compression may not help large values: If most longs can't fit in smaller representations, disable it
  4. String compression has overhead: Only enable if strings are highly compressible

Example Configuration

// For mostly numeric data - disable compression
Fory fory = Fory.builder()
  .withXlang(false)
  .withIntCompressed(false)
  .withLongCompressed(false)
  .build();

// For mixed data with arrays - enable array compression
Fory fory = Fory.builder()
  .withXlang(false)
  .withIntCompressed(true)
  .withLongCompressed(true)
  .withIntArrayCompressed(true)
  .withLongArrayCompressed(true)
  .build();
CompressedArraySerializers.registerSerializers(fory);

Related Topics