Which Java Bean Copy Library Is Fastest? A Comprehensive Benchmark

Introduction

Bean copy is widely used in development to dramatically reduce manual mapping work. This article conducts a stress test of common Bean copy tools to help developers select the most appropriate one. It continues the series started with "cglib bean copy introduction".

Bean Copy Tools

MapStruct (compile‑time generated Mapper implementation)

Selma (compile‑time generated Mapper implementation)

yangtu222 – BeanUtils (first‑time generating copy bytecode)

mica (first‑time generating copy bytecode)

hutool (reflection‑based)

Models

No Type Conversion

<code>/**
 * Source user
 */
@Data
public class FormUser {
    private Long id;
    private String nickName;
    private Integer age;
    private String phone;
    private String email;
    private String password;
    private Integer gender;
    private String avatar;
}

/**
 * Target user
 */
@Data
public class ToUser {
    private String nickName;
    private String phone;
    private String email;
    private Integer gender;
    private String avatar;
}</code>

With Type Conversion

<code>/**
 * Source user with conversion
 */
@Data
@Accessors(chain = true)
public class FormConvertUser {
    private Long id;
    private String nickName;
    private Integer age;
    private String phone;
    private String email;
    private String password;
    private Integer gender;
    private String avatar;
    @DateTimeFormat(pattern = DateUtil.PATTERN_DATETIME)
    private LocalDateTime birthday;
}

/**
 * Target user with conversion
 */
@Data
@Accessors(chain = true)
public class ToConvertUser {
    private String nickName;
    private Integer age;
    private String phone;
    private String email;
    private String password;
    private Integer gender;
    private String avatar;
    private String birthday;
}</code>

Bean Copy Benchmark Results

Environment

OS: macOS Mojave

CPU: 2.8 GHz Intel Core i5

RAM: 8 GB 1600 MHz DDR3

JVM: Oracle 1.8.0_201 64‑bit

Simple Model

Benchmark   Score       Error   Units
hutool      1939.092    26.747  ops/ms
spring      3569.035    39.607  ops/ms
cglib       9112.785    560.503 ops/ms
mica        17753.409   393.245 ops/ms
yangtu222   18201.997   119.189 ops/ms
cglibMapper 37679.510   3544.624 ops/ms
mapStruct   50328.045   529.707 ops/ms
selma       200859.561  2370.531 ops/ms

With Type Conversion (Date)

Benchmark   Score     Error   Units
mica        1186.375  64.686  ops/ms
mapStruct   1623.478  13.894  ops/ms
selma       160020.595 2570.747 ops/ms

List Model (100 items)

Benchmark   Score   Error   Units
spring      35.974  0.555   ops/ms
mica        169.066 5.460   ops/ms

Map to Bean Copy

Benchmark   Score    Error   Units
hutool      1338.551 16.746  ops/ms
mica        13577.056 27.795 ops/ms

Conclusion

The results differ slightly from those of the java‑object‑mapper‑benchmark project. Selma performed better than MapStruct in this test, possibly due to differences in the model definitions.

Feature Comparison

Selma – requires writing a Mapper, does not support Map or Collection mapping, needs manual type conversion, performance: extremely high.

MapStruct – requires a Mapper, no Map/Collection support, supports common types and complex expressions, performance: high.

BeanUtils (yangtu222) – no Mapper needed, no Map support, supports List/Set, manual type conversion, performance: high.

mica – no Mapper needed, supports Map and Collection, uses Spring type conversion, performance: high.

Spring – no Mapper needed, no Map/Collection support, no type conversion, performance: medium.

hutool – no Mapper needed, supports Map, no Collection support, no type conversion, performance: medium.

Open‑Source Recommendations

References

[1]

MapStruct – http://mapstruct.org/

[2]

Selma – http://www.selma-java.org/

[3]

yangtu222 – BeanUtils – https://github.com/yangtu222/BeanUtils

[4]

mica – https://github.com/lets-mica/mica

[5]

hutool – https://gitee.com/loolly/hutool

[6]

java‑object‑mapper‑benchmark – https://github.com/arey/java-object-mapper-benchmark