Mybatis-3 3.6.0 Ideas

[hazendaz]

Starting a public discussion here for what users would like to see as part of Mybatis 3.6.0. No actual work has started at this time. Getting list together of ideas and would like community feedback as well as up-votes on any of the many pull requests that exist that might be better fit for 3.6.x.

Current Ideas

• Move from java 8 to java 11 baseline.
• Apply all code changes related to jdk 11 baseline
• Small portion of code is currently flagged jdk16 or better. Move that part to jdk17 or better since 16 is EOL.
• Include Improve Cursor Closeable behavior #1494
• Drop support for reload4j (aka log4j1)
• Possibly discuss dropping other legacy logging formats and maybe landing on slf4j proper (most of our modules already are there)
• Release more frequently
• Introduce separate project bom

Support for 3.5.x will continue where appropriate. We have recently released 3.5.11 and have few changes since and would expect to drop that as a release when 3.6.x lands if not before. For all the mybatis modules we own, we are actively releasing all of them aligned to mybatis 3.5.11 and trying to position for more frequent releases in order to get features in users hands more quickly than previously done. Some underlying modules have already went above jdk 8 such as the caffeine cache now at jdk 11 as caffeine moved and spring module at jdk 17 and spring boot starter at jdk 17 since spring moved. Where appropriate will we will align where those extensions require while continuing to support older versions where support is still maintained.

All ideas are welcome!

[GeorgeSalu]

• I opened an issue about aot support at least initially it would be very good Add support for Ahead-of-Time (AOT) compilation #2747
• change from cglib to ByteBuddy.
• change from ognl to commons-ognl ?.
• Support the syntax of "in (#{xxx})"
• Incorrect work with immutable objects Incorrect work with immutable objects #2207
• Suporte java.util.stream.Stream Support java.util.stream.Stream as return type of mapper method #1212
• remove methods marked as deprecated

[awxiaoxian2020]

I'm working on #2691 . I think it should be included in 3.6.x too.

[awxiaoxian2020]

#2080 should be solved in 3.6.x, I think.

[GeorgeSalu]

@hazendaz @harawata how about changing the use of reflection to annotation processor like mapstruct and micronaut?

[harawata]

Even if it's possible to change the entire mapping logic to annotation processing, it will be a huge design change, so it's not going to happen in 3.x.

[GeorgeSalu]

yes I am thinking of a 4.0 version would be a very good improvement

[harawata]

MyBatis mappers are largely dynamic, so I honestly am not sure if/how annotation processing fits in the design.
If you have a clear vision for the implementation or proof-of-concept, I would like to see it (as it's off-topic here, please create a new discussion for it).

[GeorgeSalu]

I don't have knowledge about implementation I suggested annotation processor by micronaut and mapstruct to use annotation processor and compared to their competitors that use reflection api they are much more performative and in general consume less resources

[kazuki43zoo]

I want to consider for supporting the R2DBC(#1444) on core module or integration module. However, It may be something that will be supported in 4.x instead of 3.6.x.

[harawata]

R2DBC is being discussed for version 4.

[GeorgeSalu]

@harawata @hazendaz with the release of jdk 21 wouldn't it be better for the next release to go up to jdk 17???? What do you think about releasing a final 3.5.14 and concerting efforts for a new version 4 with less initial work using annotation processor?? @harawata about the annotation processor I discovered this project https://github.com/mybatis-flex/mybatis-flex which uses annotation processor I think an initial work would be to try to replace reflation with the annotation processor generating the code just like mapstruct what do you think ???

[harawata]

That project does not seem to support dynamic SQL.

Take a look at the following statement.

select * from users
<if test="name != null">
  name = #{name}
</if>

How do you generate the Java code using annotation processing for the above statement? (that's what you are proposing, right?)
Note that even the type of the parameter object could vary at runtime because it could be stored in a java.util.Map.

I have to admit that I am not familiar with annotation processing, but to me, it seems unrealistic to develop MyBatis without reflection.

[GeorgeSalu]

Well, I also don't understand much about annotation processing, in fact I only found out about this possibility when I worked with mapstruct, well, since you say that you can't use annotation processing, then that's okay, I only raised this subject because I read an article on the Micronaut website where he compares the performance of micronaut data (it uses annotation processing in fact micronaut as a whole says that it uses a lot of annotation processing) with spring data jpa and micronauta data ends up having a much better performance

[GeorgeSalu]

@harawata @hazendaz what do you think of a solution for pagination in this new version???

[jeffgbutler]

How would you imagine a pagination solution that is different from what we already support?

[GeorgeSalu]

Is there a native solution for physical paging? I'm currently analyzing this plugin https://github.com/pagehelper/Mybatis-PageHelper

[jeffgbutler]

My understanding of PageHelper is that it uses an interceptor to parse SQL and insert the database native paging clauses. As far as I know, there is no cross database native way for doing paging - every database does it differently.

In MyBatis proper, we do not parse SQL and do not alter the SQL entered by a developer. So something like PageHelper is not a good fit for the core library - although I'm sure it works well in many use cases.

[GeorgeSalu]

What do you think about cache implementations becoming a separate module and mybatis core only dealing with SQL?????

[GeorgeSalu]

@harawata @hazendaz would it be time to think about a 3.6.X version?? We already have some proposed comits

#2760
#1494

[epochcoder]

I would like to add #101 here aswell, what do you think?

[GeorgeSalu]

It would be great to include this commit in the next release, it would complete the work with immutable objects

[cheehonglee-lazada]

Hi, parallel to the current enhancement to constructors to provide better support for immutability, is there any plan to support creating instance through builder?

eg.

public class Blog {
    private final int blogId;
    Blog(int blogId) {
        this.blogId = blogId;
    }

    public static BlogBuilder builder() {
        return new BlogBuilder();
    }

    public static class BlogBuilder {
        private int blogId;
        BlogBuilder() {
        }

        public BlogBuilder blogId(int blogId) {
            this.blogId = blogId;
            return this;
        }

        public Blog build() {
            return new Blog(this.blogId);
        }
    }
}

[harawata]

For a simple class like that, it should be possible with a custom object factory + constructor-mapping.
Let me know if you have any difficulty.

Note that constructor-mapping does not support collections yet.

[GeorgeSalu]

@harawata @hazendaz Why not think about a version 3.6 now? Version 3.5 is already 6 years old and at least 2 improvement committees have been inactive for a long time #2760 #1494, and we are already moving towards Java 25lts while the master is still at version 8.

[epochcoder]

I think that is the intention, at least based on @hazendaz comment here

[hazendaz]

Hi @epochcoder, I'll be able to get back to mybatis* stuff in next few weeks. At the moment, many of the subprojects where moved to jdk 11 or 17 depending on their usage while reviewing that code and trying to stabalize it on newer jvms. Mybatis core was just waiting until we did another java 8 release, it can move up now at any time as that was just done. I think just to give folks a chance on newer version we will cut a quick java 11 version at 3.6.0 then move to 3.7.0 for java 17. That way those that need newer built one with probably a good deal of code cleanup can get a newer copy depending on their specific situation. We can always come back and patch if there is demand but that at least will set a point to start from. I realize we have lots and lots of pull requests open as well that we need to address one way or another as well. If there is anything you want to start working on, I'd say go for it. If you need to bump up to java 11 for that, look at mybatis-cdi project pom to see the java entries to do so quicker.