`returnInstanceForEmptyRow` does not work for constructor auto-mapping

[cvmocanu]

MyBatis version

3.5.10

Database vendor and version

PostgreSQL 11.6 (Debian 11.6-1.pgdg90+1) on x86_64-pc-linux-gnu, compiled by gcc (Debian 6.3.0-18+deb9u1) 6.3.0 20170516, 64-bit

Test case or example project

mybatis-bug.zip

Steps to reproduce

You need to have Java, Docker, and Docker-Compose installed.
From the root of the project:

• docker-compose up -d
• ./gradlew run

Expected result

The project prints ALL OK

Actual result

The project throws an exception got null row because all columns are null, even though returnInstanceForEmptyRow is true.

[harawata]

Hello @cvmocanu ,

Thanks for the report & repro!

I could verify a problem (i.e. returnInstanceForEmptyRow is not correctly applied when MyBatis performs constructor-automapping), so I am going to fix it.
The attached test, however, kind of misses the point and I would like to clarify that.

As seen in the log, the query in the test does not return any row.

... PersonMapper.getById - ==>  Preparing: select name from person where id=?
... PersonMapper.getById - ==> Parameters: -1(Long)
... PersonMapper.getById - <==      Total: 0

When there is no row in the result set, MyBatis returns null even if returnInstanceForEmptyRow is enabled.
"empty row" means that there is a row in the result set and all mapped columns are NULL.
The option returnInstanceForEmptyRow was added to differentiate these two cases.

[harawata]

@cvmocanu ,

I have committed the fix.

As I explained, the test you attached still fails.
Please let me know if you have any question.

Thanks again for your contribution!

[cvmocanu]

@harawata, thanks so much.

I've fixed the example, in case it helps:
mybatis-bug-good-example.zip

Do you have an idea when the fix will be released?