ExecutorException: No setter found for the keyProperty 'id'

[MikeMitterer]

MyBatis version

3.4.2

Database vendor and version

PostgreSQL 9.6.1

Test case or example project

interface TicketMapper {
    @Options(flushCache = Options.FlushCachePolicy.TRUE, useGeneratedKeys=true) // Set keyProperty as the Java variable name and keyColumn as the column name in the database.
    @Insert("INSERT INTO tickets (apikey, ticket) VALUES (#{apikey},#{ticket})")
    public void insertTicket(final TicketTO ticket) throws RecordAlreadyExists;
   }

-- Tickets für die WebSocket-Anmeldung
CREATE TABLE tickets (
  ID SERIAL PRIMARY KEY,
  crdate    TIMESTAMP DEFAULT LOCALTIMESTAMP,

	-- apikey - ist in diesem Fall ein WebToken
	apikey     VARCHAR(2048)	not null,

  -- das eigentliche Ticket, eine UUID
  ticket    VARCHAR(36)	not null
);

public class TicketTO {

    /** Verweis auf den User */
    private          String   apikey;
    private          String   ticket;
    private DateTime crdate;

    public String getAppID() {  ...  }

    public String getTicket() { ...}

    public void setTicket(final String ticket) { ... }

    public DateTime getCrdate() { ... }

    public void setCrdate(final String crdate) { ... }
}

This line produces to above error:

getSession().getMapper(TicketMapper.class).insertTicket(ticket);

It's true that TicketTO has no "id" but this should be valid!
It works with 3.4.1 but fails with 3.4.2

[harawata]

Could you please create a demo project including the configuration so that we can reproduce the issue quickly and reliably?
Here are some skeleton projects : https://github.com/harawata/mybatis-issues
Thank you!

[kazuki43zoo]

Hi @MikeMitterer,

Probably, This issue related with #782.
I think this behavior is work as designed on 3.4.2.
If you want to avoid this behavior, I think useGeneratedKeys should be set to false (or omit this option).

I have a one question. Why did you set useGeneratedKeys=true in this case ?

[harawata]

#782 explains it, it seems.
Please let us know if you think we are missing something!

[harawata]

We've got another report on the mailing list.
https://groups.google.com/d/msg/mybatis-user/H940lW-qlXQ/K3MAEL7mCQAJ
#782 might cause problem on PostgreSQL.
Reopen to reevaluate.

[harawata]

I didn't know that, but there is a default value for keyProperty when using annotation (i.e. @Insert or @Update) and this causes a problem when useGeneratedKeys is enabled in the global config and the database is PostgreSQL [1].
Here are failing tests demonstrating the issue (it uses postgresql-embedded for portability).

Although it is not ideal, we have to revert #782 for now, I think.
The right fix would be to remove the default value for keyProperty, but we must wait for the next major release because it could break backward compatibility.
Does anyone have a better idea?

[1] In PostgreSQL, ResultSetMetaData#getColumnCount() always returns the number of the columns of the inserted/updated even when there is no SERIAL column in the table.

[harawata]

This should be fixed in the latest 3.4.5-SNAPSHOT .
It would be great if you could try it and see if your problem is resolved.
Thanks!