Add new method for propagating mapper info on LanguageDriver

[kazuki43zoo]

By this changes, scripting language plugin (mybatis-freemarker, mybatis-velocity and mybatis-thymelaf) can loaded a template file automatically by rule-based if SQL is not specify on XML or annotation.

for example:

package com.example.demo.mapper;

interface DemoMapper {
  @Insert("")
  @SelectKey(statement = "", ...)
  void insert(Demo demo);
}

The scripting language plugin (e.g. mybatis-thymeleaf) can be load a template files (e.g. com/example/demo/mapper/DemoMapper/insert.sql or com/example/demo/mapper/DemoMapper/insert_selectKey.sql) automatically.

WDYT?

[kazuki43zoo]

@harawata

WDYT? If you can approve this PR, I want to add feature that load a template file automatically at first release of mybatis-thymeleaf.

[harawata]

I think I understand what you want, but it does not seem right for LanguageDriver implementations to have a logic that depends on statement ID (i.e. LanguageDriver should not need statement ID).

FWIW, it would be possible to allow users to write some LanguageDriver-specific syntax.

interface DemoMapper {
  @Insert("file:example.sqlfiles.DemoMapper_Insert.sql")
  @SelectKey(statement = "file:example.sqlfiles.DemoMapper_SelectKey.sql", ...)
  void insert(Demo demo);
}

[kazuki43zoo]

@harawata thanks for your review comment!

I think I understand what you want, but it does not seem right for LanguageDriver implementations to have a logic that depends on statement ID

I understand your opinion.
However, I think that propagating SQL definition source information(e.g. method location, XML location and element information) as meta data is not wrong solution. WDYT?
If there is no room for consideration, I will withdraw this PR.

FWIW, it would be possible to allow users to write some LanguageDriver-specific syntax.

yes! it support already as follow:

@Select("/NameMapper/findById.sql")
Name findById(@Param("id") Integer id);

[harawata]

Passing information about SQL source breaks separation of concerns, IMHO.
Those informations are used to load a SQL script from the source and that is not a LanguageDriver's task.
LanguageDriver's task is to process an already loaded SQL script and I think it's better to keep this separation.

To load SQL based on a mapper method, we already have SQL provider.
So, this might be related to #1391 .

[kazuki43zoo]

@harawata Thanks for your comment.

I understand and agree with your opinion.