alternative fix bug #921 Duplicity in package-info.java when templates are used

[pvojtechovsky]

@monperrus Hi Martin, here is the alternative PR for #932 as you suggested.

It is variant, which uses non abstract JDTBatchCompiler. FactoryCompiler and FileCompiler does not inherit from JDTBatchCompiler and they were renamed to FactoryCompilerConfig and FileCompilerConfig, because they are now used to configure the JDTBatchCompiler.

CompilationTest file and folder filtering tests are commented for now. There is no support for filtering by overwriting of createBatchCompiler. I suggest to discuss how to implement it later.

ExtendedStringLiteralTest was updated. It should be OK.

InputType was changed from enum to interface. The change is compiler compatible so source code of the clients does not have to be changed. Just recompiled. Implementation of the InputType can configure JDTBatchCompiler before the compilation is started.

	interface InputType {
		InputType FILES = FileCompilerConfig.INSTANCE;
		InputType CTTYPES = FactoryCompilerConfig.INSTANCE;

		void initializeCompiler(JDTBatchCompiler compiler);
	}

So we have standard defaults, but now they can extend and use different configurations if needed. I use it internally instead of InputType.UNIT, I am using

... createBatchCompiler(new FileCompilerConfig(sources));

[monperrus]

It is a good and creative solution. And again mostly backward compatible, congrats. See my two comments.

[monperrus]

add javadoc:
/** responsible for setting the parameters of JDTBatchCompiler, must call setCompilationUnits() */

[monperrus]

I propose that this method moves to FileCompilerConfig. And FileCompilerConfig would be responsible for calling setCompilationUnits().

[tdurieux]

It is better to remove instead of commenting.

[pvojtechovsky]

I have no problem to remove it, but I am not author. So @monperrus should agree that we will really get rid of these tests without any replacement.

[monperrus]

since your architecture is mostly backward compatible, it seems that only minor changes will enable us to keep those tests passing. can you give it a try?

[pvojtechovsky]

I will give it a try, but I would prefer if you merge my changes first, because then I can create another PRs too, related to compiler this evening. Otherwise I can fix only test and cannot do more until it is merged. Because of conflicts ...

[pvojtechovsky]

I have added comment and moved the method as suggested. I have also rebased to master