doanduyhai
commented
3 years ago Thanks for the PR @bric3
Unfortunately, Achilles processor cannot be incremental for several reasons:
1) It does not use the Filer API. Instead it is using com.squareup.javapoet.JavaFile to generate the source classes: https://github.com/doanduyhai/Achilles/blob/master/achilles-core/src/main/java/info/archinnov/achilles/internals/apt/processors/meta/AchillesProcessor.java#L118-L162
2) It does relies on internal Sun compiler classes to parse nested structures like Map<String, @Frozen MyUDT>: https://github.com/doanduyhai/Achilles/blob/b85a7f9b5e6415000787b833cdb3aa85c01ff660/achilles-core/src/main/java/info/archinnov/achilles/internals/parser/AnnotationTree.java#L274-L300
At that time, this was the only solution to parse nested annotations inside generic types. I don't know if since then things have improved on the front of Java annotation processor to avoid using internal classes. If so I'll be happy to refactor this part of the code to get rid of the Sun internal classes
bric3
Hi @doanduyhai
I'd liek to propose the following enhancement: Gradle supports incremental annotation processing, meaning the annotation processor won't have to run on every new compilation.
For that there is a simple guide to follow there to make sure the annotation processor is eligible to incremental annotation processing: https://docs.gradle.org/current/userguide/java_plugin.html#sec:incremental_annotation_processing
If I'm not mistaken achilles core is ok with these points :
[x] They must generate their files using the Filer API. Writing files any other way will result in silent failures later on, as these files won’t be cleaned up correctly. If your processor does this, it cannot be incremental.
[x] They must not depend on compiler-specific APIs like com.sun.source.util.Trees. Gradle wraps the processing APIs, so attempts to cast to compiler-specific types will fail. If your processor does this, it cannot be incremental, unless you have some fallback mechanism.
[x] If they use Filer#createResource, the location argument must be one of these values from StandardLocation: CLASS_OUTPUT, SOURCE_OUTPUT, or NATIVE_HEADER_OUTPUT. Any other argument will disable incremental processing.
I believe this annotation processor falls aggregatig category as it reaches to other elements like codecs or compiler registry. I followed the specific constraints :
"Aggregating" processors have the following limitations:
The changes looks to be minimal.
If possible it would be great to backport this change to the 5.3.x line.
Thank you in advance for the consideration.