Closed stephenjudkins closed 1 month ago
@stephenjudkins Thanks. I am sort of working on this here - https://github.com/sbt/zinc/pull/712
This is wonderful, I'm glad to know it's being worked on.
@stephenjudkins I now have tests passing with in-memory source code. I don't know how difficult it would be to abstract the JAR files, but I feel like I should just allow JAR files to be.
I assume Zinc's current virtual file system counts as "Allow operation without a filesystem". If Zinc's current implementation is not good enough, we can revisit the issue later.
Currently, input and output files are specified in the API in terms of
java.io.File
. The underlying Scala compiler, however, does not require interaction with the filesystem. If inputs and outputs could be handled through in-memory data structures it could open the door for some optimizations in some cases (in this case, in the https://github.com/higherkindness/rules_scala project).I'm aware this is a large issue, but figured I'd open it to signal support.