Add bulk read method to InputStream implementations

[pstreef]

Problem

Java's default InputStream.read(byte[], int, int) implementation calls the single-byte read() in a loop, causing up to 350x slower performance for bulk reads. Many InputStream subclasses only override the single-byte read() method but not the bulk read method.

• See Add recipe to add bulk read method to InputStream implementations rewrite#6383 for more context and original discussion.

Solution

This recipe detects InputStream subclasses that:

• Override read() but not read(byte[], int, int)
• Delegate to another InputStream

For simple delegation patterns, it adds the missing bulk read method. For complex bodies (side effects, transformations), it adds a search marker for manual review.

Handles both anonymous and named classes, ternary and if-statement null check styles, and skips FilterInputStream subclasses.

• Moved from Add recipe to add bulk read method to InputStream implementations rewrite#6383 per review feedback from @timtebeek suggesting this belongs in rewrite-migrate-java's org.openrewrite.java.migrate.io package.

[timtebeek]

Nice one! Should help folks and fix performance problems they might not even know they have.

For now we've not added this recipe to any larger composite, meaning folks would need to seek out and run this recipe specifically; shall we keep it that way?

[timtebeek]

This required a small change upstream to allow the precondition to be used: openrewrite/rewrite@e91804a

[pstreef]

For now we've not added this recipe to any larger composite, meaning folks would need to seek out and run this recipe specifically; shall we keep it that way?

I dont really know what the practice here has been in the past. What do you advise?

[timtebeek]

I think it's fine to keep as is for now: especially since we can't automatically fix any case it might be best not to repeatedly warn about any such cases if folks decide not to follow up.