john.r.rose at oracle.com
Tue Jul 18 02:12:27 UTC 2017
I have felt for some time that the BoundMethodHandle code
is not easy to work with, and I have wanted to refactor it
for a long time to separate out the specialization logic
from the BMH logic proper.
Here is an experimental refactoring that accomplishes this:
I'd like to commit it to JDK10 first as a cleanup,
but the motivation for it comes from Valhalla and Amber.
If you work on java.lang.invoke code, please take a look
and tell me what you think.
Caveat: It basically passes smoke tests only, but it
does pass its own regression test (which builds a toy
specialization schema, including nominal types the
could be replaced by value types).
More information about the valhalla-dev