> Just thinking aloud here (as always ;-). Thoughts?
My only thought is the lack of a directive to allow the compiler to tell
you the file/line of any errors. I'm rusty on my C++, but I seem to
recall they had a #file directive, so that the post-processed file, when
compiled, could signal errors with regards to pre-processed source files.
It'd be strange to have one of your 15-line files, and have the compiler
report an error on line 38 (after expansion).
What you're really doing is defining a new syntax to generate valid
bytecodes, not unlike PythonJ (or whatever it's called).
Why not just avoid the javac step, and emit bytecodes from your new
syntax anyhoo. Sure, we miss out standard compiler optimizations,
which could suck, but for the simple Easy-To-Write-JavaBean example,
it may be sufficient. Or, maybe just glue a front-end onto the
IBM Jikes back-end, where all we do is some AST translation into
what the backend normally expects.
-bob
(I like to parse)
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
