"Terry Reedy" <tjre...@udel.edu> wrote in message
news:mailman.119.1273690025.32709.python-l...@python.org...
CPython compiles Python code (a sequence of statements) to its private
bytecode (a sequence of codes and operands) and then interprets the
bytecode. So 'edit and continue' would have to recompile the statement and
patch the result into the bytecode.
I believe this is what, e.g., Visual Studio does with, e.g., C++ code -- it
compiles the modified code and then patches the binary to call the new code as
a subroutine or somesuch.
Another problem is recovery of state. If an exception is raised in the
middle of a statememt, there would need to be a means to roll back the state
of objects and namespaces to what they were before execution of the
statement started. So the interpreter would need to act like a transactional
database, which is not trivial.
Yeah, I can see this being challenging. I think it's pretty cool how some
Python debuggers will allow you to use a shell with the context set to the
current function or any of the calling functions higher up the call stack
hierarchy as well.
This would also slow normal execution, so it would have to be optional.
That'd be a small price to pay. :-) (Visual Studio only lets you do this with
unoptimized code...)
If would definitely be nice. Perhaps $500,000 in funding would do the trick.
Cool, if I win the lottery and add it to the "worty causes" list! :-)
Thanks for the information,
---Joel
--
http://mail.python.org/mailman/listinfo/python-list
