Richard Fuchs writes:
Understood. I've originally thought it was simply a bug because it didn't do what I'd expected, but I guess it's an old design decision. Thanks for clarification. Maybe for 3.4 then :)
i don't see any problem with this design decision because all results are obtainable with it. it does not matter to me if it is compatible with sed/perl or whatever. if it matters to you, for 3.4 you are welcome to introduce another mode of repl_exp behavior as long as the current one remains the default.
-- juha