Ludovic Courtès writes: > Mathieu Lirzin <m...@gnu.org> skribis: > >> One way to not break backward compatibility, would be to move (ice-9 q) >> code in (ice-9 queue) and make (ice-9 q) use it and re-export every >> variable with its old name. > > Right. > >> Finally we will need a way to give the information that (ice-9 q) module >> is deprectated to the users. I am not sure what is the best way to >> achieve this. Maybe with: >> >> (eval-when (expand load eval) >> (display "..." (current-warning-port))) >> >> What do others think? > > We could do a quick hack to iterate over the exported bindings, > re-export them, and print a message. > > However, I think it would be best to have a mechanism in the module > abstraction to /declare/ renames. That would allow both the compiler > and the run-time support to emit an appropriate warning. > > This would require something like adding a special field to ‘module’ > records, such that one can write, say: > > (define-module (ice-9 q) > #:superseded-by (ice-9 queue)) > > With that in place, it will become much easier to discuss about ice-9 > renames, IMO. > > Thoughts? > > Thanks, > Ludo’.
I think that would be useful. There's been talk about renaminig things from (ice-9) into a (guile) namespace. Likewise aliasing things from (srfi) to something more human-readable. Are the rename and aliasing thing the same step or two different things?
