Alexander Dunlap wrote:
For instances where an exception would be too intrusive, I don't see how it would be too hard to write a wrapper function
In a library that does not have direct access to the IO monad, it would be not just hard - it would be impossible. That is because of type restrictions in the current versions of catch, block, and friends. For example, if you are writing a library that exports something like: class MonadIO m => MyMonad m where... you can't block or catch exceptions, so you have to limit yourself as much as possible to operations that are unlikely to throw IO exceptions. Otherwise, you have to pass on responsibility to users of the library, which makes the library much less useful. That, in turn, would make the use of readline very awkward and complex - or in some cases even impossible - in programs based on monad transformers, one of the most beautiful and powerful idioms in Haskell. What a shame. (The other case is the one Ian mentioned, but as he points out, that could be considered a bug in the library.) On the other hand, it would never be too intrusive to write a wrapper providing an exception if that is really required. So that would be more general. The question becomes: given all of the other bindings in this readline library, is there any conceivable simple usage of readline that would usually not raise exceptions? If not, we should try to fix that. Otherwise, the situation is hopeless anyway, so then it would sadly not make any difference what we do with this particular binding. Regards, Yitz