John Meacham
Malcolm Wallace wrote:
There is also the issue that we might adopt the proposal to allow (and perhaps eventually, to require) type signatures on export lists.
All I have to say is "please, no" to the requiring part that is.
I don't seriously propose for haskell-prime that signatures should be required on exports. Just permitting them would be a large and useful step up already. But the argument about whether to specify a module interface completely and explicitly in the source file might arise again in the next standardisation process, where I can see the balance might tip more in favour of compulsion.
I totally think type signatures in export lists should be allowed optionally with the exact same meaning as if they were specified at the top level within the module.
At the moment, only one type signature is permitted per entity (except for FFI foreign exports, where the exported signature may be more specific than the definitional signature). My proposal is that the normal export list can have signatures /in addition to/ the definitional signature, but that if there are thus two signatures for an entity, they must be identical. This is to prevent the case where one modifies the definition of the function, making it more general, but forgets to modify the corresponding export. Regards, Malcolm