[MLton] MLton.Exn.setTopLevelHandler

[Matthew Fluet]

Vesa Karvonen wrote:
> AFAICT, MLton doesn't currently expose the functionality to set the
> top-level handler to a user-defined function.

I believe that is correct.

> Is there some reason why functionality
> equivalent to Primitive.TopLevel.setHandler is not exposed?

I don't see why not.  I don't think it was clear why it would be useful. 
  Generally, wrapping a 'main' function with a "handle _ => ..." usually 
accomplishes the same thing.

> Below is a patch that exposes setHandler as
> MLton.Exn.setTopLevelHandler.  It wraps the handler with a function
> that exits.  The wrapping is not absolutely necessary, but probably
> makes it easier to use.  Any objections to applying this patch (with
> or without the wrapping)?

I think the patch looks fine.  There is one oddity -- 
MLton.Exn.topLevelHandler is equivalent to the default top-level 
handler, rather than the dynamically installed top-level handler.  I 
don't think that is bad, but just a curiosity.  We might rename 
MLton.Exn.topLevelHandler to something that doesn't suggest that it is 
the 'getter' for MLton.Exn.setTopLevelHandler.

I think it does make sense to wrap the installed handler with the 
additional "handle _ => ..."; if we allow arbitrary user code to run as 
the top-level handle, it may itself raise an exception.  And MLton 
doesn't behave particularly well when an exception is raised with no 
handler in scope.