[mercury-users] Problem installing with grade java

Holger Krug hkrug at rationalizer.com
Wed Jul 25 16:42:20 AEST 2001


On Tue, Jul 24, 2001 at 12:35:08PM +0200, Tyson Dowd wrote:
> What I like about Holger's idea is that the technical implementation can
> be much cleaner that what we have now.  Well, at least the way he has
> talked about it has made me think of an interesting implemetation, but
> I'm not sure if this is what he is suggesting.

Your proposal is very similar to what I suggested. The only difference
is, that I would not use any pragmas at all to call foreign language
code from Mercury. Admittedly it would me more difficult as I
originally proposed, because at least such hints like
`will_not_call_mercury' and `thread_safe' have to be given to the
Mercury compiler.

> For each Mercury modules that compiles into {IL, Java} create a base
> clase with some abstract methods for foreign language code.  The base
> class is called (say) $modulename_base.

Yes, that's part of my proposal. But now I'm very sceptical about
it. The simple reason is, that the {IL, Java} method would be
"static". Now try to compile "static abstract" methods in Java.  The
Java compiler has to be very polite not to blame you. My fault!  Maybe
you thought about something else. Would be nice to hear about.

Nevertheless the essence of my proposal:
* getting rid of foreign code in the Mercury sources
* collecting Java code from two sources: one generated by the Mercury
  compiler and the other implemented manually (maybe based on stubs
  generated by the Mercury compiler)
could help *a bit/lot*. 

I could give you more information if necessary.

-- 
Holger Krug
hkrug at rationalizer.com
--------------------------------------------------------------------------
mercury-users mailing list
post:  mercury-users at cs.mu.oz.au
administrative address: owner-mercury-users at cs.mu.oz.au
unsubscribe: Address: mercury-users-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-users-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------



More information about the users mailing list