socket library for ocaml
Robbert VanRenesse <rvr <at> cs.cornell.edu>
1996-06-07 17:49:59 GMT
I have made an initial distribution of a socket library. It is supported not
only under Unix, but, using Winsock, also under systems like Windows 95. I
have only compiled it under Unix, and compiled and tested it under WinNT 4.0,
but expect little difficulty. For WinNT, I used Visual C++ 4.1.
The socket module supports all socket operations, select(), and gettimeofday().
It's called Socket. For documentation, see the Unix documentation (but use
Socket. rather than Unix.). For many applications, it should make the Unix
library obsolete, resulting in greater portability among platforms. For
example, it should make a port of CamlTk to Windows pretty easy, for those
whom are inclined to try this.
You can get the socket module by anonymous ftp from
ftp.cs.cornell.edu/pub/isis/horus/ocaml/socket.tar[.gz]
Untar it in the otherlibs directory, and run "make" followed by "make install"
(under Windows: nmake -f Makefile.nt). You have to define HAS_SOCKETS in
config/s.h first.
Feedback is highly appreciated. My hope is that Xavier will include this
in the official ocaml distribution, and we know how he insists on quality!
I'm currently not supporting the entire Winsock API (particularly not the
Winsock 2 API), but will probably add this some time in the foreseeable
future.
For those who are interested, I have an initial port of Horus/ML over the
socket library up and running.
Robbert
(Continue reading)
I am somewhat perplexed by the passage in the manual on virtual methods (pp. 29-30). Let's start with the first paragraph on p.30: it says that int_comparable2 is not a subtype of int_comparable, even though the former inherits from the latter! Now for a person with an experience in a non-functional OO language (C++ or Eiffel or Sather) it seems _very strange_ that the subtype and inheritance (derivation) relations aren't the same. It even seems that not having such an overlap defeats the purpose of having OO features in the first place. How is the programmer to predict where a subtype relation actually exists? Now the manual goes on to say that this occurs because `the self type appears in contravariant position in the type of method leq'. First, am I right in decoding this as `the types of the argument of int_comparable#leq and int_comparable2#leq are in covariant relation, but they would have to be in contravariant relation for a subtyping relation to exist between int_comparable and int_comparable2'? Second, and this may be the heart of the issue, _WHY_ is type 'a -> bool inferred for leq, and not < x: unit -> int; .. > -> bool , as it surely would for a global function? If these are stupid questions, I apologize again. Maybe it's a sign that the OO features need more explanation in the manual, though. Best,
RSS Feed