POPX@vax.oxford.ac.uk (Jocelyn Paine) (06/11/91)
Newsgroups: comp.lang.prolog
Subject: Specifying types
Summary:
Expires:
Sender:
Reply-To: popx@vax.ox.ac.uk (Jocelyn Paine)
Followup-To:
Distribution: comp.lang.prolog
Organization: Experimental Psychology, Oxford University, UK.
Keywords:
Is there a generally accepted notation for specifying types in Prolog?
I realise of course that Prolog itself lacks compile-time type-checking,
but there do exist programs for inferring types and checking these
against annotations added to the code. Are any of these programs in wide
enough use to merit adopting their notation for types as a de facto
standard?
If so, are any of them in the public domain? I have the type-checker
from Edinburgh Tools, but it relies on some oddities of DEC-10 Prolog
(won't run under Poplog for example), and doesn't handle modules.
Jocelyn Paineok@goanna.cs.rmit.oz.au (Richard A. O'Keefe) (06/14/91)
In article <9106111357.AA29175@ucbvax.Berkeley.EDU>, POPX@vax.oxford.ac.uk (Jocelyn Paine) writes: > Is there a generally accepted notation for specifying types in Prolog? No. > Are any of these programs in wide > enough use to merit adopting their notation for types as a de facto > standard? No. There isn't even wide agreement on what kind of types we ought to have. For what it's worth, Bristol's new Goedel language uses essentially the Hindley/Milner/Mycroft types, modulo minor syntactic details. > If so, are any of them in the public domain? I have the type-checker > from Edinburgh Tools, but it relies on some oddities of DEC-10 Prolog > (won't run under Poplog for example), and doesn't handle modules. I am unaware of any dependencies on DEC-10 Prolog. It took very little effort to make it run under C Prolog, Quintus Prolog, SICStus Prolog, NU Prolog. If there are any problems, please let me know. Of course it doesn't handle modules. There is even less agreement about modules than about types. (The latest ISO suggestion I've seen is unacceptably complex. I mean by that that if it becomes standard I will join the Pop camp and forget about Prolog; it'll be too hard to use reliably.) As far as I know, the type checker in the DEC-10 Prolog library (what is the relationship between the DEC-10 library and "Edinburgh tools"?) is the _only_ Prolog type checker which has ever been broadcast on the net. -- Q: What should I know about quicksort? A: That it is *slow*. Q: When should I use it? A: When you have only 256 words of main storage.
chandra@cs.tamu.edu (Chandrasekaran Periannan) (06/14/91)
I would like to know if netters can suggest any introductory literarture on problems with typing logic programming languages like Prolog and goes on to typed prolog. thanks chandra
lakshman@m.cs.uiuc.edu (T.K. Lakshman) (06/17/91)
In article <6333@goanna.cs.rmit.oz.au> ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe) writes: >In article <9106111357.AA29175@ucbvax.Berkeley.EDU>, POPX@vax.oxford.ac.uk (Jocelyn Paine) writes: >> Is there a generally accepted notation for specifying types in Prolog? > >No. > >> Are any of these programs in wide >> enough use to merit adopting their notation for types as a de facto >> standard? > >No. There isn't even wide agreement on what kind of types we ought to >have. For what it's worth, Bristol's new Goedel language uses essentially >the Hindley/Milner/Mycroft types, modulo minor syntactic details. > >> If so, are any of them in the public domain? I have the type-checker >> from Edinburgh Tools, but it relies on some oddities of DEC-10 Prolog >> (won't run under Poplog for example), and doesn't handle modules. > >I am unaware of any dependencies on DEC-10 Prolog. It took very little >effort to make it run under C Prolog, Quintus Prolog, SICStus Prolog, >NU Prolog. If there are any problems, please let me know. > >Of course it doesn't handle modules. There is even less agreement about >modules than about types. (The latest ISO suggestion I've seen is unacceptably >complex. I mean by that that if it becomes standard I will join the >Pop camp and forget about Prolog; it'll be too hard to use reliably.) > >As far as I know, the type checker in the DEC-10 Prolog library (what is >the relationship between the DEC-10 library and "Edinburgh tools"?) is >the _only_ Prolog type checker which has ever been broadcast on the net. These postings and a subsequent posting talk about types for Prolog. As this seems to be a topic of interest, I'd like to RE-POST a message that I posted back in February (Feb 25, 1991 I think). The main point is that THERE IS AVAILABLE BY FTP, A TYPE-CHECKING AND TYPE-RECONSTRUCTION PROCEDURE for Typed Prolog along with examples and a supporting document. Here are the details again. I'd be delighted to hear your comments and respond to any queries regarding the type system. Thanks, T.K. Lakshman PS: A paper on Typed Prolog is to appear in ILPS '91 in San Diego, California. ---------------------------------------------------------------------- TYPED PROLOG: TYPE CHECKING AND RECONSTRUCTION ---------------------------------------------------------------------- This is to (RE)announce the availability of a type checking/reconstruction procedure for TYPED PROLOG: a prescriptively typed logic programming language. Some features of TYPED PROLOG: *. Explicit type declarations (following the syntax of type declarations in the Mycroft-O'Keefe type checker) *. Typed Semantics based on many sorted logic. *. Support for a large subset of Prolog: including the (in)famous cut, not, assert and retract. *. A type CHECKING and RECONSTRUCTION procedure based on Milner's type checking algorithm for ML. The procedure performs: a. TYPE CHECKING : when given a Typed Prolog program with type declarations for all function and predicate symbols, the procedure type-checks the program based on the type rules of Typed Prolog. b. TYPE RECONSTRUCTION : when given a Typed Prolog program where type declarations for predicate symbols are omitted, the procedure reconstructs the omitted types. The user can essentially write ordinary Prolog programs in the standard syntax and use the above-mentioned implementation to type-check the programs and reconstruct omitted types for predicates. ---------------------------------------------------------------------- OBTAINING THE CODE and THE RELATED PAPER ---------------------------------------------------------------------- The type-checking/reconstruction code is written in Prolog and has been tested on Quintus Prolog (v 3.0), C Prolog (v 1.4), SB Prolog (v 2.5) and SICStus Prolog (v 0.7) This type-checking / reconstruction code is available via ANONYMOUS FTP as follows: Machine: a.cs.uiuc.edu. (On the login prompt please say anonymous). Directory: pub/reddy/typed-prolog File: Types.tar.Z (please don't forget to set BINARY) The Types.tar.Z file can be tared/uncompressed as follows: (uncompress Types.tar.Z; tar xf Types.tar; echo "y" | \ rm Types.tar >/dev/null) & This creates the Typed-Prolog directory. This directory contains in addition to the type-checking / reconstruction code and examples, a README file and a dvi and a postscript copy of a related Typed Prolog paper. ---------------------------------------------------------------------- -- INTERNET : lakshman@cs.uiuc.edu OFFICE : 217-244-5973 MAIL : Dept. of C.S., DCL, 1304 W Springfield Ave, Urbana, Il 61801. ----------------------------------------------------------------------
apolis@alw.nih.gov (Alan Polis) (06/17/91)
I noticed NU-PROLOG version was referenced in your letter. I would like to get more information on NU-PROLOG such as: who developed NU-PROLOG, where could I get a copy, and are there references to NU-PROLOG in the literature. Thanks in advance. Alan Polis National Institute of Health Bethesda, MD 20892