home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.prolog
- Path: sparky!uunet!mcsun!sunic!sics.se!torkel
- From: torkel@sics.se (Torkel Franzen)
- Subject: Re: Occurs check
- In-Reply-To: jamie@cs.sfu.ca's message of Fri, 18 Dec 1992 17:57:14 GMT
- Message-ID: <TORKEL.92Dec19121914@bast.sics.se>
- Sender: news@sics.se
- Organization: Swedish Institute of Computer Science, Kista
- References: <1992Dec17.111142.24450@dcs.qmw.ac.uk> <24435@alice.att.com>
- <TORKEL.92Dec17201903@bast.sics.se> <1992Dec18.175714.3917@cs.sfu.ca>
- Date: Sat, 19 Dec 1992 11:19:14 GMT
- Lines: 19
-
- In article <1992Dec18.175714.3917@cs.sfu.ca> jamie@cs.sfu.ca (Jamie Andrews)
- writes:
-
- >However, there are still completeness problems, with cases
- >like
- > x=f(x), y=f(y), x=y
- >where such systems typically loop instead of succeeding, and
- >even cases like
- > x=f(x,g(x)), y=f(y,h(y)), x=y
- >where such systems may loop where they should fail.
-
- Yes, and "unification without occur check" is of course a poor way of
- dealing with rational trees. Sicstus supports unification of rational
- trees, and thus treats the above cases correctly. However, the argument
- I criticized might as well have been applied to a proper implementation
- of rational trees.
-
-
-
-