[Prev][Next][Index][Thread]
the Robinson-Tennent coercion
Date: Thu, 20 Oct 88 10:34:11 PDT
To: rdt%qucis.bitnet@mitva.mit.edu, types@theory.LCS.MIT.EDU
Robinson and Tennent propose a coercion to solve the polymorphic update
problem for labeled records, but comment that its syntactic and semantic
compatibility with other coercions remains to be checked. This comment
suggested a couple of questions:
1. Is the proposed coercion meant to be an injection? This seems
plausible with the semantics for labeled records given by Cardelli
(Information and Computation, Feb/March 1988). If an injection, is
there still an issue of compatibility?
2. If for some reason, a non-trivial conversion function is required,
is there a known set of sufficient conditions to check compatibility?
The only relevant set I know is the one given by Reynolds in 1980.
That set seems limited to coercions of atomic sorts in the presence
of generic operators, and does not deal with structural coercions
(I think).
Actually, I am interested in an answer to the second question on its own
for some work I am doing right now.
Satish Thatte