samth-dissertation/popl-related.tex
Sam Tobin-Hochstadt 9c7a001a36 init
2017-07-10 13:02:10 -04:00

78 lines
3.6 KiB
TeX

There are numerous categories of related work that we must
consider. First, embedding logic in type systems or treating type
systems as logics, has a long history. Second, a few systems with
logical propositions \emph{in addition to} types have been developed.
Third, several researchers have investigated type systems for
untyped languages. Fourth, languages with intensional type analysis
allow checks similar to those performed by \scheme|number?|.
Fifth, type systems that correspond to flow
analyses have explored a related design space.
% Sixth, the idea of
%refining a type into a smaller type is familiar from refinement type
%systems.
\section{Types and Logic}
\section{Deriving Propositions}
A few systems derive formulae that describe the program that are
not simply richer types. For example, Might's Logic Flow
Analysis~\cite{might:lfa} derives propositions about the equality of
numeric values during the run of a flow analysis. The overall design of
Might's system is different, since it involves an external
theorem prover in interaction with a flow analysis, and derives
propositions that are radically different from ours. However, in a type
checking setting, we expect that Might's system would be able to
derive similar theorems to those of \lts.
\section{Types for Untyped Languages}
In the past few years, there has been a wide variety of work on type
systems for untyped languages. In addition to Typed Scheme, proposals
have been made for Ruby~\cite{ruby-static09}, JavaScript~\cite{ecma}, and
other languages. To our knowledge, no other system has yet
incorporated occurrence typing, although Furr et al state ``one of our
benchmarks could benefit from adding occurrence types''.
\section{Intensional Polymorphism}
Languages with intensional polymorphism~\cite{weirich98} also offer
introspective operations allowing programmers to analyze the type of
the data provided to functions, as Typed Scheme does. This is
typically done via a \texttt{typecase} construct. Typed Scheme and
occurrence typing in the spirit of \lts
provides significantly greater flexibility. As we
describe in this paper, Typed Scheme is able to use predicates applied
to selectors, reason about combinations of tests, abstract over type
tests, and use logical constructs such as implication to enhance the
expressivity of the system.
\section{Types and Flow Analysis}
Since the development of control flow analysis for higher-order
languages~\cite{shivers-thesis}, a number of researchers have related
these analyses to type systems. In particular, a \emph{safety
analysis} for a flow analysis can be cast as a type system. Palsberg
and O'Keefe~\cite{palsberg95} show that a form of 0CFA can be cast as
a system with recursive types and subtyping, and Palsberg and
Pavlopoulou~\cite{palsberg01} show that a polyvariant analysis is
equivalent to a system with union and intersection types.
In chapter
9 of his thesis~\cite{shivers-thesis}, Shivers describes a type
recovery analysis that includes refining the type of variables in type
tests, exploited by Wright~\cite{aw:soft-typing} and
Flanagan~\cite{cf:static-debugging} to create soft typing systems for Scheme.
We conjecture that there is a relationship between analyses of
this type and type systems such as \lts.
%% \section{Refinement Types}
%% Refinement types are types with both a base type and a predicate,
%% including only those values that are of the base type and satisfy the
%% predicate~\cite{fp:refinement-types,cf:hybrid,mwh:refinement}. \lts
%% could be viewed as a refinement type system, but these previous
%% systems have not used the tests of \scheme|if| expressions to
%% calculate the refinements. Recently,