Commit d4f84ee5 authored by Kim Nguyễn's avatar Kim Nguyễn
Browse files

Update related (connection between selfification and our work).

Update practical (mentions n0=1 for our examples).
parent a26e7df9
default:
latexmk -pdf -bibtex main.tex
clean:
rm -rf *.synctex* *.blg *.dvi *.aux *.fdb_latexmk *.log *.fls *.xcp *.out
......@@ -258,15 +258,6 @@ url={http://www.cduce.org/papers/frisch_phd.pdf}
}
@PhdThesis{Pet19phd,
author = {Tommaso Petrucciani},
title = {Polymorphic Set-Theoretic Types for Functional Languages},
school = {Joint Ph.D.\ Thesis, Università di Genova and Université Paris Diderot},
month = mar,
url = {https://tel.archives-ouvertes.fr/tel-02119930},
year = 2019,
note = {Available at \url{https://tel.archives-ouvertes.fr/tel-02119930}},
}
@article{castagna2019gradual,
......
......@@ -27,7 +27,11 @@ the algorithmic system $\vdashA$ as well as all the type operations
defined in this work. One optimization that our implementation
features (with respect to the formal presentation) is the use of a
memoization environment in the code of the $\Refine {e,t}{\Gamma}$
function, which allows the inference to avoid unnecessary traversals of $e$.
function, which allows the inference to avoid unnecessary traversals
of $e$.
Lastly, while our prototype allows the user to specify a particular
value for the $n_0$ parameter, the default value of $1$ is sufficient
to check all examples we present in the following sections.
}
\subsection{Experiments}
......
......@@ -261,13 +261,16 @@ language and their approach is extended with polymorphism, dynamic
dispatch and record types.
\rev{%%%
A problem that is faced by refinement type systems is the one of
propagating in the branches of a test the information learned on
a \emph{sub-expression} of a tested expression. A solution
propagating in the branches of a test the very precise information
learned from the test (usually that some equality between terms holds). A solution
that is for instance chosen by~\citet{OTMW04} and \citet{KF09} is to devise a
meta-function that recursively explores both a type and an expression
and propagates the type information on the sub-expressions. This
and constructs a more precise \emph{dependent} type. In the dependent
type, fresh variables are introduced to name sub-expressions and
record the new constraints.
This
process---called \emph{selfification} in the cited works---roughly
corresponds to our $\constrf$ function (see Section~\ref{sec:typenv}).
corresponds to a our $\constrf$ and \Refinef{} functions (see Section~\ref{sec:typenv}).
Another approach is the one followed by \citet{RKJ08} which is
completely based on a program transformation, namely, it consists in putting the term
in \emph{administrative normal form} (\cite{SF92}). Using a program
......@@ -377,8 +380,11 @@ Racket~\cite{THF10} the type-system associates to an expression a
quadruple formed by its type, two logical propositions, and an object which is
a pointer to the environment for the type hypothesis about the
expression and, as such, it plays the role of our extended type
environments. Likewise, the \emph{selfification} of ... \textbf{KIM
PLEASE ADD A LINE ON THIS POINT}.
environments. Likewise, the \emph{selfification} of \cite{OTMW04}
and \cite{KF09}, propagates the precise type constraints learned
during a test. However, in the latter, whole the information can be
kept at the type level, since dependent types contains terms and can
introduce variables, while in our approach the mapping is kept in an environment.
Nevertheless, we are confident that even this last non-standard aspect
of our system can be removed and that occurrence typing can be
explained in a pure standard type-theoretic setting.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment