mirror of
https://github.com/luau-lang/luau.git
synced 2024-11-15 14:25:44 +08:00
157 lines
6.8 KiB
TeX
157 lines
6.8 KiB
TeX
\documentclass[acmsmall]{acmart}
|
|
|
|
\setcopyright{rightsretained}
|
|
\copyrightyear{2023}
|
|
\acmYear{2023}
|
|
\acmConference[HATRA '23]{Human Aspects of Types and Reasoning Assistants}{October 2023}{Portugal, Spain}
|
|
\acmBooktitle{HATRA '23: Human Aspects of Types and Reasoning Assistants}
|
|
\acmDOI{}
|
|
\acmISBN{}
|
|
\expandafter\def\csname @copyrightpermission\endcsname{\raisebox{-1ex}{\includegraphics[height=3.5ex]{cc-by}} This work is licensed under a Creative Commons Attribution 4.0 International License.}
|
|
\expandafter\def\csname @copyrightowner\endcsname{Roblox.}
|
|
|
|
\newcommand{\ANY}{\mathtt{any}}
|
|
\newcommand{\ERROR}{\mathtt{error}}
|
|
\newcommand{\NUMBER}{\mathtt{number}}
|
|
|
|
\begin{document}
|
|
|
|
\title{Goals of the Luau Type System, Two Years On}
|
|
|
|
\author{Lily Brown}
|
|
\author{Andy Friesen}
|
|
\author{Alan Jeffrey}
|
|
\affiliation{
|
|
\institution{Roblox}
|
|
\city{San Mateo}
|
|
\state{CA}
|
|
\country{USA}
|
|
}
|
|
|
|
\begin{abstract}
|
|
In HATRA 2021, we presented \emph{The Goals Of The Luau Type System},
|
|
describing the human factors of a type system for a
|
|
language with a heterogeneous developer community. In this extended
|
|
abstract we provide a progress report, focusing on
|
|
the unexpected aspects: semantic subtyping and type error
|
|
suppression.
|
|
\end{abstract}
|
|
|
|
\maketitle
|
|
|
|
\section{Recap}
|
|
|
|
Luau~\cite{Luau} is the scripting language used by the
|
|
Roblox~\cite{Roblox} platform for shared immersive experiences. Luau extends
|
|
the Lua~\cite{Lua} language, notably by providing type-driven tooling
|
|
such as autocomplete and API documentation (as well as traditional type
|
|
error reporting). Roblox has hundreds of millions of users, and
|
|
millions of creators, ranging from children learning to program for
|
|
the first time to professional development studios.
|
|
|
|
In HATRA 2021, we presented a position paper on the \emph{Goals Of The Luau Type
|
|
System}~\cite{BFJ21:GoalsLuau}, describing the human factors issues
|
|
with designing a type system for a language with a heterogeneous
|
|
developer community. The design flows from the needs of the different
|
|
communities: beginners are focused on immediate goals (``the stairs
|
|
should light up when a player walks on them'') and less on the code
|
|
quality concerns of more experienced developers; for all users
|
|
type-driven tooling is important for productivity. These needs result in a design with two modes:
|
|
\begin{itemize}
|
|
\item \emph{non-strict mode}, aimed at non-professionals, focused on
|
|
minimizing false positives (that is, in non-strict mode, any program with a type error has a defect), and
|
|
\item \emph{strict mode}, aimed at professionals, focused on
|
|
minimizing false negatives (that is, in strict mode, any program with a defect has a type error).
|
|
\end{itemize}
|
|
%% For both communities, type-driven tooling is important, so we
|
|
%% provide \emph{infallible type inference}, which infers types
|
|
%% even for ill-typed or syntactically invalid programs.
|
|
|
|
\section{Progress}
|
|
|
|
In the two years since the position paper, we have been making changes
|
|
to the Luau type system to achieve the goals we set out. Most of the
|
|
changes were straightforward, but two were large changes in how we
|
|
thought about the design of the type system: replacing the existing
|
|
syntactic subtyping algorithm by \emph{semantic subtyping}, and
|
|
treating gradual typing as \emph{type error suppression}.
|
|
|
|
Semantic subtyping
|
|
interprets types as sets of values, and subtyping as set
|
|
inclusion~\cite{GF05:GentleIntroduction}. This is aligned with the
|
|
\emph{minimize false positives} goal of Luau non-strict mode, since
|
|
semantic subtyping only reports a failure of subtyping when there is a
|
|
value which inhabits the candidate subtype, but not the candidate
|
|
supertype.
|
|
For example, the program:
|
|
\begin{verbatim}
|
|
local x : CFrame = CFrame.new()
|
|
local y : Vector3 | CFrame
|
|
if math.random() < 0.5 then y = CFrame.new() else y = Vector3.new() end
|
|
local z : Vector3 | CFrame = x * y
|
|
\end{verbatim}
|
|
cannot produce a run-time error, since multiplication of \verb|CFrame|s is overloaded:
|
|
\begin{verbatim}
|
|
((CFrame, CFrame) -> CFrame) & ((CFrame, Vector3) -> Vector3)
|
|
\end{verbatim}
|
|
In order to typecheck this program, we check that that type is a subtype of:
|
|
\begin{verbatim}
|
|
(CFrame, Vector3 | CFrame) -> (Vector3 | CFrame)
|
|
\end{verbatim}
|
|
In the previous, syntax-driven, implementation of subtyping, this subtype check would fail, resulting in a false positive.
|
|
We have now released an implementation of semantic subtyping, which does not suffer from this defect.
|
|
See our technical blog for more details~\cite{Jef22:SemanticSubtyping}.
|
|
|
|
Rather than the gradual typing approach
|
|
of Siek and Taha~\cite{ST07:GradualTyping}, which uses \emph{consistent
|
|
subtyping} where $\ANY \lesssim T \lesssim \ANY$ for any type $T$, we
|
|
adopt an approach based on \emph{error suppression}, where $\ANY$ is
|
|
an error-suppressing type, and any failures of subtyping involving
|
|
error-suppressing types are not reported. Users can explicitly
|
|
suppress type errors by declaring variables with type $\ANY$, and
|
|
since an expression with a type error has an error-suppressing type we
|
|
avoid cascading errors.
|
|
|
|
We do this by defining a \emph{infallible} typing judgment $\Gamma \vdash M : T$
|
|
such that for any $\Gamma$ and $M$, there is a $T$ such that $\Gamma \vdash M : T$.
|
|
For example the rule for addition (ignoring overloads for simplicity) is:
|
|
\[
|
|
\frac{\Gamma \vdash M : T \quad \Gamma \vdash M : U}{\Gamma \vdash M+N : \NUMBER}
|
|
\]
|
|
We define which judgments produce warnings, for example that rule produces a warning
|
|
when
|
|
\begin{itemize}
|
|
\item either $T \not<: \NUMBER$ and $T$ is not error-suppressing,
|
|
\item or $U \not<: \NUMBER$ and $U$ is not error-suppressing.
|
|
\end{itemize}
|
|
To retain type soundness (in the absence of user-supplied error-suppressing types)
|
|
we show that
|
|
if $\Gamma \vdash M : T$ and $T$ is error-suppressing, then either
|
|
\begin{itemize}
|
|
\item $\Gamma$ or $M$ contains an error-suppressing type, or
|
|
\item $\Gamma \vdash M : T$ produces a warning.
|
|
\end{itemize}
|
|
From this it is straightforward to show the usual ``well typed
|
|
programs don't go wrong'' type soundness result for programs without explicit
|
|
error-suppressing types~\cite{BJ23:agda-typeck}.
|
|
|
|
\section{Further work}
|
|
|
|
Currently the type inference system uses greedy inference, which is
|
|
very sensitive to order of declarations, and can easily result in
|
|
false positives. We plan to replace this by some form of local type
|
|
inference~\cite{PT00:LocalTypeInference}.
|
|
|
|
Currently, non-strict mode operates in the style of gradual type
|
|
systems by inferring $\ANY$ as the type for local variables. This does
|
|
not play well with type-directed tooling, for example $\ANY$ cannot
|
|
provide autocomplete suggestions. Local type inference will infer more
|
|
precise union types, and hence better type-driven tooling.
|
|
|
|
At some point, we hope that error suppression will be the only difference
|
|
between strict mode and non-strict mode.
|
|
|
|
\bibliographystyle{ACM-Reference-Format} \bibliography{bibliography}
|
|
|
|
\end{document}
|