X-Git-Url: https://git.martlubbers.net/?a=blobdiff_plain;f=methods.dsl.tex;h=0ee9a243c681b908cb2fa534b2171e9e3825861f;hb=d118ff9d857683084065145df45135ef6fa06711;hp=639eec42b5d5f0c7a12c7054f7ab84f5fbcc41dd;hpb=d0c05e34c230ff8292466e58709d1f45feb05919;p=msc-thesis1617.git diff --git a/methods.dsl.tex b/methods.dsl.tex index 639eec4..0ee9a24 100644 --- a/methods.dsl.tex +++ b/methods.dsl.tex @@ -1,59 +1,132 @@ -\section{\acrlong{EDSL}s} There are several techniques available for creating \glspl{EDSL}. Each of them have their own advantages and disadvantages such as extendability, typedness and view support. In the following subsections each of the main techniques are briefly explained. -\subsection{Deep embedding} +\section{Deep embedding} A deep \gls{EDSL} means that the language is represented as an \gls{ADT}. Views are functions that transform something to the datatype or the other way around. As an example we have the simple arithmetic \gls{EDSL} shown in -Listing~\ref{lst:exdeep}. Deep embedding has the advantage that it is very -simple to build and the views are easy to make and add. However, there are also -a few downsides. +Listing~\ref{lst:exdeep}. -\begin{lstlisting}[language=Clean,label={lst:exdeep},caption={A minimal deep -\gls{EDSL}}] +\begin{lstlisting}[language=Clean,label={lst:exdeep},% + caption={A minimal deep \gls{EDSL}}] :: DSL - = LitI Int - | LitB Bool - | Var String - | Plus DSL DSL + = LitI Int + | LitB Bool + | Var String + | Plus DSL DSL | Minus DSL DSL - | And DSL DSL - | Eq DSL + | And DSL DSL + | Eq DSL \end{lstlisting} - -The expressions created with this language are not type-safe. In the given -language it is possible an expression such as \CI{Plus (LitI 4) (LitB True)} -which to add a boolean to an integer. Evermore so, extending the \gls{ADT} is -easy and convenient but extending the views accordingly is tedious and has to -be done individually for all views. +Deep embedding has the advantage that it is very simple to build and the views +are easy to make and add. However, there are also a few downsides. The +expressions created with this language are not type-safe. In the given language +it is possible an expression such as \CI{Plus (LitI 4) (LitB True)} which to +add a boolean to an integer. Evermore so, extending the \gls{ADT} is easy and +convenient but extending the views accordingly is tedious and has to be done +individually for all views. The first downside of the type of \gls{EDSL} can be overcome by using -\glspl{GADT}. Listing~\ref{lst:exdeepgadt} shows the same language, but -type-safe with a \gls{GADT}\footnote{Note that \glspl{GADT} are not supported -in \gls{Clean}. They can be simulated using bimaps}\todo{cite}. Unfortunately -the lack of extendability stays a problem. +\glspl{GADT}\cite{cheney_first-class_2003}. Listing~\ref{lst:exdeepgadt} shows +the same language, but type-safe with a \gls{GADT}. \glspl{GADT} are not +supported in the current version of \gls{Clean} and therefore the syntax is +artificial. However, it has been shown that \glspl{GADT} can be simulated using +bimaps or projection pairs\cite{cheney_lightweight_2002}. Unfortunately the +lack of extendability stays a problem. If a language construct is added no +compile time guarantee is given that all views support it. -\begin{lstlisting}[language=Clean,label={lst:exdeep},% - caption={A minimal deep \gls{EDSL}}] +\begin{lstlisting}[language=Clean,label={lst:exdeepgadt},% + caption={A minimal deep \gls{EDSL} using \glspl{GADT}}] :: DSL a - = LitI Int -> DSL Int - | LitB Bool -> DSL Bool - | Var String -> DSL Int - | Plus (DSL Int) (DSL Int) -> DSL Int - | Minus (DSL Int) (DSL Int) -> DSL Int - | And (DSL Bool) (DSL Bool) -> DSL Bool - | E.e: Eq (DSL e) (DSL e) -> DSL Bool & == e + = LitI Int -> DSL Int + | LitB Bool -> DSL Bool + | E.e: Var String -> DSL e + | Plus (DSL Int) (DSL Int) -> DSL Int + | Minus (DSL Int) (DSL Int) -> DSL Int + | And (DSL Bool) (DSL Bool) -> DSL Bool + | E.e: Eq (DSL e) (DSL e) -> DSL Bool & == e +\end{lstlisting} + +\section{Shallow embedding} +In a shallowly \gls{EDSL} all language constructs are expressed as functions in +the host language. An evaluator view for our example language then looks +something like the code shown in Listing~\ref{lst:exshallow}. Note that much of +the internals of the language can be hidden using monads. + +\begin{lstlisting}[language=Clean,label={lst:exshallow},% + caption={A minimal shallow \gls{EDSL}}] +:: Env = ... // Some environment +:: DSL a = DSL (Env -> a) + +Lit :: a -> DSL a +Lit x = \e->x + +Var :: String -> DSL Int +Var i = \e->retrEnv e i + +Plus :: (DSL Int) (DSL Int) -> DSL Int +Plus x y = \e->x e + y e + +... + +Eq :: (DSL a) (DSL a) -> DSL Bool | == a +Eq x y = \e->x e + y e \end{lstlisting} -\subsection{Shallow embedding} +The advantage of shallowly embedding a language in a host language is the +extendability. It is very easy to add functionality and compile time checks +guarantee that the functionality is available. Moreover, the language is type +safe as it is directly typed in the host language. + +The downside of this method is extending the language with views. It is near +impossible to add views to a shallowly embedded language. The only way of +achieving this is by decorating the datatype for the \gls{EDSL} with all the +information for all the views. This will mean that every component will have to +implement all views. This makes it slow for multiple views and complex to +implement. + +\section{Class based shallow embedding} +The third type of embedding is called class based shallow embedding and has the +best of both shallow and deep embedding. In class based shallow embedding the +language constructs are defined as type classes. The same language is shown +with the new method in Listing~\ref{lst:exclassshallow}. + +This type of embedding inherits the easiness of adding views from shallow +embedding. A view is just a different data type implementing one or more of the +type classes as shown in the aforementioned Listing where an evaluator and a +pretty printer are implemented. + +Just as with \glspl{GADT} in deep embedding type safety is guaranteed. Type +constraints are enforced through phantom types. One can add as many phantom +type as is necessary. Lastly, extensions can be added easily, just as in +shallow embedding. When an extension is made in an existing class, all views +must be updated accordingly to prevent possible runtime errors. When an +extension is added in a new class this problem does not arise and views can +choose to implement only parts of the collection of classes. -\subsection{Class based shallow embedding} -\todo{while iTasks is also a DSL\ldots} -\glspl{mTask} are expressed in a class based shallowly embedded \gls{EDSL}. -There are two main types of \glspl{EDSL}. -\todo{Small shallow embedded dsl intro} -\todo{Small deep embedded dsl} -\todo{Show that class based has the best of both worlds} +\begin{lstlisting}[language=Clean,label={lst:exclassshallow},% + caption={A minimal class based shallow \gls{EDSL}}] +:: Env = ... // Some environment +:: Evaluator a = Evaluator (Env -> a) +:: PrettyPrinter a = PP String + +class intArith where + lit :: t -> v t | toString t + add :: (v t) (v t) -> (v t) | + t + minus :: (v t) (v t) -> (v t) | - t + +class boolArith where + and :: (v Bool) (v Bool) -> (v Bool) + eq :: (v t) (v t) -> (v Bool) | == t + +instance intArith Evaluator where + lit x = \e->x + add x y = ... + +instance intArith PrettyPrinter where + lit x = toString x + add x y = x +++ "+" +++ y + ... +\end{lstlisting}