X-Git-Url: https://git.martlubbers.net/?a=blobdiff_plain;f=methods.top.tex;h=811e12a1c85df94705541c400588d98ef73e5880;hb=b8e0188d0d4b4f321259e036807b439c85753828;hp=52ed983bcd4f327dcbe64efe6894bb5e21f85b4c;hpb=b0fc85be57e369c5a04adaba4d6920a59d3d36e2;p=msc-thesis1617.git diff --git a/methods.top.tex b/methods.top.tex index 52ed983..811e12a 100644 --- a/methods.top.tex +++ b/methods.top.tex @@ -1,19 +1,25 @@ -\section{\acrlong{TOP}} -\subsection{\gls{iTasks}} -\gls{TOP} is a recent new programming paradigm implemented as -\gls{iTasks}\cite{achten_introduction_2015} in the pure lazy functional -language \gls{Clean}\cite{brus_cleanlanguage_1987}. \gls{iTasks} is a -\gls{EDSL} to model workflow tasks in the broadest sense. A \CI{Task} is just -a function that, given some state, returns the observable \CI{TaskValue}. The -\CI{TaskValue} of a \CI{Task} can have different states. Not all state +\section{iTasks} +\gls{TOP} is a novel programming paradigm implemented as +\gls{iTasks}~\cite{achten_introduction_2015} in the pure lazy functional +language \gls{Clean}~\cite{brus_cleanlanguage_1987}. \gls{iTasks} is an +\gls{EDSL} to model workflow tasks in the broadest sense. A \gls{Task} is just +a function that --- given some state --- returns the observable \CI{TaskValue}. +The \CI{TaskValue} of a \CI{Task} can have different states. Not all state transitions are possible as shown in Figure~\ref{fig:taskvalue}. Once a value -has gone stable it can never become unstable again. Stability is often reached -by pressing a confirmation button or for \glspl{Task} that offer a constant -value. A simple example is shown in Listing~\ref{lst:taskex} accompanied with -Figure~\ref{fig:taskex1},~\ref{fig:taskex2} and~\ref{fig:taskex3}. In this -example the first image in is the \CI{NoValue} state, the second and third -image are in the \CI{Unstable} state. When the user presses \emph{Continue} the -value becomes \CI{Stable}. +is stable it can never become unstable again. Stability is often reached by +pressing a confirmation button. \glspl{Task} yielding a constant value are +immediately stable. + +A simple \gls{iTasks} example illustrating the route to stability of a +\gls{Task} in which the user has to enter a full name is shown in +Listing~\ref{lst:taskex}. The code is accompanied by screenshots showing the +user interface in Figure~\ref{fig:taskex1},~\ref{fig:taskex2} +and~\ref{fig:taskex3}. The \CI{TaskValue} of the \gls{Task} is in the first +image in the \CI{NoValue} state, the second image does not have all the fields +filled in and therefore the \CI{TaskValue} remains \CI{Unstable}. In the third +image all fields are entered and the \CI{TaskValue} transitions to the +\CI{Unstable} state. When the user presses \emph{Continue} the value becomes +\CI{Stable} and cannot be changed any further. \begin{figure}[H] \centering @@ -21,11 +27,11 @@ value becomes \CI{Stable}. \caption{The states of a \CI{TaskValue}}\label{fig:taskvalue} \end{figure} -\begin{lstlisting}[language=Clean,label={lst:taskex},% +\begin{lstlisting}[label={lst:taskex},% caption={An example \gls{Task} for entering a name}] -:: Name = { firstname :: String - , lastname :: String - } +:: Name = { firstname :: String + , lastname :: String + } derive class iTask Name @@ -36,6 +42,7 @@ enterName = enterInformation "Enter your name" [] \end{lstlisting} \begin{figure}[H] + \centering \begin{subfigure}{.25\textwidth} \centering \includegraphics[width=.9\linewidth]{taskex1} @@ -54,10 +61,149 @@ enterName = enterInformation "Enter your name" [] \caption{Example of a generated user interface} \end{figure} -For a type to be suitable it must have instances for a collection of generic -functions that are captured in the class \CI{iTask}. Basic types have -specialization instances for these functions and show an according interface. -Generated interfaces can be modified with decoration operators. +For a type to be suitable, it must have instances for a collection of generic +functions that is captured in the class \CI{iTask}. Basic types have +specialization instances for these functions and show an interface accordingly. +Derived interfaces can be modified with decoration operators or specializations +can be created. + +\section{Combinators} +\Glspl{Task} can be combined using so called \gls{Task}-combinators. +Combinators describe relations between \glspl{Task}. There are only two basic +types of combinators; namely parallel and sequence. All other combinators are +derived from the basic combinators. Type signatures of simplified versions of +the basic combinators and their derivations are given in +Listing~\ref{lst:combinators} + +\begin{lstlisting}[% + caption={\Gls{Task}-combinators},label={lst:combinators}] +//Step combinator +(>>=) infixl 1 :: (Task a) (a -> Task b) -> Task b | iTask a & iTask b +(>>*) infixl 1 :: (Task a) [TaskCont a (Task b)] -> Task b | iTask a & iTask b +:: TaskCont a b + = OnValue ((TaskValue a) -> Maybe b) + | OnAction Action ((TaskValue a) -> Maybe b) + | E.e: OnException (e -> b) & iTask e + | OnAllExceptions (String -> b) +:: Action = Action String + +//Parallel combinators +(-||-) infixr 3 :: (Task a) (Task a) -> Task a | iTask a +(||-) infixr 3 :: (Task a) (Task b) -> Task b | iTask a & iTask b +(-||) infixl 3 :: (Task a) (Task b) -> Task a | iTask a & iTask b +(-&&-) infixr 4 :: (Task a) (Task b) -> Task (a,b) | iTask a & iTask b +\end{lstlisting} + +\paragraph{Sequence:} +The implementation for the sequence combinator is called the +\CI{step} (\CI{>>*}). This combinator runs the left-hand \gls{Task} and +starts the right-hand side when a certain predicate holds. Predicates +can be propositions about the \CI{TaskValue}, user actions from within +the web browser or a thrown exception. The familiar +bind-combinator is an example of a sequence combinator. This combinator +runs the left-hand side and continues to the right-hand \gls{Task} if +there is an \CI{UnStable} value and the user presses continue or when +the value is \CI{Stable}. The combinator could have been implemented +as follows: +\begin{lstlisting}[language=Clean] +(>>=) infixl 1 :: (Task a) (a -> (Task b)) -> (Task b) | iTask a & iTask b +(>>=) ta f = ta >>* [OnAction "Continue" onValue, OnValue onStable] + where + onValue (Value a _) = Just (f a) + onValue _ = Nothing + + onStable (Value a True) = Just (f a) + onStable _ = Nothing +\end{lstlisting} + +\paragraph{Parallel:} +The parallel combinator allows for concurrent \glspl{Task}. The +\glspl{Task} combined with these operators will appear at the same time +in the web browser of the user and the results are combined as the type +dictates. All parallel combinators used are derived from the basic parallel +combinator that is very complex and only used internally. + +\section{Shared Data Sources} +\Glspl{SDS} are an abstraction over resources that are available in the world +or in the \gls{iTasks} system. The shared data can be a file on disk, the +system time, a random integer or just some data stored in memory. The actual +\gls{SDS} is just a record containing functions on how to read and write the +source. In these functions the \CI{*IWorld} --- which in turn contains the real +\CI{*World} --- is available. Accessing the outside world is required for +interacting with it and thus the functions can access files on disk, raw +memory, other \glspl{SDS} and hardware. + +The basic operations for \glspl{SDS} are get, set and update. The signatures +for these functions are shown in Listing~\ref{lst:shares}. By default, all +\glspl{SDS} are files containing a \gls{JSON} encoded version of the object and +thus are persistent between restarts of the program. Library functions for +shares residing in memory are available as well. The three main operations on +shares are atomic in the sense that during reading no other \glspl{Task} are +executed. The system provides useful functions to transform, map and combine +\glspl{SDS} using combinators. The system also provides functionality to +inspect the value of a \gls{SDS} and act upon a change. \Glspl{Task} waiting on +a \gls{SDS} to change are notified when needed. This results in low resource +usage because \glspl{Task} are never constantly inspecting \gls{SDS} values but +are notified. -\subsection{Combinators} -\todo{Stukje over combinators, in ieder geval bind en paralel} +\begin{lstlisting}[% + label={lst:shares},caption={\Gls{SDS} functions}] +:: RWShared p r w = ... +:: ReadWriteShared r w :== RWShared () r w +:: ROShared p r :== RWShared p () r +:: ReadOnlyShared r :== ROShared () r + +:: Shared r :== ReadWriteShared r r + +get :: (ReadWriteShared r w) -> Task r | iTask r +set :: w (ReadWriteShared r w) -> Task w | iTask w +upd :: (r -> w) (ReadWriteShared r w) -> Task w | iTask r & iTask w + +sharedStore :: String a -> Shared a | JSONEncode{|*|}, JSONDecode{|*|} +\end{lstlisting} + +\section{Parametric Lenses} +\Glspl{SDS} can contain complex data structures such as lists, trees and even +resources in the outside world. Sometimes, an update action only updates a part +of the resource. When this happens, all waiting \glspl{Task} looking at the +resource are notified of the update. However, it may be the case that +\glspl{Task} were only looking at parts of the structure that was not updated. +To solve this problem, parametric lenses were +introduced~\cite{domoszlai_parametric_2014}. + +Parametric lenses add a type variable to the \gls{SDS} that is in the current +library functions fixed to the void type (i.e. \CI{()}). When a \gls{SDS} +executes a write operation, it also provides the system with a notification +predicate. This notification predicate is a function \CI{p -> Bool} where +\CI{p} is the parametric lens type. This allows programmers to create a big +\gls{SDS}, and have \glspl{Task} only look at parts of the big \gls{SDS}. This +technique is used in the current system in memory shares. The \CI{IWorld} +contains a map that is accessible through a \gls{SDS}. While all data is +stored in the map, only \glspl{Task} looking at a specific entry are notified +when the structure is updated. The type of the parametric lens is the key in +the map. + +Functionality for setting parameters is added in the system. The most important +functions are the \CI{sdsFocus} and the \CI{sdsLens} function. These functions +are listed in Listing~\ref{lst:focus}. \CI{sdsFocus} allows the programmer to +fix a parametric lens value. \CI{sdsLens} is a kind of \CI{mapReadWrite} +including access to the parametric lens value. This allows the creation of +for example \glspl{SDS} that only read and write to parts of the original +\gls{SDS}. + +\begin{lstlisting}[label={lst:focus}, + caption={Parametric lens functions}] +sdsFocus :: p (RWShared p r w) -> RWShared p` r w | iTask p + +:: SDSNotifyPred p :== p -> Bool + +:: SDSLensRead p r rs = SDSRead (p -> rs -> MaybeError TaskException r) + | SDSReadConst (p -> r) +:: SDSLensWrite p w rs ws = SDSWrite (p -> rs -> w -> MaybeError TaskException (Maybe ws)) + | SDSWriteConst (p -> w -> MaybeError TaskException (Maybe ws)) +:: SDSLensNotify p w rs = SDSNotify (p -> rs -> w -> SDSNotifyPred p) + | SDSNotifyConst (p -> w -> SDSNotifyPred p) + +sdsLens :: String (p -> ps) (SDSLensRead p r rs) (SDSLensWrite p w rs ws) (SDSLensNotify p w rs) + (RWShared ps rs ws) -> RWShared p r w | iTask ps +\end{lstlisting}