deliverables/report/eval.tex

   1 \section{Evaluation}
   2 All in all we are quite happy with \SPLC. Our added syntactic sugar and the
   3 higher order functions make our implementation of \SPL{} quite comfortable to
   4 program in. Due to the UNIX style interface it is quite easy to combine
   5 usage of the compiler with \SSM.
   6
   7 There are however of course a lot of things to can be improved. We discuss a few
   8 of the most interesting and most pressing issues. First of all error handling
   9 in \SPLC{} can be greatly improved. Currently a lot of error happen at position
  10 (0,0), most notably all type errors happen at this position. It would be a
  11 great improvement in usability if errors would always be reported at the
  12 correct position. A second issue is that in \SPLC{} it is not possible to
  13 spread a program over multiple modules, when developing larger programs this
  14 would be an absolute requirement. Thirdly \SPLC{} currently does not allow for
  15 multiple recursion, and all functions need to be declared before they are
  16 called. It would be nice to have multiple recursion and to not be forced to
  17 structure your program in the order in which functions are called.
  18
  19 \subsection{Known Bugs}
  20 For higher order functions there is a small bug that when an assignment is done
  21 to a variable with a higher order function type where the number of arguments
  22 is bigger than one and one of the, not-last, arguments is a higher order
  23 functien itself the behaviour is unspecified. The bug happens because the
  24 function is placed on the heap but the heap is distorted because one of the
  25 arguments also modifies the heap. We are certain that with some time we can fix
  26 this problem since it is an assembly issue.
  27
  28 \subsection{Work division}
  29 \begin{description}
  30         \item [Lexing \& parsing]:
  31                 \begin{description}
  32                 \item [\Yard] Pim
  33                 \item [Lexing] Mart \& Pim
  34                 \item [Parsing] Mostly Mart, some Pim
  35                 \item [Sugar] literal string: Mart, literal lists: Pim, Variable
  36                         arguments printing: Mart, Let-expansion: Pim, Lambdas: Pim
  37                 \end{description}
  38         \item [Semantical analysis]:
  39                 \begin{description}
  40                 \item [Sanity checks] Mart
  41                 \item [Type inference] Mostly Pim, some Mart
  42                 \item [Lambda lifting] Pim
  43                 \end{description}
  44         \item [Code generation]:
  45                 \begin{description}
  46                 \item [RWST monad] Mart
  47                 \item [Basic Generation] Mart \& Pim
  48                 \item [Higher order functions] Mart
  49                 \end{description}
  50 \end{description}
  51
  52 \subsection{Metrics}
  53 Between the February the seventh and June the sixth, $18.283$ lines of code
  54 been written of which $12.497$ have been removed later on. Detailed records
  55 can be found on the open git repository available on
  56 \url{https://github.com/dopefishh/cc1516}. Most of the code is written with
  57 pair programming.
  58
  59 The source code of \SPLC{} contains $2051$ lines of code, the rest of the lines
  60 of code committed can be attributed to the presentations, test programs and
  61 grammar specification.