Analysis Of Helium Programs Obtained Through Logging
Room: BBL room 416
Title: Analysis of Helium programs obtained through logging
(master's thesis defense)
In the recent years the Helium compiler is used to teach Functional Programming at
Universiteit Utrecht. This compiler is equipped with a logger, which
collects information about each compilation done by a student in one of
the many lab sessions of the course. This vast collection of loggings
hides information regarding the behavior of students programming in
Haskell and the performance of the Helium compiler regarding the quality
of error message.
In this talk we present the work of collecting and analyzing the data set.
The research work forms a framework for evaluating a software tool, like a
compiler, through the use of loggings. We present our approach of
collecting and analyzing the data set using a custom developed tool set,
which includes a combinator library for doing descriptive statistical
analysis. With this tool set we present a set of interesting analyses,
providing a view on programming aspects like: the ratio of actual code
lines and documentation lines over time, ratio of type hints in the
compiler output for a set of loggings, the time needed to solve type
errors, with and without supportive hints.