On Using Scheme to Introduce Prolog

1
On Using Scheme to Introduce Prolog

• Daniel E. Stevenson and Michael R. Wick
• Department of Computer Science
• University of Wisconsin - Eau Claire
• Eau Claire, WI 54701

2
Road Map

• Motivation
• Reductionist Approach
• Scheme-to-Prolog Strategy
• The Debate
• Experiment Design
• Results and Observations
• Follow-up Experiment
• Results and Observations
• Future Work
• Contact Information

3
Our Motivation

• Want to give students a concrete strategy to move
  from a blank page to a working Prolog program
• We follow a reductionist approach to teaching
  Programming Languages
• Includes use of translational semantics
• Scheme is our target language
• Simple syntax
• Level playing field
• Expressive power

4
A Reductionist Approach to Programming Languages

• Focus on language building blocks
• What is a paradigm?
• Collection of strategic decisions involving the
  blocks
• What is a language?
• Collection of tactical decisions involving the
  blocks
• Our Implementation
• Cover basic syntax description
• Demonstrate building blocks in Scheme
• Implement paradigms in Scheme
• Assignments are in other languages

5
A Prolog/Scheme Correspondence (1)
append(,L,L). append(HT,L,HNewT) -
append(T,L,NewT).
(define (append L1 L2) (cond ((null? L1)
L2) (else (car L1)
(cdr L1)
,L,L
H
T
L
HNewT
append(T,L,NewT)
(cons ))))
L2
(append )

• Many problems in Prolog are list-processing
  problems.
• The concepts of list processing and recursive
  definitions are similar between Scheme and Prolog
• Prologs pattern matching plays the role of car
  and cdr in Scheme
• Prolog rules map to Scheme cond clause

6
A Prolog/Scheme Correspondence (2)
isMember(A, AL). isMember(A, XL) -
isMember(A, L).
(define (isMember A L) (cond ((null? L)
f) (car L)
(cdr L)
A
A, A
A
L
isMember(A, L)
((eq? A ) t)
A
(else (isMember? )))

• Many problems in Prolog are predicate problems.
• The concepts are again similar between Scheme and
  Prolog
• Prologs pattern matching plays the role of car
  and cdr in Scheme
• Prolog positive rules map to Scheme cond clauses
• Scheme adds negative cond clauses

7
A Scheme-to-Prolog Strategy

• To Translate a Scheme Program into a Prolog
  Program
• Write the required behavior (predicate versus
  procedure) in Scheme
• Define the corresponding Prolog relation. If the
  Scheme program is a predicate, use the same
  number of parameters in both versions.
  Otherwise, use one additional parameter in the
  Prolog version to represent the Scheme return
  value.
• Translate each positive clause in the Scheme
  version to Prolog.
• If a Scheme clause requires access to the car or
  the cdr of an input parameter, use the pattern
  HT in the head of the corresponding Prolog
  definition.
• If a Scheme clause requires two values to be
  equal, use the same variable in Prolog for each
  corresponding occurrence of the two values in
  Scheme.
• If a Scheme clause requires the results of other
  Scheme functions, use Prologs subgoaling to
  query and name the results of these functions.

8
The Debate

• Does our Scheme-to-Prolog strategy help or hurt?
• Do students waste too much time in Scheme and not
  enough time learning Prolog?
• Given Schemes relationship to mathematics, is
  there any difference in learning between using
  purely mathematical recursive definitions and
  using recursive Scheme procedures?
• Does the use of Scheme hurt students when it
  comes to learning the backtracking concepts of
  Prolog?
• Two Hypotheses
• Null Hypothesis I Our use of Scheme to introduce
  students to Prolog does not impact the students
  ability to learn early Prolog.
• Null Hypothesis II Our use of Scheme to
  introduce students to Prolog does not impact the
  students ability to learn later Prolog.

9
Experiment Design (1)

• A total of 44 students in our Programming
  Languages course
• Students split into control and experiment groups
• 22 students each
• Same CS course background
• Six students in each group had discrete including
  brief intro to Prolog
• Exact same average CS GPA
• Exact same average Overall GPA
• Both groups given one-hour lecture on logic
  programming
• Both groups given Prolog Tutorial from Bucknell
  University
• Control given recursive definition strategy
• Experiment given Scheme-to-Prolog strategy

10
Experiment Design (2)

• Students given four one-hour sessions to answer
  10 questions
• Positives(L1, L2). L2 are the positive numbers
  from L1.
• Max(N, L). N is the maximum number in L.
• Palindrome(L). L is a palindrome.
• Prefix(P, L). P is a prefix of L.
• Last(X, L). X is the last element of L.
• Rem(X, L1, L2). L2 is L1 with all occurrences of
  X removed.
• NextTo(X, Y, L). X and Y are consecutive
  elements of L.
• Replace(N1,N2,L1,L2). L2 is L1 with every
  occurrence of N1 replaced by N2.
• Parent(P, C). P is a parent of C.
• Ancestor(A, C). A is an ancestor of C.
• Students also given in-class exam at completion
  of logic programming section of course
  (emphasizing backtracking)

11
Results and Observations (1)

• Number of Students Who Answered Each Question
  Correctly
• Total Number of Questions Answered Correctly

12
Results and Observations (2)

• Students Scores on In-Class Exam on Logic
  Programming

13
Follow-up Experiment

• Not a warm fuzzy
• Are we helping our students learn new languages
  at all?
• Final Exam
• Give students 10 questions to implement in a new
  language
• Give students BNF, reference manual, and tutorial
  as resources
• Pessimistic Experiment
• Pick language/paradigm similar to known
  (Java/C)
• Ruby object-oriented with familiar syntax
• Should be least difference between control and
  experiment
• Divide into control (students after CS 330) and
  experiment (students before CS 330)
• Comparison between two different student bodies
• Neither reported knowledge of Ruby
• Control GPA 3.11
• Experiment GPA 3.09

14
Results and Observations

• Whew!

15
Future Work Contact Information

• Modify pre/post exams to use Haskell, ML, Snobol,
  
• Test on students with deeper exposure to Scheme
• Question-by-question analysis of pre/post exams
• Daniel Stevenson (stevende_at_uwec.edu)
• Michael Wick (wickmr_at_uwec.edu)
• Department of Computer Science
• University of Wisconsin Eau Claire
• Eau Claire, WI 54701