Class 15: Repetition with Recursion

Held: Wednesday, 12 February 2003

Summary: Today we begin to consider one of the most powerful tools in Scheme, recursion. Recursion allows you to repeat operations.

Related Pages:

• EBoard.
• Reading on Recursion (assigned today).

Assignments

• Read Repetition with Recursion.

Notes:

• Reflection: What did you learn from the lab on CGI?
• Are there questions on Homework 2?
• Today's class will be conducted primarily in a lecture + discussion style. Be prepared to participate!
• You may want to read an inteteresting note on using click here by folks even more opinionated than I.

Overview:

• Repetition.
• Recursion.
• Examples.
• Recursion in Scheme.

Repetition

• You may recall that when we first considered algorithms we identified a number of key aspects of algortihms:
  • variables: the ability to name things;
  • conditions: the ability to choose between things;
  • procedures: the ability to name (and parameterize) collections of steps;
  • repetition: the ability to do something multiple times;
  • input and output: the ability to get and report values.
• We've already seen how to do almost all of these things, except for repetition.
• Examples of repetition from baking:
  • Stir the mix 50 times
  • Knead the bread dough until it feels like your earlobe
  • Bake until golden-brown.
• Examples of repetition from mathematics:
  • Sum these values
  • Find the smallest of these values
• Examples of repetition from everyday life:
  • Naively find a name in the phone book
  • Do I have a CD by Van Morrison?
• Examples of repetition from previous Scheme exercises:
  • How long is this list?
  • Is X a member of this list?

Recursion

• In Scheme, the most common mechanism for repetition is recursion.
• To do something that involves repeated actions, you
  • Do one action
  • Repeat the rest
  • Combine the results if necessary.
• For example, to stir your cake mix 50 times, you stir it one time and then stir it 49 more times.
• More generally, to stir a cake mix n times, you stir it one time and then n-1 more times.
• Similarly, to knead dough until its the right consistency, you knead it a little, check the consistency, and, if it's not the right consistency, knead it until its the right consistency.
• In the case of mathematics, to sum a list we might add the first value to the sum of the remaining values (or add the last value to the sum of the initial values).
• There are a few key aspects to recursive design:
  • You need to know when you're done (and what to do when you're done). This aspect of recursive design is called the base case.
  • You need to know what to do when you're not done. Here, you should do a little, try again, and then perhaps combine the results. This aspect of recursive design is called the recursive case.
  • You need to be sure that you're getting closer to the base case (otherwise you'll never stop).

Some Examples

• We'll look together at how we might solve a few recursive list-based examples as a way of thinking about the problem. I've included a variety of possibilities.
• sum, sum the elements in a list.
• member: Determine if a value is in a list.
• length: Find the length of a list
  • We'll assume that the built-in length procedure does not exist.
• smallest: Find the smallest value in a list of numbers.
• average: Find the average value in a list of numbers.
• increment-elements: Add two to each number in a list of numbers.
• increment elements (revised): Add two to each number in a list of many different kinds of values.
• The code is not in the outline because I don't want to bias your answers.

Recursion in Scheme

• Here's the form of a typical recursive procedure:

  (define proc
    (lambda (val)
      (if (base-case-test)
          (base-case val)
          (combine (partof val)
                   (proc (update val))))))
  

• When the value you're working with is a list and your base case is the null list, the form is somewhat simpler:

  (define proc
    (lambda (lst)
      (if (null? lst)
          null-case
          (combine (onestep (car val))
                   (proc (cdr val))))))