Class 50: Records

Held: Tuesday, 2 December 2003

Summary: Today we consider how to build records in Scheme. Records group data and permit access to those data by name.

Related Pages:

• EBoard.
• Reading on Records in Scheme.
• Lab on Records in Scheme.

Assignments

• Read Objects in Scheme.

Notes:

• I will provide an opportunity for you to ask questions about exam 3.
• I've written a sample interface for our student search service.

Overview:

• Grouping Data.
• The Record.
• Records in Scheme.
• Lab.

Grouping Data

• As you've seen in a number of examples, it is often convient to group data together into what I typically call an entry.
• You currently know two basic ways to group data: lists and records.
• Lists provide dynamic-sized groups of data but have costly access.
• Arrays provide fixed-sized groups of data with quick access to elements by number.

Records

• However, as we've seen in our work, it is often more convenient to refer to the parts of a group of data by name.
• For example, given a record for a student, we might ask for the student's first name, last name, or grade.
• Computer scientists use record as the name for a fixed-size group of data with quick access to elements by name.
• In the programming language Pascal, one can define a record of that form with a command like

  type
    Student = record
      firstName: string;
      lastName: string;
      grade: real;
    end;
  

• Unfortunately, Scheme lacks a built-in record type.

Implementing Records

• Fortunately, we can implement records on our own.
• The easiest thing to do is to build a vector and treat the elements of the vector as the fields of the record.
  • We write procedures to access each part of the vector.
• For example

  (define student
    (lambda (lname fname gpa)
      (vector 'student lname fname gpa)))
  (define student-lastname
    (lambda (student)
      (if (not (eq? (vector-ref student 0) 'student))
          (error "Not a student")
          (vector-ref student 1))))
  

• However, we'd like to be able to identify whether or not each vector is supposed to represent one of our desired objects.
• One clever trick is to build a special value that others find it hard to duplicate and to put it at the front of the vector.
• Consider

  (define produce-type-mark
    (let ((type-mark (list 'compound)))
      (lambda () type-mark)))
  

• The let creates a new symbol table with one entry whose name is type-mark and whose value is (compound).
• The lambda gets that value.
• Any other list of the symbol compound will be equal? to the type mark, but not eq?.
• Hence, we can use that value.
• Is this ideal? No. Next class, we'll see another way to build things like records.

Lab

• Do the lab on records.
• We'll reflect a little bit at the start of class tomorrow.