Data Abstraction

• In our main program, we don't really know how the bignum is implementd.
• Is this good?
  • Portions of programs can be developed in parallel
  • Underlying details can be changed without changing the program.
  • (Think about reversing how the digits are stored)
  • Or any other change in how we do things...
• It turns out that we often want to do this.
• Data Abstraction is the separation of a data type's logical properties form its implementation.
• For example: integers
  • High Abstraction:
    • Integers represent whole numbers {..., -3, -2, -1, 0, 1, 2, 3, ...}
    • They support the operations +,-, comparison, *, and /
  • Implementation:
    • an integer is represented by 32 bits
    • integers are stored as a two's compliment number
    • To add two integers
      • align least significant bit
      • add bits, with a carry to the next most significant bit
      • If the final carry out does not equal the carry in, overflow.
  • We talk of a specification, which tells us what a function or a data type does.
  • We also have an implementation, which is the structure or algorithm to accomplish the specification.
• An Abstract Data Type (ADT) is a data type whose properties (domain and operations) are specified independently of any implementation.
• Think about our bignum package again
  • The abstraction
    • An signed integer up to 100 digits
    • Operations
      • addition subtraction
      • multiplication
      • compare, swap
      • I/O
  • The implementation
    • A struct, containing ...
• Let's try another, how about a counter (like a turnstile)
• Often we have the following catigories of routnes
  • Constructors
  • Transformers
  • Observers
  • Iterators