CS 152, Programming Paradigms Fall 2011, SJSU

1. CS 152, Programming ParadigmsFall 2011, SJSU Jeff Smith

2. Parameter passing methods • Parameters may be used for input or output or both. • Parameter values may be copied from argument values. • Or parameters may be aliases of actual arguments.

3. Parameter passing in Ada • Ada distinguishes among IN, OUT, and IN OUT parameters. • The first case is the default. • In Ada, IN parameters cannot be assigned to. • So they may be passed by reference. • If OUT parameters can be read, they're just IN OUT parameters

4. Conventional terminology • pass by value • IN parameter, passed as copy • pass by reference • IN OUT parameter, passed as reference • pass by result • OUT parameter, passed as copy • pass by value-result • IN OUT parameter, passed as copy

5. Pass by value • is often the default method • is eager evaluation • may require much copying • required in C • recall that array variables are effectively pointers • required in Java • recall that object variables are effectively pointers

6. Pass by value and constant parameters • If pass-by-value parameters are not mutable, they may be implemented as references • thus saving the expense of copying • If pass-by-value parameters are mutable, they may be implemented as local variables • initialized to the actual arguments

7. Pass by reference • is available in C++ and Pascal • It can be implemented just by passing the address of the argument. • and having the procedure dereference it as needed. • In C this can be done by the programmer • by passing pointers. • The actual arguments must be L-values -- so they can't be null.

8. Pass by value vs. pass by reference • Unlike pass by value, pass by reference leads to aliasing (cf. T&N, p 233). • But unlike pass by value, pass by reference saves the expense of copying.

9. Pass by value-result • can be implemented with one copy in each direction • can differ from pass by reference • if there is aliasing and side effects • cf. T&N, p. 234, or Louden, p. 321 • IN OUT parameters in Ada-95 are passed by value-result if they are elementary. • and by reference if they are composite.

10. Pass by name • is different than the other methods. • It is often confusing but sometimes useful. • Think of it as textual substitution • of the arguments for the parameters • It can give results different from the other methods, when side effects are involved.

11. Pass by name and lazy evaluation • Pass by name allows the defining of functions that don't evaluate all of their arguments. • or don't evaluate them all immediately • This approach is called lazy (or delayed) evaluation, vs. eager evaluation. • It requires arguments to be passed unevaluated • if they’re variables, their names are passed. • Scheme and Haskell use it to construct (implicitly) infinite objects.

12. Applicative vs. normal order • A related issue is the distinction between applicative and normal order in evaluation. • Applicative order corresponds to bottom up evaluation of a syntax tree. • It’s more commonly used in languages. • In normal order, arguments are passed unevaluated, as in pass by name. • Note: (f x) is evaluated twice in normal order evaluation of (square (f x)).