Re: First-class data types

Bill Torcaso:
          There is a nit to pick with the definition of 'first class
      datatype' that asserts any operation can be applied to any object
      of any first class datatype. Consider a language in which functions
      are first-class:


Is that supposed to be a statement about the domain of every function
in the language? [e.g. that the empty function, which has no domain,
is not a proper member of the language?] If so that statement strikes
me as suspicious.


          What is the XOR of two functions? What is the AND of two
          functions?


let h = f XOR g where f and g are functions.
then h(x) is f(x) XOR g(x)


Similarly for AND.


          What is the function-invocation of the integer constant 17? Of
          the floating-point constant 0.5?


let h = the function 17
then h(x) is 17


Similarly for 0.5


      Any language that admits arithmetic types and arithmetic operations
      will have a hard time supporting first-class types. And of course,
      once you subset the domain of datatypes and the operators that
      apply in each sub-domain, the definition loses all generality.


This assertion may be true, but the above cases are easily resolvable.
[For people who aren't aware of this yet: both of the above features
are available, albeit with slightly different syntax, in the language
J. But so far J's just an interpreted language.]


--
Raul Deluth Miller-Rockwell <rockwell@socrates.umd.edu>
--

Return to the comp.compilers page.
Search the comp.compilers archives again.