[Aldor-l] [Axiom-developer] spad: language and compiler

[Gabriel Dos Reis]

Jacques Carette <carette at mcmaster.ca> writes:

| Ralf Hemmecke wrote:
| > Compile time evaluation in full 
| > generality introduces a way make it really hard to find bugs.
| >
| > But anyway, maybe Aldor should allow compile time evaluation.
| >   
| Quick remark --  it has been shown that:
| 1) C++'s template language is a Turing Complete PL
| 2) Haskell's class types (with common extensions) is also a Turing 
| Complete PL
| 3) All meta-programming systems allow arbitrary compile time evaluation
| 
| Yes, it does make debugging harder.  But the advantages seem to _far_ 
| outweigh the problems.  One just develops new debugging (and coding) 
| techniques to deal with the added power/complexity.
| 
| I could re-use Stephen's brilliant closing line from yesterday's email:
|     "These kinds of errors have to be seen as bugs in programs, just as 
| division by zero is a programming error and not an invalidation of 
| integer arithmetic." -- S.M. Watt
| [where 'These kinds of errors' is now /Programs with infinite loops in 
| types/]

I agree with most of what you said.  However, the slogan "well-typed
programs don't go wrong" does some value that I would heisate to
compromise...

-- Gaby