Re: Are these all really true ?

  rfg@monkeys.com (Ronald F. Guilmette) writes:
|> Gabriela de Vivo (UCV). <gdevivo@conicit.ve> wrote:
|> >* Strongly type language result in better programs.


I'll assume that "strongly typed" means "compile-time type-checked".


|> No. Strongly typed languages force people to think more about what they
|> are doing. It is the _thinking_ that results in better programs.


The contrary is true in my experience : In a compile-time type-checked
language I tend to be more sloppy. If I am not sure if something is
correct, I sometimes let the compiler at it, and if it typechecks ok,
I assume it is correct (until the bugs show themselves :-).


With run-time typechecked languages (like Prolog) I am a bit more
cautious, but still there are cases where I simply run the program,
just to see if it works (for my test case).


I am most cautious and think most about my program in languages
without type-checking like Forth. As a consequence, I catch errors at
programming time that a typechecker would not catch. The price I pay
for this is a few errors that the type checker would catch; however,
these type errors are usually easy to find, if they show themselves
(but some type errors just result in poor portability). My experience
is confirmed by Stroustroups observation in the HOPL-II proceedings
about the interaction of typechecking and programmer behaviour: Better
compile-time typechecking reduces type errors, but increases the time
spent on each type error.


My conclusion: Compile-time type-checked languages do not result in
better programs, they result in a false sense of security.


- anton
--
M. Anton Ertl
anton@mips.complang.tuwien.ac.at
http://www.complang.tuwien.ac.at/anton/home.html


--

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