[E-Lang] Thanksgiving present for E programmers

[Mark S. Miller]

At 01:23 AM 11/22/00, Tyler Close wrote:
>... Using the ":type" syntax, an
>E programmer can write static type checks. ...

I agree with your entire point except for the use of the word "static" in 
the line quoted above.  I strongly suspect we have a terminology-only 
disagreement.  E's :type syntax, as you clearly fully understand from the 
rest of your message, expresses a check that must be performed at runtime. 
Of course, just like a compiler can perform "3 + 4" at compile time instead 
of runtime, a compiler may also use static type checking techniques to 
perform many of E's :type checks at compile time.  But in both cases, the 
compiler must transparently maintain the illusion that computation 
happens only at runtime.

E's :type checking is like static type checking in that it causes early 
rejection of conditions that may lead to a protocol mismatch problem.  But 
it is unlike static type checking with regard to how early that is.  By 
"static" checking, I've always meant rejection sometime before the program 
is allowed to begin execution.  By contrast, E's :type checking causes 
rejection when a variable would be bound or a value returned.

OTOH, E's :type syntax is also designed so that a programmer (or a 
programming shop) can build/buy and use other tools outside the E 
implementation itself to truly statically check types or other statically 
checkable quality issues, and, if they wish, impose on themselves the 
discipline of only allowing themselves to commit to programs that pass their 
internal static quality checks.  The variety of kinds of static quality 
checks a shop might want to impose on itself is so great, and getting global 
agreement on which ones are good is so hard, it is important to pull these 
out of the language definition or implementation.  The older folks among us 
may imagine one such tool might be called "ELint".


        Cheers,
        --MarkM