OpenEuphoria

Re: WISHLIST.TXT

• Posted by Kat <gertie at PELL.NET> Aug 23, 2002
• 439 views

On 23 Aug 2002, at 11:55, C. K. Lester wrote:

> 
> > :restartloop:
> > -- initvars
> > for loop = 1 to something do
> >  --code
> >   if whatever1 then goto restartloop end if
> >   if whatever2 then goto endloop end if
> >   if whatever3 then goto resumeloop end if
> > --code
> > :resumeloop:
> > -- possible code
> > end for
> > :endloop:
> 
> I've never gotten involved in any arguments regarding goto because I really
> don't care. I don't use them... wouldn't use them... etc... But, I don't care
> if
> they're part of the language, as long as code without gotos doesn't suffer.
> 
> Kat, my question is, would there be some extra garbage collection required
> if you jump out of a loop that hasn't completed? In your example above, when
> whatever2 is true, would the interpreter know that you've exited the loop and
> will never return? Does the interpreter even need to know that? I'm just
> thinking that stuff would be left hanging, and that doesn't seem very
> efficient
> to me unless the interpreter is smart enough to know where you've gone and
> what
> you've left undone.

The way i see it, the loop is treated like a local function in Eu (it was in
turbo
pascal), which is why you don't haveto declare the loop variable. Any jump to 
*outside* this function will generate default cleanup code exactly as it does 
now. Forward jumps are easy to add, backwards jumps to *before* the loop 
start are more difficult, it's a few more steps (exit loop, cleanup, reinit the 
loop var, find the target (that's easy, we passed over it before, we know 
where it is), and jmp there with everything else unchanged. The fun part is 
the loops are simply exited, triggering the same cleanup as now exists, in 
the compiled code,, i assume the interpreter is smart enough too.

This way of treating the loops allowed another neat trick in TP: the 
embedded procedure or function. Which would fullfill another request made in 
this thread, a request made before a few times too: a 3rd level of var scope. 
In Pascal, i can do this (pseudocode):

procedure 1()
declare vars for procedure 1()

function 2()  -- vars' scope is procedure 1()
end function2

procedure 4()  -- vars' scope is procedure 1()
end procedure

procedure 1()'s code here
call function 2()
more procedure 1()'s code here
call procedure 4()
more procedure 1()'s code here
procedure 1()'s code here
call function 2()

end procedure 1()

For scoping problems, this was ideal, and in the disassembler, it looked a lot 
like the loop's machine code, anything in the loop was scoped to the calling 
procedure as well.

Also, see: Karl's Bach / Bliss in the archives. 
 
> Just curious,

Curious is good.

Kat