EndProcedure value

[Danilo]

Let me re-iterate. The "hidden" return will only be replaced if "EndProcedure value" is used.
All use of ProcedureReturn inside the procedure will continue to function just like they do today.

Let me simplify this further. "EndProcedure value" let you replace the xor eax,eax with a mov eax, value

That's it. That is the only thing I'm suggesting, I'm not suggesting to change ProcedureReturn at all.

With some dead-code-elimination at PB and assembly level, this could be avoided at all.
It then would be optimized automatically, without the need for syntax changes or additional syntax.

For example, all code after a major ProcedureReturn statement is dead code and can be removed
by the compiler - if control flow does not jump to a label after the ProcedureReturn statement.
Likewise, at assembly level, dead code can be eliminated if never executed and not reachable by control flow -
but all that would probably be easier with a second pass over the generated ASM.

A more simple approach:
If the last statement in a procedure is ProcedureReturn (and that's often the case),
it is not required to generate code for returning the default 0-value.
'EndProcedure Value' wouldn't be required. It is just a small optimization to the current system.

According to my understanding, you just want some more optimizations for generated procedure code.

[uwekel]

+1

Nice idea! Would save me more than hundred lines of code! I am sure the discussion in this thread took more time than implementing this feature by the PB team

[Dude]

all code after a major ProcedureReturn statement is dead code and can be removed

Say what? What about times when you have multiple ProcedureReturns depending on a decision?

[Little John]

With some dead-code-elimination at PB and assembly level, this could be avoided at all.
It then would be optimized automatically, without the need for syntax changes or additional syntax.

Exactly.

[wilbert]

With some dead-code-elimination at PB and assembly level, this could be avoided at all.
It then would be optimized automatically, without the need for syntax changes or additional syntax.

Exactly.

Seems hard to implement.
I have sometimes procedures like this.

Code: Select all

Procedure Test()
  !mov eax, [l_test]
  ProcedureReturn
  !l_test: dd 5
EndProcedure

Debug Test()

How would PB detect it has to stay in ?

[Little John]

Hi wilbert,

I was referring to assembler code that has been produced by PB itself from plain PB code entered by the programmer.
As far as I understand, lines which start with ! are not touched by PB at all, but are just passed to the assembler. I think that should not be changed.

[dobro]

a,b,c = someprocedure(foo)

Not possible in purebasic, right?

maybe :

Code: Select all

Declare  foo(in)

foo(2)

debug a
debug b
debug c

Procedure foo(in)
	shared a,b,c
	a=in*2
	b=in*4
	c=in*8
EndProcedure

[StarBootics]

a,b,c = someprocedure(foo)
Not possible in purebasic, right?

You are right it's possible in MatLAB or FreeMat. By the way my favorite way to handle multiple return :

Code: Select all

Declare  foo(in, *OutA, *OutB, *OutC)

foo(2, @a.l, @b.l, @c.l)

Debug a
Debug b
Debug c

Procedure foo(in, *OutA, *OutB, *OutC)
   
   PokeL(*OutA, in*2)
   PokeL(*OutB, in*4)
   PokeL(*OutC, in*8)
   
EndProcedure

Best regards
StarBootics

[Bo Marchais]

Very handy!

[TI-994A]

a,b,c = someprocedure(foo)

Not possible in purebasic, right?

Using a non-global array, and without passing the array to the procedure:

Code: Select all

Procedure someprocedure(foo, *bar)
  For i = 1 To 3
    PokeI(*bar + offset, i * foo)
    offset + SizeOf(Integer)
  Next i
EndProcedure

Dim a(2)
someprocedure(123, a())

Debug a(0)
Debug a(1)
Debug a(2)

Or, if your feeling especially adventurous:

Code: Select all

;#### WORKS BUT NOT RECOMMENDED! ####
;assumes contiguous memory allocation

Procedure someprocedure(foo, *bar)
  For i = 1 To 3
    PokeI(*bar + offset, i * foo)
    offset + SizeOf(Integer)
  Next i
EndProcedure

Define.i a, b, c
someprocedure(123, @a)

Debug a
Debug b
Debug c

[dobro]

with String

Code: Select all

Declare.s  foo()

o.s=foo()

For i=1 to 3
	debug StringField(o.s,i,"," )
Next i

Procedure.s foo()
	;Huey, Dewey and Louie
	a.s="Huey"
	b.s="Dewey"
	c.s="Louie"
	
	out$=a.s+","+b.s+","+c.s ; group
	ProcedureReturn out$
EndProcedure

; Epb

or

Code: Select all

Declare.s  foo()

foo()
debug a.s
debug b.s
debug c.s

Procedure.s foo()
	shared a.s,b.s,c.s
	a.s="Huey"
	b.s="Dewey"
	c.s="Louie"
EndProcedure
; Epb

[wilbert]

with String

Another way

Code: Select all

Procedure MulDiv(a, b, *mul.Integer, *div.Integer)
  *mul\i = a * b
  If b 
    *div\i = a / b
  EndIf
EndProcedure
 
Procedure MulDivS(a, b, *mul.String, *div.String)
  *mul\s = Str(a * b)
  If b
    *div\s = Str(a / b)
  EndIf
EndProcedure


MulDiv(20, 5, @a, @b)
MulDivS(16, 2, @c.String, @d.String)

Debug a
Debug b

Debug c\s
Debug d\s

[Lunasole]

Then it is outside. Procedure / EndProcedure it's a code block, so after EndProcedure it would be outside.

It is not "outside". There already is For/Next code block and Next can take argument (for example).

I like any syntax improvements making code shorter, more usable, less ugly and raw/"looking like a .bat file" without loss of readability and ability to parse it easily with some script (for case you need to translate it to other language). This idea match all criterias, so might be nice.

[DontTalkToMe]

It is not "outside". There already is For/Next code block and Next can take argument (for example).

LOL

OK, I get your point above, Still I wouldn't put a clear-cut code unit like procedure-endprocedure on the same plane of repeat-until or for-next.
You can even write for i - next i (so i is after the next).

[Lunasole]

@DontTalkToMe, hah, It happens. It is hard for me to read whole threads in english (takes much more time than native langs) this time posted without reading and missed, sorry