Procedure Inline keyword

[Dreglor]

it has proably been asked before but i didn't find it while searching :\

i think a inline function would be nice.
it would improve speed in some cases and improve readability

there just some times when a procedure call would be nice but the code inside is so small its slower than having inlined.
my raytracer uses alot of this inlining to improve speed becasue of the vector math procedures like DotProducts are tiny but are hardto read and understand in the soup of code when inlined.

so maybe a special procedure keyword such as

Code: Select all

ProcedureInline.f DotProduct(ax.f,ay.f,az.f,bx.f,by.f,bz.f)
  ProcedureReturn ax * bx + ay * by + az * bz
EndProcedure

i don't think it would too difficult to implement just replace all procedures with there code

[Dräc]

Wait for PB v4 and keyword "Macro"

[Fred]

yes, macros will be perfects for such stuffs

[remi_meier]

And what about local variables in macros? That's the difference between
inline-procedures and macros!
Fred: Do you plan to implement them like #define or like inline procedures?

[Fred]

mostky like #define (with some more flexibility)

[Dare2]

Good stuff.

In the meantime, Dreglor, why not write the "inline" code in an include file and include whereever needed? That means only one place to make a change rather than hunting through all places where the particular sequence of code is used.

[tinman]

Are there any plans to add inline functions (either forced as Dreglor requested or automatic based on something like executable size) or are we stuck with macros?

[Mistrel]

What do you need an inline function for that a macro can't provide?

[tinman]

What do you need an inline function for that a macro can't provide?

Local variables that do not interfere with the variables already defined everywhere you use the macro.

You can take the address of a procedure (although not sure how that would work if it was forced inline).

A macro can not have a return value (you need to use a fixed known variable name or your macro can only be a single statement).

[Mistrel]

Ahh, I see. I agree that it would be a nice feature then.

[luis]

Continuing what tinman was saying:

- macro's shortcomings (no return values AND side effects) removed

- ability to validate params at compile time

- easier to debug


I was looking to see if a request for inline procedures was already made and found this one.

So, even if we have macros, I would like to have inline procedures as well.

[Shield]

Inline Procedures should be up to the compiler, imho. That means that the compiler
should decide when to inline a function and when not.

In C/C++ for example, the keyword 'inline' is just a hint to the compiler that this
function might be a case for inlining, but it's entirely up to the compiler.

While I'm usually in favor of such additions I think PB should just decide on its own when it makes sense
to inline a function and when it doesn't. Programmers should just always use a procedure.

[netmaestro]

Ha, got me. I read this post over twice and was about to reply, 'how is a macro different from what you're describing?' and then I saw all the replies and realized this thread is 8 years old!

[luis]

... and yet a macro is still different from an inline procedure even if 8 years are passed, some differences are outlined in the posts above

or read here -> http://en.wikipedia.org/wiki/Inline_function

[Rinzwind]

Why stil no inline procedure?

Very different from macro's... Inline procedures are perfect for short small read/check functions

Short macOS example:

Code: Select all

Procedure kOS(constant.s)
  Protected t
  t = dlsym_(#RTLD_DEFAULT, constant)
  If t
    t = PeekI(t)
  EndIf
  ProcedureReturn t
EndProcedure

How to do this with a macro?