Re: Passing Parameter as Variable vs as address of variable

Hi Patrik

Just a touch more.

When a parameter is passed from RPG, it will be passed by reference by default. In this case, there is no copying of the variable contents to another address, so no time taken to duplicate anything.

If, as Barbara mentioned, specify CONST on the parameter of the prototype, then a copy of the variable is made at a new address, and that address is passed.

Also, I did mean to say that a pointer is only a "look at what is at this address" - if you pass a pointer by value, there is no data type information included - so if you are passing a pointer to an integer, there is no check that, in fact, it is an integer. That is the danger of passing a pointer by value.

In another exchange here, you asked, why was BY VALUE added? The reason is, certain system APIs and Unix-style functions require it. An example is the access() function, which tells you whether a file can be accessed in a certain manner - many of us use it to see if an IFS object exists.

This is the C/C++ declaration -

int access(const char/*path/, int/amode/);

You have to pass the 2nd parameter by value - and use the VALUE keyword in the prototype in RPG.

Regards
Vern

On 8/26/2023 4:22 PM, Patrik Schindler wrote:

Hello Vern,

Am 26.08.2023 um 19:17 schrieb Vern Hamberg via RPG400-L<rpg400-l@xxxxxxxxxxxxxxxxxx>

I think you might be used to programming in C or C++ - in those languages, parameters are passed by value, by default. To pass by reference, you have to prefix the parameter name with an asterisk. As you know.

Thanks. Your assumption about my skill is correct. :-)

In IBM i languages, CL and RPG (I don't know about COBOL), the default is that parameters are passed by reference. This is automatic, unless you specify "by value".

I'm aware.

So it is extremely _*not*_ recommended to pass a pointer (by value). You completely lose any data type checking when compiling that program.

I was not aware that passing a pointer includes not only a bare "look at this address", but also some additional data. :-O

Also, you cannot define an RPGLE program to have an incoming parameter passed in by value - so all RPG (and CL) programs expect to receive a pointer to whatever parameter is listed. I just verified this, both with the default MAIN procedure that uses the cycle and with a MAIN control-option where you specify a procedure the does not use the cycle.

Thanks for taking time to run a test!

I think, the "conversion to a value" is merely a duplication of the data to a new memory location and *that* location is passed to the called application. Yes?

Performance-wise, as soon as data duplication happens, there is a (negligible?) performance penalty.

:wq! PoC