Pages

Wednesday, September 4, 2019

Variable length arrays in RPG

arrays with varying number of elements

One of the new additions to RPG in IBM i 7.4, and not to the equivalent TR of 7.3, was the ability to have variable length arrays.

Having read the documentation, and having had a chance to "play", my findings mean I can divide this post three sections:

  1. Setting the number of elements in the array
  2. Expanding number of elements
  3. Compile time array

Before I get started I want to show what the maximum size of an array, this number has not changed since at least IBM I 7.2 . It is not the number of elements in the array that is the limit, it is the total size of the array. The array must not exceed 16,773,104 bytes. For example, this is valid as an array of 16,773,104 elements of one character is valid.

dcl-s Array1 char(1) dim(16773104) ;

If I increase it by one byte the program will not compile:

000500 dcl-s Array2 char(1) dim(16773105) ;
======>                         aaaaaaaa
*RNF3452 20 a   000500  The parameter for keyword DIM is not 
                        valid; defaults to 1.

Or increase the size of the individual elements to 10:

000800 dcl-s Array3 char(10) dim(16773104) ;
======>      aaaaaa
*RNF0376 30 a   000800  Total length 167731040 of data item
                        ARRAY3 exceeds 16,773,104 bytes.

If I reduce the number of elements by a factor of ten, then the array is valid.

dcl-s Array4 char(10) dim(1677310) ;

This is just for your information, I am not encouraging anyone to create arrays of that size.

 

Setting the number of elements in the array

When I define my array the format of the DIM keyword is different:

dcl-s ArrayVar char(1) dim(*var:1000) ;

Rather than give a just number for the number of elements I have the *VAR value followed by the maximum number of elements the array can have. I chose 1,000 just because it looked like a nice round number to me, you can pick any number for your own array.

As I have not set the array to any number of elements if I display the number of elements in the array there are zero:

dsply ('1. Var = ' + %char(%elem(ArrayVar)) ) ;

> DSPLY  1. Var = 0

If I try to move a value into one of the elements of the array I get an error, which is not a surprise as the array has zero elements.

ArrayVar(1) = '1' ;

RNQ0121: An array index is out of range (C G D F).

To set the size of the array I use the %ELEM built in function, then I fill the array up to that number of elements:

%elem(ArrayVar) = 5 ;

for Counter = 1 to %elem(ArrayVar) ;
  ArrayVar(Counter) = %char(Counter) ;
endfor ;

dsply ('3. Var = ' + %char(%elem(ArrayVar)) ) ;
dsply ('4. Var(5) = <' + ArrayVar(5) + '>') ;

> DSPLY  3. Var = 5
> DSPLY  4. Var(5) = <5>

Then I can just reset the array back to having zero elements:

%elem(ArrayVar) = 0 ;
dsply ('5. Var = ' + %char(%elem(ArrayVar)) ) ;

> DSPLY  5. Var = 0

What happens to the values in that I placed in the array when I increase the number of elements back to five? The values in the array's elements were cleared.

%elem(ArrayVar) = 5 ;
dsply ('6. Var = ' + %char(%elem(ArrayVar)) ) ;
dsply ('7. Var(5) = <' + ArrayVar(5) + '>') ;

> DSPLY  6. Var = 5
> DSPLY  7. Var(5) = < >

I know it goes without saying, but if I try to increase the number of elements to greater than the 1,000 maximum I defined in the array's definition I get a compile error:

%elem(ArrayVar) = 1001 ;

RNF7563: The expression is not valid for assignment to 
         built-in function %LEN or %ELEM.

Sometimes IBM adds something I consider bizarre, the following example is one of those.

dcl-s ArrayVar2 char(1) inz('*') dim(*var:1000) ;

In this array I want to initialize every element with an asterisk ( * ). If I add ten elements to this array then all ten contain an asterisk:

%elem(ArrayVar2) = 10 ;
dsply ('1. Var2(10) = <' + ArrayVar2(10) + '>') ;

> DSPLY  1. Var2(10) = <*>

What if I now want to have another ten elements, but not have them initialized with the asterisk.

%elem(ArrayVar2) = 0 ;
%elem(ArrayVar2:*keep) = 20 ;
dsply ('2. Var2(20) = <' + ArrayVar2(20) + '>') ;

> DSPLY  2. Var2(20) = < >

In the second line of code the %ELEM BiF has a second parameter, *KEEP, this stops the new elements, 11-20, from being initialized with the asterisk. Why you want to initialize the elements in an array with one value and then not want to do it for others is, well, beyond me.

 

Expanding number of elements

To use an error with an "expanding" number of elements I need to use the *AUTO value in the DIM of the array's definition:

dcl-s ArrayAuto char(1) dim(*auto:1000) ;

The array starts off with zero elements…

dsply ('1. Auto = ' + %char(%elem(ArrayAuto)) ) ;

> DSPLY  1. Auto = 0

And I can add elements to the array just by referencing them, no need to allocate a number of elements as I did before. For example:

for Counter = 1 to 5 ;
  ArrayAuto(Counter) = %char(Counter) ;
endfor ;
dsply ('2. Auto = ' + %char(%elem(ArrayAuto)) ) ;

> DSPLY  2. Auto = 5

As before I can set the number of array elements to another number just by using the %ELEM BiF:

%elem(ArrayAuto) = 0 ;
dsply ('3. Auto = ' + %char(%elem(ArrayAuto)) ) ;

> DSPLY  3. Auto = 0

A new keyword that can be used with the expanding array has been introduced, *NEXT. Rather than having to give the number of the next available array element I can use, the *NEXT does that for me. For example:

%elem(ArrayAuto) = 0 ;

for Counter = 1 to 5 ;
  ArrayAuto(*next) = %char(Counter) ;
endfor ;
dsply ('8. Auto = ' + %char(%elem(ArrayAuto)) ) ;
dsply ('9. Auto(5) = <' + ArrayAuto(5) + '>') ;

> DSPLY  8. Auto = 5
> DSPLY  9. Auto(5) = <5>

The statement in the For group uses the *NEXT for the next array element to use. At the end of the For group there are five elements in the array.

The most common types of arrays I use are data structure arrays. Can the number the elements in these expand when used?

01  dcl-ds Data qualified dim(*auto:9999) ;
02    Name varchar(40) ;
03    PlaceOfBirth varchar(50) ;
04  end-ds ;
05  dcl-s Rows uns(5) ;

06  exec sql DECLARE C0 CURSOR FOR
07             SELECT RTRIM(LNAME) || ', ' || FNAME,
08                    IFNULL(PLACEBIRTH,'*UNKNOWN')
09               FROM PERSON
10              ORDER BY LNAME,FNAME
11                FOR READ ONLY ;

12  exec sql OPEN C0 ;

13  dsply ('1. Elements = ' + %char(%elem(Data)) ) ;

14  exec sql FETCH C0 FOR 9999 ROWS INTO :Data ;

15  exec sql GET DIAGNOSTICS :Rows = ROW_COUNT ;

16  exec sql CLOSE C0 ;

17  dsply ('2. Rows fetched = ' + %char(Rows) ) ;
18  dsply ('3. Elements = ' + %char(%elem(Data)) ) ;

Line 1: The DIM keyword contains the *AUTO and the maximum number of elements I want this array to have.

Line 5: This variable will be used to contain the number of rows retrieved by the SQL statement. I have defined this variable as an unsigned integer, as I cannot retrieve a fraction of a row or a negative number of them.

Lines 6 – 11: The definition of the cursor I will be using to fetch rows from the table. I am using the IFNULL, line 8, in the Place of Birth column to convert any columns that have a value of null to something more user friendly.

Line 12: Open the cursor.

Line 13: I want to know the number of array elements in my data structure just before I fetch the results.

Line 14: I perform the fetch for the maximum number of possible array elements. In the past I have used a variable here that was defined with the number of data structure array elements, but in this case it will not work. You will see why in a moment. Therefore, I have had to "hard code" the number of elements, 9,999.

Line 15: I use GET DIAGNOSTICS to retrieve the number of rows fetched.

Line 16: Close the cursor.

Lines 17 and 18: Display the number of rows fetched and the number of elements in the array.

When I run this program I see:

DSPLY  1. Elements = 0
DSPLY  2. Rows fetched = 41
DSPLY  3. Elements = 41

This shows that before the fetch statement the array has no elements. 41 rows were fetched into 41 elements of the array. The allocating of the elements for the results was done automatically, without me having to do anything.

 

Compile time array

Compile time arrays, you know those ones where you give the list of values at the bottom of the program.

dcl-s ArrayCt1 char(1) ctdata dim(5) ;

** ArrayCt1
1
2
3
4
5

It has always been one of my pet peeves that the RPG compiler knows how many items I have in my compile time array, but I always have to give it in the definition statement.

Now that peeve can be laid to rest, as I can now define an compile time array thus:

dcl-s ArrayCt1 char(1) dim(*ctdata) ;
dsply ('1. Ct1 = ' + %char(%elem(ArrayCt1)) ) ;

> DSPLY  1. Ct1 = 5

Now I can add to or delete elements from the compile time array and not have to count how many elements there are and change the definition.

If you do define compile time tables this new way you will need to place them in your code before any other compile time arrays defined in the previous way. If not your program will not compile.

RNF0203: DATA FOR ARRAYS AND TABLES WITH DIM(*CTDATA) MUST
         PRECEDE ALL OTHER COMPILE-TIME DATA.

 

 

This article was written for IBM i 7.4, and will not work for earlier releases.

5 comments:

  1. "In the second line of code the %ELEM BiF has a second parameter, *KEEP, this stops the new elements, 11-20, from being initialized with the asterisk. Why you want to initialize the elements in an array with one value and then not want to do it for others is, well, beyond me."

    I'm afraid you have completely missed the purpose of *KEEP Simon. It does indeed preclude initialization - that's its whole raison d'ĂȘtre - to _retain_ any values currently in memory. Major reason for its existence is that variable length arrays cannot be passed as parameters currently (i.e. they can't be defined in a proto) this may be a temporary limitation but it may never be changed. So suppose you use one as a parm (you define it as *Varsize) and the called routine _adds_ elements. That is fine but when the called routine returns you need to set the %Elem count to the new current value _but_ must preserve the values added. _That_ is what *KEEP is all about. You are not likely to use it under any other circumstances as there is no guarantee as to what the new elements will contain - could be any old rubbish left lying around.

    I have described this in detail in an article that will shortly be published by IBM Systems Mag. Should be out in a couple of weeks.

    ReplyDelete
  2. Always good stuff

    ReplyDelete
  3. Big minus:
    It's not applicable on nested datastructures.

    ReplyDelete
  4. Thanks for the info! When I create a variable length array, is the entire memory for the entire array allocated at runtime? For example, if I create a DS array for transactions and specify dim(*auto:1000), will the job dynamically allocate memory as the array grows?

    ReplyDelete

To prevent "comment spam" all comments are moderated.
Learn about this website's comments policy here.

Some people have reported that they cannot post a comment using certain computers and browsers. If this is you feel free to use the Contact Form to send me the comment and I will post it for you, please include the title of the post so I know which one to post the comment to.