DATA-GEN and DIM(*AUTO) Serendipity

[mlopez01]

Hello.

I came across something nice today. Whether it's intentional or not I don't know, and maybe Scott Klement can chime in since I'm using his YAJLDTAGEN.

Using a varying dimension array (DIM(*AUTO) I found that DATA-GEN will serialize based on the number of array entries actually used, thereby eliminating the need for using "countprefix".

Here's code:

h dftactgrp(*no) option(*srcstmt: *nodebugio: *noshowcpy)
h bnddir('YAJL/YAJL')
h decedit('0.')

dcl-s jsonDoc varchar(250);

dcl-ds names dim(*auto: 10) qualified inz;
  name varchar(50);
end-ds;

names(*next).name = 'John';
names(*next).name = 'Doe';

data-gen names
             %data(jsonDoc)
             %gen('YAJL/YAJLDTAGEN');

*inlr = *on;
return;​

The result is variable "jsonDoc" contains [{"name":"John"},{"name":"Doe"}]. No need for an additonal field (like num_names).

If I take out *AUTO so that it's just a DIM(10) then I get an additional eight empty names.

Pretty cool, if it's by design.

I couldn't find any documentation about this in the ILE RPG Reference but found this in Open Access guide for RPG (subfield totalElems looks like the key):

https://www.ibm.com/docs/en/i/7.5?topic=code-parameter-passed-data-gen-generator#roaDataGenParm__QrnDgArray_T

Any thoughts?

Mike

[Barbara Morris]

It's the way varying-dimension arrays work in general. If the current dimension of "names" is 2, then "DATA-GEN names" generates 2 elements.

If you didn't have *VAR or *AUTO for your array, you would use %SUBARR to control the number of elements generated.
DATA-GEN %SUBARR(names : 1 : 2) ...

[mlopez01]

Although varying-dimension arrays are a nice "feature" as it pertains to DATA-GEN it's very limited since a varying-dimension array can only be at the top level, and typical JSON docs (what I'm using DATA-GEN for) usually contain many nested levels.

Is there any talk around the water cooler about allowing nested varying-dimension arrays?