Using identity fields in RPG requires a little SQL, but that SQL can coexist with traditional record-level access, as this monitor program illustrates.

In my last two articles (1, 2), I presented a business requirement that could be met by creating a data model with an identity field. Identity fields are very powerful, but they are best handled using SQL. This particular program makes use of a number of different ILE techniques, so we should dive right in.

The Definitions

If you review the previous articles, you'll know that we have two files, SQLSTM and SQLJOB. In this program, the I/O to SQLSTM will be done via SQL, while SQLJOB will be maintained using record-level access. The program uses some data structures as well, and I'll explain why as we go along. But let's get through the definitions.

     ctl-opt actgrp(*new) option(*srcstmt:*nodebugio);

     // All I/O to this file is through qualified DS

     dcl-f SQLJOB keyed usage(*update:*output);

     dcl-ds ioSQLJOB extname('SQLJOB':*all) qualified end-ds;

     // These are used to define all the fields for SQL

     // The fields are unqualified

     dcl-ds dsSQLSTM extname('SQLSTM') end-ds;

     dcl-ds dsSQLJOB extname('SQLJOB') inz end-ds;

     // Monitor DS to bring job down

     dcl-ds dsSQLMON len(1) dtaara('SQLMON') end-ds;

     /COPY QCPYLESRC,DEFUTL

Several things are defined here. First, of course, is control option. I tend to use *NEW for the activation group for most utility programs, and the *SRCSTMT and *NODEBUGIO are there for debugging purposes. Next are the definitions used to update the detail file, SQLJOB. This file is maintained via record-level access, so I need a file declaration. I also use a data structure for all I/O, and you'll see why shortly.

The next two data structures are defined solely to provide work fields for my SQL statements. One of the great benefits of externally described files is the fact that you don't have to define all the fields yourself. But by taking advantage of externally described data structures, I at least don't have to define all the fields individually. Please note that these fields are unqualified; that's merely to keep my SQL statements from getting too verbose.

Next, I define a data area. Since this program sits in a loop, I wanted a clean way to exit, and a data area is one of my preferred methods. Finally, I have a quick /COPY that brings in the prototype for the sleep() function.

The Main Loop

     exec sql set option commit = *none;

     init();

      dow dsSQLMON = '';

      processJobs();

      sleep(60);

     in dsSQLMON;

      enddo;

     *inlr = *on;

This is the whole mainline. It's quite simple: I run an initialization routine and then sit in a loop until the data area is set. The loop process all the current ODBC jobs and then sleeps for 60 seconds. After I’m done, I set on *INLR and get out. One of these days, I'll get around to using NOMAIN in my utility programs. I should do that sooner rather than later, but I just haven't done it yet. One other thing: you may notice the EXEC SQL at the beginning; this allows me to use SQL statements without the trappings of commitment control.

Initialization

      dcl-proc init;

     in *lock dsSQLMON;

      dsSQLMON = *blank;

     out dsSQLMON;

     exec sql declare c cursor for

        select JOB_NAME, CPU_TIME, TOTAL_DISK_IO_COUNT, JOB_USER_IDENTITY,

                CLIENT_IP_ADDRESS, CHAR(SQL_STATEMENT_TEXT, 500)

        from table(ACTIVE_JOB_INFO(JOB_NAME_FILTER=>'QZDAS*',

                                    DETAILED_INFO=>'ALL'))

          where SQL_STATEMENT_TEXT <> ' ';

      end-proc;

The initialization routine is simple. I just clear the data structure used to end the program. As a habit, I usually store my non-executable SQL cursor definitions here as well just to keep them out of the way; they can be quite large. This one is not large, but it does use a pretty cool DB2 Service, ACTIVE_JOB_INFO. I'm not going to go into it in detail, but the first salient point is that my query returns the fully qualified (with slashes) job name, the total CPU time, the total disk I/O, the current user for the job (very important in server jobs), the IP address of the initiating job, and the first 500 characters of the SQL statement being executed. The other point is that I limit it to the QZDAS job because those are my external ODBC jobs, and I only include jobs that are actually executing SQL. Quite a powerful little query!

Main Processing

Processing is simple:

• Get the next statement.
• Create a new SQLSTM record if necessary.
• Create or update the SQLJOB record.

      dcl-proc processJobs;

     dcl-s ni int(5);

     exec sql open c;

     exec sql fetch from c into

        :SJFQJN, :SJCPU, :SJDISK, :SJUSID, :SJCLIP:ni, :SSSTMT;

     dow SQLCOD = 0;

I start by opening the cursor and fetching the first record into the fields defined in the two unqualified data structures. As I said earlier, I like to use unqualified fields just because it reduces the clutter in the SQL statement. One thing to be careful about is null-capable fields. There are a couple of ways to handle them, but perhaps the easiest is just to create a null indicator (the ni field) and use it wherever a value can be null. As it turns out, only one field can be null; that's the IP address, SJCLIP.

        // Find statement ID, if not found add new one and get ID

        exec sql set :SJSQID:ni =

          (select SSID from SQLSTM where SSSTMT = :SSSTMT);

        if ni < 0;

          exec sql insert into SQLSTM (SSSTMT) values(:SSSTMT);

          exec sql set :SJSQID = identity_val_local();

        endif;

This is the code that handles the identity field. It's so very cool. In a moment, I'm going to write or update an SQLJOB record, and to do that, I need the ID associated with the SQL statement. The statement is currently in the field SSSTMT, and I need to either get the ID if that statement is already in the file, or add it to the file and get the ID of the newly added record. The code is not complex. First, try to get the ID of a matching record using the SET syntax (this is a variant of SELECT INTO, and I prefer it because it's more like an EVAL). A matching record will set SJSQID to that record's ID. Otherwise, the SET returns a null and the ni indicator will be set to a negative value. If that occurs, I have to do two things: add a record and get its ID. The INSERT statement adds a record. Note that it only adds the SSSTMT field; the SSID field is the auto-generated identity field, so I don't specify a value. The system generates it. But that means I have to retrieve that value for the following SQLJOB I/O. That's what the identity_val_local() function does; it returns the last ID value generated.

So to recap: either I get the ID of a matching record or I add a new record and get the ID that the system generated for it.

        // Now find job, and either update or add

        chain ( SJFQJN: SJUSID: SJCLIP: SJSQID) SQLJOB ioSQLJOB;

        if %found(SQLJOB);

          // If found, only update cumulative data and audit fields

          // Get current values for cumulative data

          ioSQLJOB.SJCPU = SJCPU;

          ioSQLJOB.SJDISK = SJDISK;

          // Audit fields

          ioSQLJOB.SJLCTS = %timestamp();

          ioSQLJOB.SJCHGS += 1;

          update SQLJOBR ioSQLJOB;

        else;

          // If not found, populate I/O record from unqualified fields

          // Initialize job values and cumulative data

          //   The only fields in dsSQLJOB with data are the ones that

          //   are populated by the FETCH

          ioSQLJOB = dsSQLJOB;

          // Parse fully qualified job name into components

          parseFQJN( ioSQLJOB);

          // Get ID for SQL statement

          ioSQLJOB.SJSQID = SJSQID;

          // Audit fields

          ioSQLJOB.SJCRTS = %timestamp();

          write SQLJOBR ioSQLJOB;

        endif;

This part of the logic is more familiar to us traditional record-level access programmers, with just a slight twist: I use data structures. But there's a method to my madness as we'll see. If a record exists, the logic is very simple: CHAIN to the record, update the fields, update the record. All I do is store the cumulative data and update the audit fields. The slightly trickier bit comes when I don't find a record; this is where I use data structures. Remember that the SQL statement fetches all that data into the dsSQLJOB data structure. So rather than move the fields individually, I can move everything all at once using a simple EVAL. Yes, it's lazy, but I'm a programmer. I thrive on lazy code, as long as it's architecturally sound.

Once I've moved all those fields (the fully qualified job name, IP address, etc.), then I can initialize the other fields. I parse the job name into individual fields, store the ID, and initialize the audit fields. Time to write the record!

        // Get the next record

        exec sql fetch from c into

          :SJFQJN, :SJCPU, :SJDISK, :SJUSID, :SJCLIP:ni, :SSSTMT;

     enddo;

     exec sql close c;

      end-proc;

The rest of the procedure is mundane. Fetch the next record, loop if one exists, otherwise close the cursor and exit.

Buttoning It Up

That's the end of the program, with one exception. Let me show you the parseFQJN function.

      dcl-proc parseFQJN;

     dcl-pi *n;

        uSQLJOB likeds(ioSQLJOB);

     end-pi;

     dcl-s slash2 int(3);

     // Find second slash (we know first is at 7)

      slash2 = %scan( '/': uSQLJOB.SJFQJN:8);

     // Parse the segments

      uSQLJOB.SJJOBN = %subst( uSQLJOB.SJFQJN: 1: 6);

      uSQLJOB.SJUSER = %subst( uSQLJOB.SJFQJN: 8: slash2-8);

      uSQLJOB.SJJOB = %subst( uSQLJOB.SJFQJN: slash2+1);

      end-proc;

This is the code to parse a fully qualified IBM i job name, such as 654321/JPLUTA/MYJOB. To parse the string, you have to know the positions of the two slashes. But here's the trick: you already know the position of the first slash; it's the seventh character, because job number is always six digits. So all you have to do is get the second slash. The %scan BIF does that and stores it in the slash2 variable. After that, it's just a matter of using %subst to break up the chunks.

And that's it for this program. With the exception of the sleep procedure (which is just a prototype to the system-supplied sleep function), this is all the code you need to implement this procedure. Enjoy!

Joe Pluta is the founder and chief architect of Pluta Brothers Design, Inc. He has been extending the IBM midrange since the days of the IBM System/3. Joe uses WebSphere extensively, especially as the base for PSC/400, the only product that can move your legacy systems to the Web using simple green-screen commands. He has written several books, including Developing Web 2.0 Applications with EGL for IBM i, E-Deployment: The Fastest Path to the Web, Eclipse: Step by Step, and WDSC: Step by Step. Joe performs onsite mentoring and speaks at user groups around the country. You can reach him at This email address is being protected from spambots. You need JavaScript enabled to view it..

---

MC Press books written by Joe Pluta available now on the MC Press Bookstore.

		Developing Web 2.0 Applications with EGL for IBM i Joe Pluta introduces you to EGL Rich UI and IBM’s Rational Developer for the IBM i platform. List Price $39.95 Now On Sale
		WDSC: Step by Step Discover incredibly powerful WDSC with this easy-to-understand yet thorough introduction. List Price $74.95 Now On Sale
		Eclipse: Step by Step Quickly get up to speed and productivity using Eclipse. List Price $59.00 Now On Sale

Also Read

SQL 101: DDL Recap—Improving Our Basic SQL Tables, Part 2

TechTip: Automate Complex Tasks Using SQL and User-Defined Table Functions

SQL 101: DDL Recap—Improving Our Basic SQL Tables, Part 1