RE: SQL update/insert for a string contain " ' "

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> -----Original Message-----
> From: Clapham, Paul
>
> You would suggest, I think, to replace each of these SQL statements by a
> call to an RPG program.  That would help out with the source management
> issue.  But I'm not entirely sure what that RPG program would look like;
> would I be writing a new one for each SQL statement I needed, or
> would there
> be just one program per table?

Actually, I break data access into categories: CRUD servers, QUERY servers
and TRANSACTION servers.

CRUD (Create/Read/Update/Delete) servers are designed to maintain master
files.  For a simple master file (Company Master?), the CRUD server only
accesses one file.  For a hierarchical data object with header and detail,
like an order file, the CRUD server handles all the related records.

QUERY servers are more complex.  They're designed to support retrieving
related information, using foreign keys.  For example, you can use a QUERY
server to retrieve order information, but it is smart enough to be able to
resolve the item description and include it in te returned data.

Finally, TRANSACTION servers are specifically designed to support
transaction events.  For example, a shipment event may need to relieve
inventory, post to shipping and update accounts receivable (depending on
your business).

In any event, if your SQL statements are maintaining hierarchically linked
files, they'd be a single CRUD server.  If they're querying a bunch of data
from different files, they'd could use a single QUERY server.


> I have heard that RPG in V5R1 will include the ability to create
> Java objects and call their methods.  So why shouldn't I write the program
> in RPG, access the database in the usual way, and have it create a Java
> XML-writer object and call the appropriate methods?  Not especially OO, I
> admit, but the object that encapsulates our 166-field inventory
> item file is a LONG way off.

I have heard this as well, but there are some reasons against it.  I've
talked with Blair Wyman about this, and the issue is that each job will have
its own JVM.  JVM startup is a lot of overhead, and you don't get to share
data with other jobs.  I'd much prefer any Java work to be done in a single
JVM that can be shared by all the jobs in my system.  I've asked that IBM
provide the ability to "hot link" to a specific JVM, but there are some
technical hurdles that make them none too keen on the idea.

Joe