Wednesday, November 19, 2014

Planning for trouble: comments on my latest Oracle Magazine article.

In my November/December 2014 article for Oracle Magazine, Planning for Trouble, I urge developers to realize that regardless of best intentions, not everything related to our apps is under our control, and we need to assume that trouble might be coming our way.

I received today, the following comments from Gary Malandro, which I thought you might enjoy reading:

Enjoyed your article in Oracle Magazine, and I have a few comments.

1.       You mentioned “Documents that spell out naming conventions…fit very nicely inside desk drawers”.  On our team, we have a number of policies that include the requirements for a technical design document, a software change request, source control, and some regarding style.  First thought upon hearing that is probably what-a-load-of-bureaucratic-nonsense.  Well, there are reasons for creating and enforcing these standards.  Compliance for one.  Another is we had code going into production systems that performed poorly, contained logic errors, was difficult to understand, already existed, no exception handling, etc.  With standards, we are able to peer review designs and code to make sure it is understood.  These are necessary when doing reviews otherwise it is a matter of style between developers – what 1 developer thinks is cool another finds difficult to understand.  Anyway, since we’ve implemented these policies our defect rate to test and error rate to production has improved considerably.  Besides PLSQL, we also develop in Java and .Net.  I’ve heard all the excuses as to the evil of standards and documentation (and probably said them many times myself), but the people who say this typically have to rework their code multiple times.  So pay now or pay later. 

2.       In my previous role I was supporting/developing a large volume of PL/SQL code.  I created a separate package (XXSA_DEBUG) which contained procedures to output data in various ways (a very, very, very poor man’s log4j).  In procedures I would include a way to get a debug variable and could either decide to call the output package, or just directly call it and let it decide where/how to output.   Production access is limited, so being able to change the debug flag using an application, or even running with a stub program, greatly aided when troubleshooting was required.

3.       Another thing we enforce is exception handling with clear messages.  If it’s something going to a log file or table (aimed for the developer) it should contain the procedure name and where the error occurred.  This allows any developer to trace where the error occurred.  Too many times we had log files loaded with “Failed”, and no explanation.  And thousands of lines of code that contained “Failed” for the output message. 

Thanks, Gary, for taking the time to share your experience. I would be very happy to hear from others!

1 comment:

  1. its really nice and awesome story i nice informative ,thanks for sharing for this Chicago Public Relations