Comments on: 4 Building Blocks Of SQR Log Files

By: administrator

administrator — Tue, 28 Apr 2009 04:00:49 +0000

You’re right, that is something different. It seems interesting, but we could only use it for errors that the programmer detected. If the SQR interpreter finds an error, it won’t dump a stack for us. Perhaps some of the third party debuggers would.

By: Bob Josephson

Bob Josephson — Mon, 27 Apr 2009 18:28:25 +0000

I’m not sure we’re talking about the same thing. My idea was that, instead of writing a log command in each function or module, just push the name of the function on a stack (i.e. a growing array). Pop it when you exit the function. Then, when you want to log a fatal error, call a routine that logs all the function names on the stack before logging the error message.

By: administrator

administrator — Mon, 27 Apr 2009 16:26:32 +0000

Remember, Edward, we could write a program that met the users stated requirements perfectly, but failed when the user (or the database) fell out of conformance. I had a case where the original user specified that my program should produce a report for a given pay period. Then, a different user ran the report without entering a pay period. I learned a lesson about validating user inputs, but I learned it from the log file; the second user didn’t mention that she hadn’t entered a value.

By: administrator

administrator — Mon, 27 Apr 2009 16:21:03 +0000

SQR does have a facility somewhat like severity level. Any line can begin with “#debug”, and SQR will compile that line if there is a “-debug” parameter in the SQR command line. Further, a line can begin with “#debuga” (where “a” is a single letter or digit). Then the SQR command line can have something like “-debugab12″ which will cause SQR to compile any line beginning with #debuga, #debugb, #debug1, or #debug2.

SQR doesn’t have a “fail-only output” function. However, we could create a file and write to it. Then, the last thing we would do before the successful end of the program is to close it and delete it.

By: Edward deShelly

Edward deShelly — Mon, 27 Apr 2009 14:00:41 +0000

Bob, once again you’re aiming low and succumbing to lousy programming habits. You’re just resigning yourself to having bugs, and trying to CYA with error messages.

Maybe if you cared more about the quality of your work, you wouldn’t need to log so many errors.

By: Bob Josephson

Bob Josephson — Mon, 27 Apr 2009 13:58:14 +0000

It’s impossible to make it failsafe, so you might as well make it failsoft.

Is there any way to assign a severity level to a log message and specify at runtime what severity to display?

An even cooler feature would be where the program pushes each function entry onto a stack and pops it on exit. The stack only gets dumped if a severe message (i.e. an error) gets logged. That way, you have a stack trace for your errors, but you don’t create huge log files when everything is OK. (And even when an error happens, you have a succinct log file.)