I started a new job a few weeks ago. I forgot to mention it.
Sadly, you will not see as much posting from me.
I hope you can contain your disappointment.
Anyway what is interesting about this is I have been hired to help replace a huge document management system that I built 11 years ago for a large pharmaceutical company.
Imagine the nostalgia when I saw AWK and shell scripts written by me in 1995 and 1996.
Thrill to the idea that validation and testing documents still exist that I wrote so long ago; in Word Perfect.
With my name even!
Especially nice were the greetings I received from old colleagues.
Good ol’ awk. I once interviewed for a co-op position at Cisco
in the 90’s. One of the interviewers wanted someone to help
write sed and awk scripts for something. He said “all these
wippersnappers want to use perl, and I guess you can do that
if you have to.” I didn’t even know perl at the time, so I didn’t
get the job. To this day I’ve shunned sed and awk just to spite
him.
Here’s the tragic tale of how I became an avid AWK user.
The system in question had three main parts residing on HP UNIX boxes. Metadata
in an Oracle database, huge amounts of content and customizations written in C.
Through a twisted turn of events it came about that the system had to be moved to
a Solaris platform. I was given the responsibility of orchestrating the move.
This was my first real gig as a “comp sci” graduate at the ripe age of 35 and I didn’t want to blow it.
Because of the link between metadata and content I was forced to utilize the vendor supplied methods of transformation.
Needless to say, the testing environment proved a catastrophe.
However, through much anger and sweat I finally ran a successful test run and we were
set to go.
Now, at this time, one must remember, remote access had not reached today’s standards.
Therefore on the weekend of truth I, out of fear of failure, spent the entire weekend in my little cubicle watching the conversion. I slept on the floor. I played Empire on my PC while carefully ensuring no fluxuations occurred in the anticipated scheme.
On Monday morning workers would arrive. They’d log into the system to retrieve their SOPs and begin manufacturing the drugs upon which we rely.
On Sunday evening the entire process was complete. Everything worked perfectly…
…Except…the customizations written in C would not work.
On Monday morning workers would arrive. They’d log into the system and NOT retrieve their SOPS.
Secretaries would faint.
Workers would cry.
Stockholders would require coronary bypass surgery.
Production would grind to a halt.
I would be canned!!!
Yes, I had tested the C programs but for some reason they didn’t work in the real world!!!
What was I to do?
OMGWTF!!!
Quickly, late Sunday night I realized I didn’t have time to call the UNIX support dude and have the compiler fixed.
I decided to rewrite them all in AWK.
Fast, simple and quick.
Grabbing a famous O’Reilly book I learned AWK and rewrote everything.
I rewrote the testing scripts.
I rewrote the validaion scripts.
I singlehandedly slapped everything into production and let 'er rip.
It worked.
On Monday the workers came in and gathered their SOPs.
Secretaries enjoyed their coffee.
Stockholders didn’t raise an eyebrow.
A new job where the people are nice and the pay is better—that would be hard to beat. Congratulations!! Keep up with that bike riding and stay in touch.
I’ve done enough migrations from HPUX to Solaris to recognise the kind of problem. (mostly, different flavours of C, and manufacturers who don’t see why they should bother to ship a complete C compiler as standard.)
I occasionally see old programs dating back to 1998. That’s quite far enough. It’s the comments that are embarrassing. Although, heaven knows, it should be the lack of comments that embarrass.
I meant that you learned something new and applied it, rather than
curling up in a fetal position and whimpering, as I’ve seen people
do in similar situations.
Or different non-standard extensions, or compiler directives, or ??? For a while, it seemed each vendor regarded having non-standard extensions to their language implementation was a point of honor.
Back in the mists of time (or at least the mid 80s) I had a similar problem - the place I worked had 5 different embedded code products, averaging 20-30 K lines each of (very undocumented, and mostly uncommented) 6802 assembler that we were contractually required to maintain. And the assembler they were written for only ran on the PDP-11. (For non-computer people, think of it as being written in Chaucer’s English on hand-made paper - even by mid-80’s standards, it was archaic).
They finally retired our wheezy old PDP-11/20 (with 5 Mb disk packs, yet). And replaced it with a MicroVax running HPUX. Since I was the only engineer with any Unix experience, I was appointed administrator and put in charge of the great code migration. As mentioned, our original assembler wasn’t available, so locating a replacement was my first task. I was able to find a far better 6802 assembler, but the syntax and directives were different, and non-compatible.
I thought (briefly) about converting 150K lines of undocumented code by hand. As an experiment, I started a hand conversion of the smallest product to see how much work it was. After spending the better part of a day, it was obvious that 2-3 days per product was a minimum to convert and test. And the chances for introducing a conversion error were uncomfortably high.
So I took a couple of days and wrote a conversion shell script that filtered the source files through awk and sed. And had all 5 products converted and running by the end of the week. It firmly cemented my largely undeserved reputation as a Unix guru with my teammates.
When Perl became available, I largely switched over to it for new projects. But I still retain a lingering fondness for the Bourne shell, awk, and sed.
