Programming quality control

[jonnysnow]

I have been asked to help develop guidelines for verifying the output of programming projects. (As an exercise, I was also asked to check another programmer's work on a specific project, and complete the same project and compare results.)

Since our office primarily works in SAS, I thought I would post in this forum to see what other people say-- what would you suggest, have you ever developed/worked with quality control guidelines, what has worked or not worked?

Any suggestions or references would be very helpful.
Thanks!

 

[ChrisW75]

A previous place I worked at had a pretty good set up. All work had to be run in batch, and log and list files kept.
The brief for the job should have specific deliverables mentioned that can be ticked off against.
At the completion of the project (or at stages throughout) a full walkthrough is performed with another programmer to make sure that the deliverables have been met, and that there are no problems in the code. Observation counts are useful here as you can keep track of that to check for inadvertant cartesian joins (multiple duplicates on both datasets) and things being dropped that shouldn't be.
I also wrote a short script using the GREP command which searched the log files for certain key words which would signify a problem for instance "ERROR:", "WARNING:" and the note you get when you perform a merge on 2 datasets with duplicates on both, I think it's something like "multiple values of the by variable found on more than one dataset". I think I also searched for "0 observations" to pick up cases where we ended up with no data in a dataset due to an incorrect filter.
The results of the GREP would be checked before the walkthrough with another programmer.
Code walkthroughs bring about another advantage in that other programmers get to see each others methods and can learn new techniques. It also means that all the knowledge for each job is not all just in one head, which is very important I think.
I hope that this helps.

Chris
Business Analyst, Code Monkey, Data Wrangler.
SAS Guru.