| || |
automated testing and hand held scanners
Currently I'm working on a project that is using automated testing to test a handheld scanner. They've hard wired the scanner so that the buttons are simulated as being pushed. They've gotten the project to the point of testing one large continuous testplan that in my opinion was piece mealed together and could be improved upon. The test is used to regression test the version previous to the version that was just released to the field and they are basically using it as a data creation tool and verifying that nothing in the old version breaks the new version of the database. There isn't any data validation set up though. So basically the scanner screens and button functions are tested and the test does a cliff dive into the hands of manually testing and checking the rest of the process including large amounts of data and a couple of reports.
My group is getting together to talk about what to do to improve the process and are actually considering dropping the automation altogether. I need some suggestions to go to the meeting with but I'm not sure what direction to go. I can't see that they would want to automate the future version because that would be too unstable and turn into a maintenance disaster. I would like to talk them into getting the automation done (or at least started) immediately after the product is sent to the field (close to being completely stable). What do you think?
Re: automated testing and hand held scanners
What's the purpose behind the testing?
Test that the scanner is reading BC's correctly?
That the application does the right thing based on the bar code scanned?
One approach (off the top of my head), might be to break the testing into the above pieces.
One part simply makes sure that the BC are being read correctly by the scanner. How to automate this? I am not sure since it needs to physically read something from somewhere. Possibly get a Krispy Kreme machine and engineer it to scroll the bar codes for the scanner to read *shrug*.
This part of the test can simply just check that the BC's that are being reported match the ones that are fed in. Hopefully the ones being fed in won't change too much, so you can predict the output.
But then the other piece would simply ensure that the program is doing the right thing given a set of bar codes. Feed the program a bunch of bar codes without getting them in via the scanner).
Feeding the bar codes programatically should ensure that the input will never change, and therefore, you should be able to predict the output easily.
depending on the situation, I would run these tests separatly to ensure that at each stage of processing, you are getting the right results.
Just something to think about.