1 Replies - 455 Views - Last Post: 27 September 2011 - 04:30 PM

#1 sbromley   User is offline

  • D.I.C Head

Reputation: 21
  • View blog
  • Posts: 127
  • Joined: 20-May 09

Model Based Testing

Posted 27 September 2011 - 03:28 PM

I've been interested in model based testing for a while. I still have some concerns in its practicality though. How scale able is it? I would want to be testing enterprise size applications. The first thought that comes to mind is just breaking things down into smaller finite state machines, but is that a good idea? How much of the process can be auto generated from code, say a Java class, a Spring project, or from a maven goal?
Is This A Good Question/Topic? 0
  • +

Replies To: Model Based Testing

#2 Martyr2   User is offline

  • Programming Theoretician
  • member icon

Reputation: 5227
  • View blog
  • Posts: 14,010
  • Joined: 18-April 07

Re: Model Based Testing

Posted 27 September 2011 - 04:30 PM

Well practicality is always going to be dependent. There are usually a bit of an initial investment to setup such automation testing and if you are testing something super simple then is it really practical? If it was an entire system, then yeah for sure it is practical.

It is always a good idea to break down problems and complex systems into smaller testing units. That is the whole idea behind unit testing. Test a class or a subsystem with a test harness program etc. The idea is that you can then build it up to more complex systems with some level of reliability that it is going to work. Of course you then move to integration testing between tested sub systems. At this point I would then be looking at model based testing.

As for auto generated from code, there I am not a huge fan. Anything auto generated from code I find can be buggy. Simply because all the nuances of reading the code, understanding what it is attempting to do and then building a test that fully encompasses the range of conditions is a hard thing to do for a computer. Sure a range check may be easy to spot and test for, same with an off by one error, but what about fully understanding that 87 million miles is not representative of distance between Earth and the Sun (which is closer to 93 million miles of 1 AU)... it doesn't even know what Earth and the Sun are.

I think model based testing is a good tool, especially for processes that are very clean and clear cut (financial systems etc) but obviously it has its drawbacks. Either way it will need to always be reviewed in the end to know if the testing criteria is complete enough and if the results are clear enough.


This post has been edited by Martyr2: 27 September 2011 - 04:33 PM

Was This Post Helpful? 1
  • +
  • -

Page 1 of 1