posted 10-20-2002 04:22 PM         

Misconception of what automation is and how it can be beneficial is the biggest problem anyone faces. A proper expectation must be set with management and other affected parties (development and marketing). The best way to explain it is that automation needs to be treated just like regular application development. Even if you go and get one of the best test tools and add on a Keyword driven library you will have development time to build the data sets, driver scripts, error handling routines, and report engine. Even despite what some of the Keyword library advocates say it will take time.

[It would be beneficial for those of you promoting Keyword driven testing to specify how long it took you to implement this approach the first time.]

This will be on top of time to learn the tools and their functionality. Also, once all this is done you will have maintenance time to consider. Maintaining and debugging scripts is just as time consuming as regular programming, even more at times. But if you plan properly and build your automation system in a step by step process you will have better control.

Another thing to consider is resources allocated (both personel and equipment). You will need someone full time to do this, and have dedicated machines (controlled environment which development & other test staff are not allowed to touch). Doing it as a part time effort and on volatile platforms is just going to make the job more difficult and prone to failure.

Phew!.... I gotta stop here and let some other people chime in. I gotta go pick up a pizza. This is making me hungry.

Remember "automation is not automagic".

Best of luck to ya.

posted 10-20-2002 11:07 PM           

Learning curve of the tool is also one major area you have to consider while evaluation apart from the features ..

posted 10-21-2002 09:58 AM         

are you thrusted this job against your will ?? i see a lot of dislike in what you are doing.

posted 10-22-2002 11:42 AM         

Try browsing General Discussion or other areas of this forum, situations like what you are in, "Is the Industry Standard". You will find very few companies who have the "ideal" environment. Look at my posts, we seem to be in the same kind of situation. Size of the company need not necessarily mean it will have better processes.
Would do you mean by "lack of input and output control/analyzation". You as a tester have to create this process, if i understand it correctly.
"we have not been provided a testing DB". Does 'we' mean QA? if it does i understand you have to create your own DB for testing, populate the DB with test data depending on set criteria.

posted 10-23-2002 05:53 AM         

quote:

---

I am now getting facts to back up my stand that I cannot automate this, at least in the beginning.

---

The widely followed practice of treating Test Automation as the silver bullet still continues However here's something that will help boost your fact gathering.

Lessons in Test Automation : Elfriede Dustin
Success with test automation

Hope this helps

posted 10-23-2002 07:43 PM         

You can download Oracle 9i for free and use it leagally for internal development use. All you need is a reasonable P3 or better system with 512megs of RAM. This isn't too bad, since SDRAM is really cheap. (I'm sure you must have 1 MSDN subscription in house? If so, you can use the Win2K server.). If you are even more tight on budget or unix savvy, you can consider going with Linux version (make sure you don't run GNOME, or KDE GUI interface to save tons of memory and processing.).

So you have a Oracle database. Now what's the execuse of the company to not to allow you to copy the database? All you need is to export their database, and import it using the Oracle tool. All you need is enough disk space to export it, and copy it over and to your new space, and you are done. (Make sure you create a new tablespace and user for each one to make it clean...).

Whenever someone says "I can't", or "It's too difficult", I just do it. That is why I have a E250 with 2 gig of ram sitting next to me in the cube running Oracle. This beast used to sit in dev. mgr's office, and it got logged in once in the first year of it's ownership. And they said "We can't, since we need it for performance testing.". Well, I've had it for almost a year, and nobody has ever asked to use it for performance testing. All it took for me was "Can I just borrow it? I promise to return it, when you need it.".

Often, it's easier to ask for forgivness than permission. I live by Steve McConnel's "How to manager your manager" section. 6 months ago, my boss said NO to ant. Last week, he ordered me to use it...

As for testing pl/sql, you can consider using perl. Simply write a 'qx/sqlplus user\/password \@filename', then call a perl DBI module to execute the proper call SQL call to validate the data. If you want, you can even consider using something like JUnit using JDBC. Both of these tools are free, and will be able to easily integrate with build tools, thereby making nightly regression a possibility. This may be a good enough bait to get everyone interested and support your automation effort.

If your company is against automation, then lean toward free tools, that will easily integrate with the build process. Once you get the momentum going, you can get more budget and tools/apis to do more testing.

posted 10-25-2002 12:32 AM         

quote:

---

Originally posted by cnote:
Thank for your input. Everyone has been a great help.

---

Actually for unit testing PL/SQL you might consider utPL/SQL - a framework created by one of the greates PL/SQL experts Steven Feuerstein. You can download it from http://oracle.oreilly.com/utplsql/
The problem I see with unit testing is that it should be done mainly from the developers which is a rare case when the testing and QA is not a strategic company enterprise..nevertheless you have to try