The Lounge is rated PG. If you're about to post something you wouldn't want your
kid sister to read then don't post it. No flame wars, no abusive conduct, no programming
questions and please don't post ads.
I find unit tests helpful when initially developing some new functionality, as they allow me to focus on the new functionality and getting that working in isolation. Then I can look at integrating the new functionality when I know it works. And definitely agree with your point about regression testing. I think this is possibly the single most powerful reason to use them. If I make a change to the code, I want to be sure I have changed all the affected areas, and breaking unit tests gives me exactly that.
"There are two ways of constructing a software design: One way is to make it so simple that there are obviously no deficiencies, and the other way is to make it so complicated that there are no obvious deficiencies. The first method is far more difficult." - C.A.R. Hoare
I completely agree with the comment about regression testing.
You have to at least run the code to make sure it works right? Why not frame that test in a way that can be run again? If you have a decent framework set up and are familiar with how to use it then it won't take much longer anyway.
Then sometime down the road you have to modify it (internal enhancement for performance or change in functionality) and it gives you so much more confidence to make the change knowing that you had good coverage tests from before.
Do you proofread your own prosa? Or do you review your own bookkeeping?
Testing your own stuff is not such a great idea. You tend to build your misconceptions into the tests and sometimes you try to prove how good you are a little too hard. And these are just some unintentional reason why this can go all wrong.
"I don't know, extraterrestrial?" "You mean like from space?" "No, from Canada."
If software development were a circus, we would all be the clowns.
I've worked in numerous roles in IT so this method works for me. I'm destructive to software.
I started out in support at a medium sized place with a smaller software group (25 people). The dev sat a row over. I asked him once, "Hey, this customer makes a great point. Why does this do that?" GreyBeard: Shut up!
I then moved into QA. Once the devs were having a design session that leaked into the hallways.
Another support guy -- who really used the software to help users -- said, "Hey, I think you're forgetting about X."
GrayBeard Devs: Shut up! We will cross the bridge when we get to it. You are wrong about that design element. Support Guy: Wait guys. Seriously. Think about-- GrayBeards: Shut it! Support guy:
6 months later after new software release with this new design... Support guy: <hangs up phone> Remember that design element I mentioned? Well, the customer just fell into that hole. Graybeards: Shut! Up! Support Guy: I'll enter it as a bug in the system.
Later in my QA career, devs would say, "hey, monkey, go test and we'll give you a banana..." I entered a 10,000 character URL and hit post. It crashed Oracle instance!! Hahaha... (no, you cannot do that now, but you could in IE 2.x,3.x) Enter steps into bug tracker and submit. Heh, heh, heh. There's your treat, dev. Dev: What do you want me to do with that bug you entered. me: I don't care. It kills the Oracle instance and your app dies so it doesn't mean anything to me. Ignore it if you like. If you feel confident to go to prod with that. Dev: <sulks off into another direction. > For the win!!
I've worked with people like that, but I've also worked with people who were specifically hired as testers.
What I've seen is that the arrogant kind of dev you're talking about (and they exist just about everywhere) tends to be a lot more careful -- and involved! -- when they know that there is someone who will be confirming that everything works OK.
I know that the next comment my annoy some people here, but this is somewhere where Agile works exceedingly well. The morning stand-up ritual builds really good connections between devs and testers.
I wanna be a eunuchs developer! Pass me a bread knife!
I can actually see what you mean in the case of (true) Agile, because in that case the hand-off from dev to QA should be so tight that it is as if the hand-off were simply the one person who played the role as dev would now put on her QA hat and do the tests. In that case you would be right and that is really how it should work and that is again directly tied to ownership.
But, i assure you, long ago when that dev said that to me, it was entirely because testing was beneath him. Great conversation here though.
We're set up here now with SonarQube hooked into the CI system so if test coverage on new code is below 80% the change is automatically rejected (and it does reject changes).
We still get people saying "I don't understand why I need to test", which I hear as "I don't understand why I need to write higher quality code and catch issues earlier when they can be fixed more rapidly and where it's cheaper to do so".
I find that if you can set up a system to show unit tests being run, coverage information and actions being taken when issues are found then there tends to be more corporate buy-in. We had a senior manager here who wanted screens up so he could see the information each time he walked by the different teams.
Eagles may soar, but weasels don't get sucked into jet engines
It would never occur to me to give someone something attributable to me without making sure it works.
That's exactly how I feel about it. It was literally like listening to a crazy person when he said he hadn't even shot one element of data through the thing. Crazy! And, you're not old, you have a fine attention to The Craft of Software Development. (Also, I'm old too.)
If for nothing else, I find unit tests the quickest way to debug new features and check they actually work. Far quicker than manually clicking through UI controls until you get to your new feature / bug fix.
Although I have worked with devs who pass on to QA without even running through the code themselves . .
Absolutely! Great comment. Griping about unit tests and/or just not doing them is akin to griping about Source Control.
If you don't understand how much value Source Control brings as a developer and you only see it as overhead, then you probably don't really understand a lot about real dev. Source Control is annoying at times, but it saves so much work. Same thing for unit tests.
I've only been in the trade for 4 years so perhaps that is before my time. I can't even imagine working without source control. That includes personal projects, but particularly for collaboration.
What is the alternative to source control when working as a team? Emailing code around? Store on network drive? Word of mouth? Passenger pigeon? Reminds me of the time I received a printed XML in the post..
That's a good way to look at this and it sparked a thought in me. User Acceptance testers can decide whether the software meets the user requirements and bascially fulfills the outward contract. So the testing could determine that the software is 100% correct.
However, without unit tests the design underneath could be so terrible that as soon as the customer wants to add an "easy change" the design may be so brittle that everything falls apart. So unit tests and other tests can/may indicate where the design is broken or bad and those kinds of tests would never be found by acceptance testers.
Sure, programmers should do some unit and module testing.
But too often, programmers test their software when used the way it is supposed to be used. It is not like the "five year old test": Place a five year old by the keyboard and tell him "You just do what you want to do, and you'll have another ice cream cone every time you make the program crash".
Real testing must be done by "misbehaved" testers. People who are paid by the number of bugs they find, not by the number of code lines they write. 30 years ago, before software testing was an established discipline, we used student summer interns: If they stayed with us for more than half a year, they learned the programmer's way of thinking, using the software the intended way, and the number of bugs detected gradually. Every summer we got a new group of students, unfamiliar with the software, and the detected bugs count rose sharply.
Nowadays, we have a separate testing team - full time, but then: They are educated in the discipline of testing. Their profession is to identify corner cases, defining stress testing, managing bugs databases (programmer solution: add a comment in the code: "To do: Fix this bug when time allows" ), and rating the severeness of the bugs, making sure that the all the serious bugs are really fixed before the product is released.
Besides: Their job is to bark at the programmers when bugs are found that should never have been made at all. Programmers don't like to be barked at. So they do proper unit and module tests just as much to quiet down the testers as to satisfy the customers
I think the t-shirt I am wearing today is somehow related to my own program/bug-writing experience - it says "Experience - the ability to recognize a mistake when you repeat it". That certainly describes my experience when my code goes to the testers.
Great comments. I agree. There are definitely different types of testers and testing. as a matter of fact, the book I mentioned also touches upon these and how they are all different and should be considered.
I also find that if you want to make testable code, stay as "functional" in your programming as much as possible.
I recently had to rework a monolithic module. The final result was 4 independent modules (separation of concerns) that were pure functional (inputs, outputs, no side effects) which generated almost the same information, but it was a lot easier to test/debug/capture intermediate results/replay/prove/etc.
The functional/separated modules actually were a lot cleaner and faster (less iterations) since one data structure in the old monolithic was split into separate,focused structures in the new modules.
After the refactor was complete, then the new requirements were added with clearly observable/diff-able results in outputs of the affected modules.
Excellent commentary. My primary customer makes a lot of different industrial machines. One of the primary engineering tasks is testability - how do we know this thing actually works? Keeps the electrical and mechanical engineers busy.
No reason why the same principles could not be applied to the software.
Charlie Gilley <italic>Stuck in a dysfunctional matrix from which I must escape...
"Where liberty dwells, there is my country." B. Franklin, 1783 “They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Last Visit: 31-Dec-99 19:00 Last Update: 22-Jan-17 20:32