Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Kevin Lamping
    @klamping
    sometimes it's quicker to write a throw away test
    and sometimes not worth the effort, esp. if it's a new product and the requirments are still in flux
    Clint Hoagland
    @vsComputer_twitter
    "It depends," right? There's def gonna be cases where you want a throwaway test, nothing wrong with that.
    Personal example: I write "unit tests" that use real concrete things that hit the outside world (e.g. file systems, web APIs etc) just so I know what spits out of them so that I can accurately characterize them in mocks.
    I throw them out once I know how they work and the mocks are in place.
    Natalie J. Bennett
    @njbennett
    Tests can serve different purposes:
    1. Design aids
    2. Documentation
    3. Regression checking
      And there's at least one more that I'm not thinking of at the moment... but the point being, a test that serves #1 but doesn't help #2 or #3 much can be a great test to write and then an equally great test to delete.
    Rey Winters
    @reywinters03
    I'm okay with throwaway tests as long as people don't depend on them. Essentially you're providing saas for people ..
    Kevin Lamping
    @klamping
    that's the trick though. anytime you put anything useful out there, people will start depending on it :)
    Clint Hoagland
    @vsComputer_twitter
    I really like the point Natalie made about using tests to help design your code and then realizing that it's ok to delete them if they don't provide any regression safety
    Natalie J. Bennett
    @njbennett
    What do you mean by "depend on them?" What does it look like to depend on a test?
    Kevin Lamping
    @klamping
    just that they're using it for validation of something
    Clint Hoagland
    @vsComputer_twitter
    I think if a test is telling you something you still want to know, it's implicitly not a throwaway test.
    In my mind a throwaway test is something you just slap together to see how something is working and after you know you don't care anymore because you assume -for whatever reason- that it isn't going to change.
    Clint Hoagland
    @vsComputer_twitter
    (or if you don't care whether it changes I guess)
    Rey Winters
    @reywinters03
    Story time! So there was a developer who created a demo server to show off the newly created framework. Management decided (because of a time crunch) to use this demo server as the actual server. Needless to say it couldn't support a lot of users on at the same time.
    Clint Hoagland
    @vsComputer_twitter
    Stories of Protoduction with Rey Winters
    Rey Winters
    @reywinters03
    That said ... Throwaway tests should be separated and not publicized or at least explicitly stated that they are not written with best practices in mind :)
    Haha
    Clint Hoagland
    @vsComputer_twitter
    OK so what I'm hearing is not that these are what I think of as "throwaway tests" but rather tests that are either bad themselves, or that test something that is itself bad(?)
    Rey Winters
    @reywinters03
    Both...?
    but that's a good point ... What does a throwaway test look like to you
    Clint Hoagland
    @vsComputer_twitter
    http://content.screencast.com/users/CH_JingTesting/folders/Snagit/media/1832c601-20c3-4a5c-bdc7-a20fafb9e947/10.05.2016-21.06.png
    I wrote a test to tell me how many directories get returned by Directory.GetDirectories if I point it at my C drive. I did this only to find out how that method works; incidentally I found out how many (98). Now I can throw that test away because it's not testing anything I care about.
    what - in my opinion - you're describing as throwaway tests are what Justin Searls - in this very excellent talk from RubyConf (http://blog.testdouble.com/posts/2015-11-16-how-to-stop-hating-your-tests.html)
    ...described as "crappy tests"
    Clint Hoagland
    @vsComputer_twitter
    Not sure of the exact minute in the presentation he said it but he described what he does about them: moves them to a folder called "crappy-tests" and then replaces them with good tests when he has time.
    Rey Winters
    @reywinters03
    Ahhh haha yes! To me your throwaway tests are utility scripts
    Kevin Lamping
    @klamping
    @reywinters03 that story makes me crack up, because I can totally see it happening
    Rey Winters
    @reywinters03
    @klamping it was based on a true story ... I believe the actual upper bound limit of users were two 😂
    Kevin Lamping
    @klamping
    haha
    Rey Winters
    @reywinters03
    How many of you have or had created testing frameworks that ultimately were not used by the people you created it for ...?
    Kevin Lamping
    @klamping
    :raised_hand:
    was shut down after a year of development and some use
    Clint Hoagland
    @vsComputer_twitter
    <------- this guy
    Rey Winters
    @reywinters03
    Sigh me as well ...
    Rey Winters
    @reywinters03
    Does anyone here go into the SUT source code and change it to be more testable?
    Natalie J. Bennett
    @njbennett
    "Does anyone here go into the SUT source code and change it to be more testable?"
    All the time!
    I don't like to play the "best practice" card, but it's tremendously valuable. Often a small change in the production code can eliminate giant terrible swaths of test code.
    Kevin Lamping
    @klamping
    agreed. as a front-end dev, I don't mind an ID here and there to make testing easier
    Clint Hoagland
    @vsComputer_twitter
    This is a theological discussion I've argued with many many devs but my position is that code that is more testable is implicitly better and if you can change code to make it more testable it will make the code better. Full stop. So you should.
    Kevin Lamping
    @klamping
    yep
    i can see purists fighting back, but I think it's overwhelmingly worth the drawbacks
    Natalie J. Bennett
    @njbennett
    What kind of arguments do devs have against changing code for test ability?
    I work in a hardcore TDD shop so that viewpoint seems utterly alien
    Kevin Lamping
    @klamping
    well, from an html standpoint, you're adding IDs that don't serve a purpose other than for testing
    in theory that clutters up the code and increase page size
    but that's a pretty far stretch IMO
    Clint Hoagland
    @vsComputer_twitter
    IME bcz it breaks their OO design and "exposes class internals," usually because I want to do shit like inject a mock
    Natalie J. Bennett
    @njbennett
    @vsComputer_twitter ... have they heard about dependency injection?
    Clint Hoagland
    @vsComputer_twitter
    They have and they're against it. (Those that fight me on it)