Loading ...
Sorry, an error occurred while loading the content.

Re: [XP] Agile requirements

Expand Messages
  • David Chelimsky
    ... I ve generally operated this same way, but I ve been having some second thoughts. I was going to just respond here, but it got somewhat verbose so I posted
    Message 1 of 21 , Dec 2, 2005
    View Source
    • 0 Attachment
      Jim Shore wrote:
      > I would suggest using NUnit to test everything that you as programmers
      > feel should be tested. I don't see Fit as a testing tool.
      >
      > I use Fit to provide examples of complicated requirements. I don't try
      > to test everything with Fit; I mainly just focus on examples of domain
      > logic. I only occasionally provide examples of UI interaction or data
      > translation: as a rule of thumb, I don't do it unless the UI interaction
      > or data translation is complicated or Fit would facilitate discussion
      > with non-programmers.
      >
      > I see NUnit and Fit as being orthogonal. They solve different problems
      > and it's not that important that they both end up comparing 'expected'
      > and 'actual' results.
      >
      > What's complicated about the application you're building? What's the
      > "secret sauce"--the magic know-how that your application provides that
      > no one else does? Provide Fit examples of that.
      >
      > I use TDD for everything, even if it has Fit examples too. When I write
      > my NUnit tests, I use different data than my Fit examples. I TDD from a
      > programming perspective... using data that reflects my knowledge of the
      > program's edge cases, zero-one-many scenarios, etc.
      >
      > Cheers,
      > Jim
      >
      I've generally operated this same way, but I've been having some second
      thoughts. I was going to just respond here, but it got somewhat verbose
      so I posted it:
      http://butunclebob.com/ArticleS.DavidChelimsky.WhyLimitFit.
      Coincidentally, that decision allowed me to better express some of the
      things I had started to write in an email.

      Looking forward to your (Jim's and everyone else's) comments either here
      or on the blog.

      Thanks,
      David
    • William Wake
      On 12/2/05, David Chelimsky wrote: [In his blog entry] ... This is a place I see good leverage. I see the difference when a Fit test
      Message 2 of 21 , Dec 2, 2005
      View Source
      • 0 Attachment
        On 12/2/05, David Chelimsky <david@...> wrote:
        [In his blog entry]
        >"I've noticed that these tables lead me and my team to classes that operate
        >directly on the basis of the described conditions." -

        This is a place I see good leverage. I see the difference when a Fit
        test is ready first and drives things - some object will directly
        represent the decisions in the table. When the table is done
        afterwards, there may or may not be such an object. (The correct
        behavior may be there, but arises out of the interaction of objects.)

        Having the "policy" object directly present in the system seems like a
        big benefit to me - it gives you a traceability that the system
        directly reflects the conditions.


        Like everybody else, it seems, we're also trying to hit the right
        balance between "tests as specification" versus "test as
        verification", and of "customer-written" vs. "programmer-written". I
        too find people easily fall into the "trap" of writing lots of
        verification scripts rather than expressing clearly and directly what
        they want to happen.

        --
        Bill Wake William.Wake@... www.xp123.com
      • yahoogroups@jhrothjr.com
        From: David Chelimsky To: extremeprogramming@yahoogroups.com
        Message 3 of 21 , Dec 2, 2005
        View Source
        • 0 Attachment
          From: "David Chelimsky"
          <david.at.objectmentor.com@...>
          To: "extremeprogramming@yahoogroups.com"
          <extremeprogramming.at.yahoogroups.com@...>
          Sent: Friday, December 02, 2005 8:38 AM
          Subject: Re: [XP] Agile requirements


          > Jim Shore wrote:
          >> I would suggest using NUnit to test everything that you as programmers
          >> feel should be tested. I don't see Fit as a testing tool.
          >>
          >> I use Fit to provide examples of complicated requirements. I don't
          >> try to test everything with Fit; I mainly just focus on examples of
          >> domain logic. I only occasionally provide examples of UI interaction
          >> or data translation: as a rule of thumb, I don't do it unless the UI
          >> interaction or data translation is complicated or Fit would
          >> facilitate discussion with non-programmers.
          >>
          >> I see NUnit and Fit as being orthogonal. They solve different
          >> problems and it's not that important that they both end up comparing
          >> 'expected' and 'actual' results.
          >>
          >> What's complicated about the application you're building? What's the
          >> "secret sauce"--the magic know-how that your application provides that
          >> no one else does? Provide Fit examples of that.
          >>
          >> I use TDD for everything, even if it has Fit examples too. When I
          >> write my NUnit tests, I use different data than my Fit examples. I
          >> TDD from a programming perspective... using data that reflects my
          >> knowledge of the program's edge cases, zero-one-many scenarios, etc.
          >>
          >> Cheers,
          >> Jim
          >>
          > I've generally operated this same way, but I've been having some second
          > thoughts. I was going to just respond here, but it got somewhat verbose
          > so I posted it:
          > http://butunclebob.com/ArticleS.DavidChelimsky.WhyLimitFit.
          > Coincidentally, that decision allowed me to better express some of the
          > things I had started to write in an email.
          >
          > Looking forward to your (Jim's and everyone else's) comments either here
          > or on the blog.

          I think there's an element here that really does need expressing: _choice_.

          I'd like to keep a separation between the external specification, which
          ultimately comes from the customers and which needs to be expressed
          in the terms natural to the business we're supporting, and the internal
          specification, which, while necessarilly related to the external
          specification,
          may be related in ways which may not be directly expressable.

          However, "may" does not equal "must". I've also noted that there are
          a class of tables which seem to map, very plainly and simply, to
          objects (or something) in the software, and where separate tests are
          not only redundant, they are a violation of the DRY principle: Don't
          repeat yourself.

          The biggest example I've got here (within FIT) is the mapping from
          labels to programming language identifiers. The unittest (unittest is
          Python's version of xUnit) code for this duplicates the table in
          Rick Mugridge's specification suite for ExtendedCamelCase.
          I've got other examples, and I've noticed the tendency to write
          some HTML, especially to drive fixtures from the doTable,
          doRows or doRow methods.

          I don't think there's a real way around this since so much of
          FIT revolves around the parse tree. I don't think I'd do this
          in most real world applications, since the closest they get
          to HTML is the FIT fixtures. The problem just doesn't occur
          unless you're either working on FIT or on an HTML part of
          a web application.

          Much of the rest is simply tool support. Living in a tool
          impoverished world gives me a somewhat different perspective.
          I don't, for example, have anything that lets me click on a line
          in a stack trace and go directly to the line in the code. I run
          all my tests from a script in a command window. This may be
          dark ages, but it does give me more leverage in running tests.
          If I want to run FIT in my "t all" script, then it's very simple:
          just add the necessary commands to the script. All I need
          then is a way to specify which tests are wired directly to
          functionality, and I'm almost home.

          As it turns out, it's relatively easy in FitNesse to say
          which tests to run: that's the virtual suite functionality.
          I've got the same functionality (plus more) in batch that
          will be in my 0.8 release. What's now missing is a way
          of looking at the output and pinpointing anything that
          went wrong (that is, stack traces).

          In counterpoint though, I think Jim's viewpoint has one
          really major plus: keeping things separate reduces coupling,
          and coupling impacts our ability to change. Do I really want
          the customer changing tables that I'm using as part of my
          program specification suite? Especially if the change impacts
          the design in a way that I can no longer use the table directly
          as a developer specification?

          John Roth


          > Thanks,
          > David
          >
          >
          >
          >
          > To Post a message, send it to: extremeprogramming@...
          >
          > To Unsubscribe, send a blank message to:
          > extremeprogramming-unsubscribe@...
          >
          > ad-free courtesy of objectmentor.com
          > Yahoo! Groups Links
          >
          >
          >
          >
          >
          >
          >
          >
        • David Chelimsky
          ... John - I appreciate the notion of keeping things separate. I don t want the customer changing tables that *I m* using as *programmer tests*. So that s a
          Message 4 of 21 , Dec 2, 2005
          View Source
          • 0 Attachment
            yahoogroups@... wrote:
            > In counterpoint though, I think Jim's viewpoint has one
            > really major plus: keeping things separate reduces coupling,
            > and coupling impacts our ability to change. Do I really want
            > the customer changing tables that I'm using as part of my
            > program specification suite? Especially if the change impacts
            > the design in a way that I can no longer use the table directly
            > as a developer specification?
            John - I appreciate the notion of keeping things separate. I don't want
            the customer changing tables that *I'm* using as *programmer tests*. So
            that's a problem if we're trying to reduce duplication - especially with
            the current lack of good FIT refatoring tools. I'd probably end up w/
            duplication between the customer's fit tests and mine :)

            However, I do get a little jealous of my customers who get to express
            their specs in this beautiful, simple, easy to glean tabular format when
            I feel sort of stuck w/ a less expressive tool (for certain problems) in
            xUnit. Obviously xUnit is extraordinarily expressive, and I have no
            interest in replacing xUnit w/ FIT as a unit testing tool. But I do want
            to be able to take advantage of FIT where it makes sense to do so.

            -David Chelimsky
          • Rob Park
            My 2 cents (or at least how we treat it) are that the details of your MVP, subsequent DAOs, and any of the other little individual things it must do are
            Message 5 of 21 , Dec 2, 2005
            View Source
            • 0 Attachment
              My 2 cents (or at least how we treat it) are that the details of your MVP,
              subsequent DAOs, and any of the other little individual things it must do
              are subjects for unit testing (aka programmer tests). Fit fixtures can be
              written to tap into your presenter (as just another view). Then you can
              right FitNesse tests (aka customer tests) to simulate and test important
              customer actions, which inevitably are very closely aligned with the story
              (more than the code itself).

              Hope that helps.

              .rob.

              >...
              >
              >Now we are starting to gather the requirements. My usual way, the XP way
              >is to plan the first relase first, by having a pile of user stories, then
              >adding details to them as acceptance tests, more when the iteration is
              >planned. However, writing FIT tests in fitnesse seems like a very nice
              >thing to do, but what confuses us is what should be automated as
              >unit/integration tests with NUnit/NMock and what should be automated as
              >Fit tests?
              >
              >For instance, obviously an application like this will have an order form,
              >order will have a list of order details. My way would be to build the
              >form, using TDD and the MVP pattern. At the same time, I think that I
              >could construct a test as a fit table, them make an adapter, and build the
              >part of the form to make the test pass, build a new one and so on until I
              >have m order form constructed. Fit tests are much more visual, but a lot
              >slower. What should I do? What should really be Fit tests and what NUnit?
              >Where's the line?
              >
              >Thanks,
              >Dan
              >
              >
              >
              >
              >On Thu, 01 Dec 2005 06:20:22 +0200, Jim Shore <jshore@...>
              >wrote:
              >
              > > I've just posted a new essay on my blog titled "How I Use Fit." It's
              > > based on my recent post to this group about Fit and TDD.
              > >
              > > As I look back, I realize that I've now closed the circle on a pretty
              > > substantial cycle of essays on agile requirements. I now cover
              > > everything from planning months of releases down to how to write a
              > > single section of Fit document, the work of an hour. Pretty cool.
              > > (Hey, wait! Shouldn't I demand big bucks for this stuff?)
              > >
              > > * "Beyond Story Cards" describes how I prefer to handle requirements
              > > over the course of developing an entire software product.
              > > * "Fit Workflow" describes how I use Fit to facilitate collaboration on
              > > requirements during a single iteration.
              > > * "A Vision for Fit" gives a concrete example of using Fit to facilite
              > > collaboration.
              > > * Describe-Demonstrate-Develop, in "How I Use Fit" describes how I
              > > prefer to develop the Fit document, fixtures, and how actual software
              > > development comes into play.
              > > * "Elaborating on a Theme," also in "How I Use Fit," describes how I
              > > structure Fit documents and their examples.
              > >
              > > Find links to these essays at
              > > http://www.jamesshore.com/Blog/Agile-Requirements.html
              > >
              > > Cheers,
              > > Jim
              >
              >
              >
              >--
              >Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
              >
              >
              >To Post a message, send it to: extremeprogramming@...
              >
              >To Unsubscribe, send a blank message to:
              >extremeprogramming-unsubscribe@...
              >
              >ad-free courtesy of objectmentor.com
              >Yahoo! Groups Links
              >
              >
              >
              >
              >
              >
              >
            • Rick Mugridge
              ... Hi David, I have no problem, in principle, in using Fit for unit tests and JUnit for storytests. It depends on the abilities of the human readers and the
              Message 6 of 21 , Dec 2, 2005
              View Source
              • 0 Attachment
                David Chelimsky wrote:

                > ...
                >
                >However, I do get a little jealous of my customers who get to express
                >their specs in this beautiful, simple, easy to glean tabular format when
                >I feel sort of stuck w/ a less expressive tool (for certain problems) in
                >xUnit. Obviously xUnit is extraordinarily expressive, and I have no
                >interest in replacing xUnit w/ FIT as a unit testing tool. But I do want
                >to be able to take advantage of FIT where it makes sense to do so.
                >
                >-David Chelimsky
                >
                >
                Hi David,

                I have no problem, in principle, in using Fit for unit tests and JUnit
                for storytests. It depends on the abilities of the human readers and the
                writers, and then on the form of the tests as to which is best. (I also
                use Fit for defining builds and other things besides.)

                I do agree that the Customers' storytests need to be kept separate from
                the programmers' Fit unit tests. I also want to keep separate the Fit
                storytests that define the PUBLISHED LANGUAGE (DomainDrivenDesign) for a
                subsystem. And also the storytests that are specific to installation and
                on-site maintenance. And etc.

                Cheers, Rick
              • Rick Mugridge
                Yes, I think they re just tackling different levels of the system. With good tool support, I could image using Fit for all my unit testing as well. I look
                Message 7 of 21 , Dec 2, 2005
                View Source
                • 0 Attachment
                  Yes, I think they're just tackling different levels of the system. With
                  good tool support, I could image using Fit for all my unit testing as
                  well. I look forward to trying this out.

                  Cheers, Rick

                  Steven Gordon wrote:

                  > ...
                  >
                  >And, as Dadi points out, xUnit supports the automated verification of TDD
                  >specifications, allowing them to also function as automated unit tests for
                  >any software that purports to satisfy those specifications.
                  >
                  >Likewise, Fit/Fitnesse supports the automated verification of Application
                  >Domain specifications, allowing them to also function as automated
                  >integration tests for any software that purports to satisfy those
                  >specifications.
                  >
                  >Two sides of two similar coins.
                  >
                  >Regards,
                  >
                  >Steve
                  >
                  >
                  >


                  [Non-text portions of this message have been removed]
                • Rick Mugridge
                  Hi David, Yes, I agree with your point of view and have seen it work very well in practice. With calculation rules like your validation table, it makes much
                  Message 8 of 21 , Dec 2, 2005
                  View Source
                  • 0 Attachment
                    Hi David,

                    Yes, I agree with your point of view and have seen it work very well in
                    practice. With calculation rules like your validation table, it makes
                    much sense to have them expressed directly in the code, taking a domain
                    driven design approach. I've noticed that tthe business logic often gets
                    little attention, and this helps bring a focus to it.

                    Of course, there still needs to be unit tests for driving the
                    infrastructure (non domain) parts.

                    Cheers, Rick

                    David Chelimsky wrote:

                    >Jim Shore wrote:
                    >
                    >
                    >>I would suggest using NUnit to test everything that you as programmers
                    >>feel should be tested. I don't see Fit as a testing tool.
                    >>
                    >>I use Fit to provide examples of complicated requirements. I don't try
                    >>to test everything with Fit; I mainly just focus on examples of domain
                    >>logic. I only occasionally provide examples of UI interaction or data
                    >>translation: as a rule of thumb, I don't do it unless the UI interaction
                    >> or data translation is complicated or Fit would facilitate discussion
                    >>with non-programmers.
                    >>
                    >>I see NUnit and Fit as being orthogonal. They solve different problems
                    >>and it's not that important that they both end up comparing 'expected'
                    >>and 'actual' results.
                    >>
                    >>What's complicated about the application you're building? What's the
                    >>"secret sauce"--the magic know-how that your application provides that
                    >>no one else does? Provide Fit examples of that.
                    >>
                    >>I use TDD for everything, even if it has Fit examples too. When I write
                    >>my NUnit tests, I use different data than my Fit examples. I TDD from a
                    >>programming perspective... using data that reflects my knowledge of the
                    >>program's edge cases, zero-one-many scenarios, etc.
                    >>
                    >>Cheers,
                    >>Jim
                    >>
                    >>
                    >>
                    >I've generally operated this same way, but I've been having some second
                    >thoughts. I was going to just respond here, but it got somewhat verbose
                    >so I posted it:
                    >http://butunclebob.com/ArticleS.DavidChelimsky.WhyLimitFit.
                    >Coincidentally, that decision allowed me to better express some of the
                    >things I had started to write in an email.
                    >
                    >Looking forward to your (Jim's and everyone else's) comments either here
                    >or on the blog.
                    >
                    >Thanks,
                    >David
                    >
                    >
                    >
                    >To Post a message, send it to: extremeprogramming@...
                    >
                    >To Unsubscribe, send a blank message to: extremeprogramming-unsubscribe@...
                    >
                    >ad-free courtesy of objectmentor.com
                    >Yahoo! Groups Links
                    >
                    >
                    >
                    >
                    >
                    >
                    >
                    >


                    [Non-text portions of this message have been removed]
                  • Dominic Williams
                    ... Surely, evolving an expressive and consise domain-specific language for the programmer tests is exactly the same exercise as what XP tells us to do with
                    Message 9 of 21 , Dec 6, 2005
                    View Source
                    • 0 Attachment
                      David Chelimsky started:

                      > However, I do get a little jealous of my customers
                      > who get to express their specs in this beautiful,
                      > simple, easy to glean tabular format when I feel sort
                      > of stuck w/ a less expressive tool (for certain
                      > problems) in xUnit.

                      and Rick Mugridge continued:

                      > With good tool support, I could image using Fit for
                      > all my unit testing as well. I look forward to trying
                      > this out.

                      Surely, evolving an expressive and consise
                      domain-specific language for the programmer tests is
                      exactly the same exercise as what XP tells us to do
                      with the code?

                      I am wondering what it says about our programming
                      languages or our ability to use them creatively that
                      programmers should be wanting to use FIT rather than
                      code for programmer tests...

                      Would you still want to if you were coding in Lisp or
                      Smalltalk or Ruby or something?

                      Anyway, even with C++ or Java, I don't think I'd want
                      to do programmer tests in FIT: when coding tests is
                      painful, I know I need to improve the design.

                      Regards,

                      Dominic Williams
                      http://www.dominicwilliams.net

                      ----
                    • Rick Mugridge
                      ... Please notice that I m not advocating this as a general approach! And I doubt that I ll end up wanting to change. But I think it s worth trying it. TDD
                      Message 10 of 21 , Dec 6, 2005
                      View Source
                      • 0 Attachment
                        Dominic Williams wrote:

                        >David Chelimsky started:
                        >
                        >
                        >
                        >>However, I do get a little jealous of my customers
                        >>who get to express their specs in this beautiful,
                        >>simple, easy to glean tabular format when I feel sort
                        >>of stuck w/ a less expressive tool (for certain
                        >>problems) in xUnit.
                        >>
                        >>
                        >
                        >and Rick Mugridge continued:
                        >
                        >
                        >
                        >>With good tool support, I could image using Fit for
                        >>all my unit testing as well. I look forward to trying
                        >>this out.
                        >>
                        >>
                        >
                        >Surely, evolving an expressive and consise
                        >domain-specific language for the programmer tests is
                        >exactly the same exercise as what XP tells us to do
                        >with the code?
                        >
                        >I am wondering what it says about our programming
                        >languages or our ability to use them creatively that
                        >programmers should be wanting to use FIT rather than
                        >code for programmer tests...
                        >
                        >
                        Please notice that I'm not advocating this as a general approach! And I
                        doubt that I'll end up wanting to change. But I think it's worth trying
                        it. TDD would be unknown if someone didn't try something that seemed
                        very odd to begin with. I often learn something from pushing the
                        boundaries, even if I just understand my assumptions better.

                        >Would you still want to if you were coding in Lisp or
                        >Smalltalk or Ruby or something?
                        >Anyway, even with C++ or Java, I don't think I'd want
                        >to do programmer tests in FIT: when coding tests is
                        >painful, I know I need to improve the design.
                        >
                        >
                        Fit could be useful is in setting up compex object structures, which can
                        be a real pain in Java. Ruby, of course, provides much better support
                        for this. But I personally prefer statically typed languages as a design
                        medium, having seriously tried a variety of types of language.

                        So, yes, it highlights the limits of the design of programming
                        languages, which are still (mostly) in the days of ascii.

                        Cheers, Rick

                        >Regards,
                        >
                        >Dominic Williams
                        >http://www.dominicwilliams.net
                        >
                        >


                        [Non-text portions of this message have been removed]
                      Your message has been successfully submitted and would be delivered to recipients shortly.