Agree with what's being said here, though I think that there's two separate
threads.
1. yes, TDD is the best way to develop applications. For Isis, I think
that integration tests work best, because it is "outside-in thinking", a la
the GOOS book (well worth a read if you've not come across it). Unit tests
are great for filling in the holes, ie the tricky algorithmic bits of
logic. I like to separate these out into domain services so that they can
be tested in isolation.
2. completely separately, it's also valuable to run up the app, then modify
the production code and see the changes through hot reloading. You might
be changing the production code because you've just written a failing
integration test. Or you might be spiking out/prototyping a new feature
with the end-users, figuring out what the requirements actually are.
There is always a risk that if you are prototyping new features without
tests (just exploring the domain) then you'll forget/not bother to write
any integration tests. The TDD purists would say that you should throw
away any production code not written as a result of a failing tests ("TDD
as if you meant it" [2]) , and that tests alone should drive out the
design.
With Isis though I don't think it's worth being quite so dogmatic.... for
myself I'm happy to put the tests in after. One reason for that is because
the framework provides plenty of feedback already as to whether the
requirements are met. Also, the framework provides some pretty powerful
design tools (such as mixins/contributions and the event bus) that if used
correctly are likely to keep the design clean. A tool such as Structure101
is a good way to ensure that the overall architecture is sound.
My 2c
Cheers
Dan
[1] http://www.growing-object-oriented-software.com/
[2] http://www.infoq.com/presentations/TDD-as-if-You-Meant-It
On 17 October 2015 at 08:43, Óscar Bou - GOVERTIS <[email protected]>
wrote:
> Hi to all.
>
> Just our experience.
>
> We currently have in production an Isis-based system with more than 4.600
> integration tests fully automated through Jenkins, giving us a code
> coverage over the 70% barrier normally used as reference in Silicon Valley.
>
> We currently have “no UI” testing at all, in all those screens
> automatically generated through the Isis meta-model.
> Simply it’s not needed ...
>
> And we have great confidence that every new commit passing successfully
> all integration tests (including the new ones introduced) is
> production-ready, as domain logic has been properly tested and the UI,
> simply, is not needed ...
>
> We cannot envision that in any other platform than Isis.
>
> So yes, please, my advise to all community is to advance on the
> integration tests path.
> We currently have a wonderful platform for that.
>
> Cheers,
>
> Oscar
>
>
>
>
> > El 17 oct 2015, a las 6:30, Stephen Cameron <[email protected]>
> escribió:
> >
> > yes, I assume so, they get enhanced only when I use a maven goal, my use
> of
> > JUnit/JMock is just within Eclipse and no so DN enhancement is is
> > occurring. Obviously I need to go another step and do integration testing
> > using enhanced classes too, and I will get to that very soon.
> >
> > If Naked Objects as a paradigm has merit, then the UI development,
> > including testing via the UI should, to my mind, be a minor part of the
> > process. l am finding things 'blowing up' in the UI, but those issues
> will
> > mostly be caught with integration tests I feel. I particular a kind of
> > error that appears to be a complex interaction between wicket, the Isis
> > meta-model and DataNucleus that can take some sorting out, its the 'array
> > index out of bounds' error.
> >
> > [also, my idea of separating into data and behaviour classes using
> > inheritance won't work, I should try to get better sleep!]
> >
> > On Sat, Oct 17, 2015 at 7:04 AM, Stephen Cameron <
> [email protected]
> >> wrote:
> >
> >>
> >>
> >> On Sat, Oct 17, 2015 at 6:53 AM, Stephen Cameron <
> >> [email protected]> wrote:
> >>
> >>>
> >>>
> >>> If you move to TDD and use JMock, which I am trying now, less of your
> >>> time is spent testing things in the UI.
> >>>
> >> and you are testing with unenhanced classes.
> >>
> >>>
> >>> So far my thoughts of prototyping in real time have not happened, maybe
> >>> one day.
> >>>
> >>> So class reloading is now less of an issue for me, but DCEVM sounds
> good.
> >>>
> >>> It seems to me that putting behaviour in services is moving away from
> >>> (Naked) Objects towards procedural code, that services are essentially
> >>> modules of procedures. I am not against this outright, quite the
> contrary,
> >>> but maybe there are downsides to consider?
> >>>
> >>> Another option is to split classes into an abstract data-model parent
> and
> >>> a concrete behaviour-model child using inheritance. Could this
> alleviate
> >>> the need for DN enhancement with each change issue?
> >>>
> >>> Steve
> >>>
> >>> On Sat, Oct 17, 2015 at 2:13 AM, Rade, Joerg / Kuehne + Nagel / Ham
> GI-PS
> >>> <[email protected]> wrote:
> >>>
> >>>> Hi Dan,
> >>>>
> >>>> changes to non-entity classes could be reloaded?
> >>>> I assume I have to rebuild the war. Correct?
> >>>>
> >>>>> One reasonable workaround is to use contributed actions, and/or event
> >>>> bus,
> >>>>> and have them in essence "decorate" the class. As your app gets
> larger
> >>>> you'll
> >>>>> probably want to be using these anyway. And the contributing services
> >>>> can be
> >>>>> reloaded by DCEVM without requiring any enhancing going on.
> >>>>
> >>>> Can you point me to an example? Or give more details?
> >>>>
> >>>> Thanks in advance
> >>>> Jörg
> >>>> -----Ursprüngliche Nachricht-----
> >>>> Von: Dan Haywood [mailto:[email protected]]
> >>>> Gesendet: Freitag, 31. Juli 2015 13:13
> >>>> An: users
> >>>> Betreff: Re: faster development in eclipse
> >>>>
> >>>> I've been meaning to do add some docs and a quick screencast about
> DCEVM
> >>>> for a while; maybe this weekend..
> >>>>
> >>>> To answer Martin's question: in fact, if running in prototype mode,
> then
> >>>> the Wicket viewer automatically invalidates the metamodel of the
> (class of
> >>>> the) displayed object... this is why editing the .layout.json file and
> >>>> reloading changes causes updates to be picked up automatically.
> >>>>
> >>>> What is more tricky is if the entity is changed itself, ie its
> >>>> structure/method signatures. If the internals are changed then
> regular
> >>>> Java Hotspot can mostly cope, and I suspect that DCEVM would make
> that more
> >>>> reliable still. But if the entity has changed then the class needs
> >>>> re-enhancing. Unfortunately DN doesn't provide any easy way to do
> that, so
> >>>> there's little point in Isis providing an API to invalidate a class
> if the
> >>>> underlying DN metamodel hasn't also been updated.
> >>>>
> >>>> One reasonable workaround is to use contributed actions, and/or event
> >>>> bus, and have them in essence "decorate" the class. As your app gets
> >>>> larger you'll probably want to be using these anyway. And the
> contributing
> >>>> services can be reloaded by DCEVM without requiring any enhancing
> going on.
> >>>>
> >>>> ~~~
> >>>> Alternatively you could dust off the JRebel plugin that I worked on
> >>>> [1]. I did get this working reasonably well for me, though I'm not
> sure if
> >>>> anyone else did. (And, having said that I haven't found myself
> reaching
> >>>> for it, but partly because I switch context between projects all the
> time).
> >>>> Anyway, perhaps worth checking out.
> >>>>
> >>>> HTH
> >>>> Dan
> >>>>
> >>>>
> >>>> [1] https://github.com/isisaddons/isis-jrebel-plugin
> >>>>
> >>>>
> >>>>
> >>>> On 31 July 2015 at 09:03, Martin Grigorov <[email protected]>
> wrote:
> >>>>
> >>>>> Hi,
> >>>>>
> >>>>> I use DCEVM for code reloading. It is free and there is nothing to
> >>>> learn.
> >>>>> But it doesn't help with updating Isis meta model.
> >>>>> It would be nice if Isis exposes REST/JMX/... interface that triggers
> >>>>> rebuild of the model.
> >>>>> @Dan: is this possible with the code of isis-mavel-plugin?
> >>>>> On Jul 31, 2015 7:45 AM, "Stephen Cameron"
> >>>>> <[email protected]>
> >>>>> wrote:
> >>>>>
> >>>>>> Hi again,
> >>>>>>
> >>>>>> After playing around with options I now think there isn't much
> >>>>> opportunity
> >>>>>> to speed things up, other than to use JRebel or change the way I
> >>>>>> develop (not start the webapp so much). The webapp project looks for
> >>>>>> jars in the maven repository not the war, so I only need to do maven
> >>>>>> install on the
> >>>>> dom
> >>>>>> project when I change a dom class. That should have been apparent
> >>>>>> sooner
> >>>>> :(
> >>>>>>
> >>>>>> Bottomline: the webapp takes some time to start, but that is fine.
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>> On Fri, Jul 31, 2015 at 11:27 AM, Stephen Cameron <
> >>>>>> [email protected]> wrote:
> >>>>>>
> >>>>>>> Hi,
> >>>>>>>
> >>>>>>> I would greatly appreciate some advice on speeding up development
> >>>>>>> in Eclipse. Using the simpleapp project with separate dom,
> >>>>>>> fixture,
> >>>>>> integtest
> >>>>>>> and webapp subprojects.
> >>>>>>>
> >>>>>>> I'd at least like the webserver (jetty,tomcat) to restart when I
> >>>>> modify a
> >>>>>>> dom class. This is what I am used to with Dynamic Web Projects,
> >>>>>>> however after a bit of trying today I still cannot get that to
> >>>>>>> work, using
> >>>>>> Tomcat.
> >>>>>>>
> >>>>>>> It seems most are using JRebel with IntelliJ, but I am not so keen
> >>>>>>> to
> >>>>> go
> >>>>>>> there yet as I am (re)learning on many fronts at present.
> >>>>>>>
> >>>>>>> The issue with Tomcat is a jetty dependancy, this was noted
> >>>>>>> previously [1], but I could not get around it by removing
> >>>>>>> jetty-all-9.2.11.v20150529.jar as suggested, I then got a
> >>>>>>> null-pointer exception.
> >>>>>>>
> >>>>>>> At the moment I have to rebuild the whole project after a class
> >>>>>>> change
> >>>>> in
> >>>>>>> the dom-project then restart the server to see the change. Is the
> >>>>> because
> >>>>>>> the .war has to be rebuilt for mvn jetty:run to use the new
> >>>>>>> version of
> >>>>>> the
> >>>>>>> class file?
> >>>>>>>
> >>>>>>> I've just tried to use the eclipse jetty plugin but that is
> >>>>>>> telling me that the dom classes have not been enhanced by
> >>>> datanucleus.
> >>>>>>>
> >>>>>>> Hoping for some insights.
> >>>>>>>
> >>>>>>> Thanks
> >>>>>>> Steve C
> >>>>>>>
> >>>>>>>
> >>>>>>> [1]
> >>>>>>
> http://permalink.gmane.org/gmane.comp.apache.incubator.isis.devel/99
> >>>>>> 31
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>
> >>>> Kühne + Nagel (AG & Co.) KG
> >>>> Rechtsform: Kommanditgesellschaft, Bremen HRA 21928, USt-IdNr.: DE
> >>>> 812773878.
> >>>> Geschäftsleitung Kühne + Nagel (AG & Co.) KG: Reiner Heiken (Vors.),
> >>>> Dirk Blesius, Martin Brinkmann, Matthias Heimbach, Jan-Hendrik
> >>>> Köstergarten, Nicholas Minde, Lars Wedel, Jens Wollesen.
> >>>> Persönlich haftende Gesellschafterin: Kühne & Nagel A.G., Rechtsform:
> >>>> Aktiengesellschaft nach luxemburgischem Recht, HR-Nr.: B 18745,
> >>>> Geschäftsführendes Verwaltungsratsmitglied: Karl Gernandt.
> >>>> Geschäftsleitung Region Westeuropa: Yngve Ruud (Vors.), Richard Huhn,
> >>>> Björn Johansson, Jan Kunze, Bruno Mang, Stefan Paul, Holger Ketz,
> Dominic
> >>>> Edmonds.
> >>>>
> >>>> Wir arbeiten ausschließlich auf Grundlage der Allgemeinen Deutschen
> >>>> Spediteursbedingungen (ADSp), jeweils neuester Fassung. Wir verweisen
> >>>> insbesondere auf die vom Gesetz abweichenden Haftungsbeschränkungen
> von
> >>>> Ziffer 23 und 24 ADSp. Den vollständigen Text der ADSp übersenden wir
> Ihnen
> >>>> gerne auf Anfrage und können Sie auch unter
> http://www.kuehne-nagel.com
> >>>> einsehen. Ergänzend wird vereinbart, dass (1) Ziffer 27 ADSp im Rahmen
> >>>> internationaler Übereinkommen weder unsere Haftung noch die
> Zurechnung des
> >>>> Verschuldens von Leuten und sonstigen Dritten zu Gunsten des
> Auftraggebers
> >>>> erweitert, und (2) wir in den im deutschen Seehandelsrecht
> aufgeführten
> >>>> Fällen des nautischen Verschuldens oder Feuer an Bord nur für eigenes
> >>>> Verschulden und (3) im Sinne der CMNI genannten Voraussetzungen nicht
> für
> >>>> nautisches Verschulden, Feuer an Bord oder Mängel des Schiffes haften.
> >>>>
> >>>
> >>>
> >>
>
>
