elementary-dev-community team mailing list archive

[Julien <spautz.julien@xxxxxxxxx>]

So if I'm not mistaken, TDD requires you to write tests before 
implementing something.

This might work well for teams with enough developers that don't have 
to rely on volunteers and drive-by contributors, but I don't see this 
working out for elementary, at least not at this point.

Many contributors fix a few small bugs here and there and don't have 
much experience with development, which is why we try to keep the entry 
barrier really low for new devs. When I started here, I had no 
experience with Vala, Gtk and Launchpad at all, and I just fixed some 
easy bugs, implemented some small features in various apps and so on.

If we used TDD, we would have to reject most code by new devs, because 
they didn't implement tests, and then we'll have to explain to them how 
to use tests, and then they'll have to write tests, but that's bad 
because you have to write them before you write the actual code ...

It's going to be a mess.

What might work are regular unit tests, implemented by people who kind 
of know what they're doing (we need some documentation for that, maybe 
add it to the dev guide). This will give us some of the TDD benefits 
w/o deterring potential new devs.

On Wed, Aug 21, 2013 at 3:29 PM, Craig <weberc2@xxxxxxxxx> wrote:
> > I have never heard a project die because they decided to move to 
> > TDD. And I heard about lots of hits on the matter.
>
> This. I've heard (and witnessed) a lot of projects drop their defect 
> percentage from 20% to 3%.  A lot of new-to-TDD developers don't 
> like it at first because it feels slower, but I don't think those 
> people remember how much time is spent bug hunting at the end of a 
> release cycle.
>
> I think the next step is to find a pilot project and get the lead 
> developer(s) to agree to work toward TDD. This will probably look 
> like:
>
> 1) Modifying the project structure to include _test directories
> 2) Creating tests with GLib.Test or some such
> 3) Coaching/mentoring the developers at TDD
> 4) Performing code reviews in which insufficiently tested code is 
> sent back for revision
>
> After testing it out for a few months, the community can see how they 
> like it.
>
> Does this sound like a reasonable approach? If so, would any project 
> leads like to volunteer their project? And who in the community would 
> like to participate in such an experiment?
>
>
> On Wed, Aug 21, 2013 at 4:34 AM, Alex Lourie <djay.il@xxxxxxxxx> 
> wrote:
> > Albert
> >
> > We don't need to exaggerate. Though TDD is, indeed, a development 
> > methodology, it is not supposed to completely change the way 
> > everyone works.
> >
> > Just consider that writing a good test takes about 10 minutes, and 
> > that each developer writes one or two tests for new stuff they add 
> > (or the old one they fix) from time to time. Then, in time, you'll 
> > have part of your code tested, which is exactly what we're aiming 
> > for. Beginning is hard, continue something is easier.
> >
> > I have never heard a project die because they decided to move to 
> > TDD. And I heard about lots of hits on the matter.
> >
> >
> > On Tue, Aug 20, 2013 at 3:37 AM, Albert Palacios 
> > <optimisme@xxxxxxxxx> wrote:
> > > Hi Craig,
> > >
> > > Thanks for your explanation and the linked video. I am still 
> > > agnostic, but I don't want to be the only one who complained about 
> > > the proposal. 
> > >
> > > At this point I have more questions than answers, and those will 
> > > probably be solved with a working example.
> > >
> > > But remember, TDD is a development methodology not a testing 
> > > methodology. It will change our development work flow, and will 
> > > probably move potential volunteers away from the project. 
> > >
> > > TDD like every other development methodology does not fit every 
> > > project or team. It can be a hit, and it can even the death of the 
> > > project.
> > >
> > > Albert
> > >
> > >
> > >
> > >
> > > On Tue, Aug 20, 2013 at 12:50 AM, Craig <weberc2@xxxxxxxxx> wrote:
> > > > Hi Albert,
> > > >
> > > > Thanks for your response, you asked a lot of great questions. In 
> > > > addition to Gufran's earlier response.
> > > >
> > > > >  Can you prove that there will be huge benefits in 
> > > > > time/resources?
> > > >
> > > > Well, that depends on what you consider "proof". In the spring, my 
> > > > company paid several thousand dollars to send me to a conference 
> > > > in San Jose in which many software development authorities 
> > > > recited, "test driven development pays itself off in iteration 0". 
> > > > That means the very first time you write the code it has already 
> > > > paid for itself (because even before you get the 
> > > > automatic-regression testing benefits, you've already got the 
> > > > benefits of a better architecture and documentation--because the 
> > > > tests _are_ the documentation!). I'm sure I can dig up lots of 
> > > > other resources, but I think it should suffice to say that I've 
> > > > never heard an expert comment on TDD except to say it's fastest 
> > > > way to develop quality software. For more information addressing 
> > > > your specific concerns, see the Wikipedia article's "benefits" 
> > > > section (and read on to the "shortcomings" as it is also 
> > > > relevant: http://en.wikipedia.org/wiki/Test-driven_development#Benefits. 
> > > > Even more importantly, this short 
> > > > video: http://www.youtube.com/watch?v=DodJQyHsmHI
> > > >
> > > >
> > > > > Can you prove that there will be less bugs? (looks like that if 
> > > > > tests are not right, bugs will populate equally).
> > > > It's pretty hard to write bad tests if you're practicing TDD, 
> > > > because you write the test first, watch it fail, insert the code 
> > > > you need to make it pass, and then hopefully watch it pass. If you 
> > > > wrote a bad test, it very probably will pass before you've written 
> > > > the code to make it pass (which serves to alert the programmer 
> > > > that his test is bad or his software is doing something 
> > > > unexpected) or the test will fail after he has correctly written 
> > > > the next line of code (which serves to alert the programmer to 
> > > > review both the code and his test and identify the source of the 
> > > > problem). For this reason alone, many, many bugs are eliminated.
> > > >
> > > > > From what's been said, looks like there will be an extra effort 
> > > > > on development, adding complexity and more tools to know (not to 
> > > > > say maintain).
> > > > Besides the initial learning curve, development actually goes 
> > > > _faster_ with TDD (see the aforementioned Wikipedia article--I can 
> > > > provide more resources on demand) because debugging time becomes 
> > > > exponentially more expensive as time passes after the bug has been 
> > > > introduced. This is because the bug can live anywhere in any code 
> > > > that has been added *since the last time the tests were run* and 
> > > > because the programmer will have an increasingly difficult time 
> > > > remembering the code he wrote at the time of the bug as time 
> > > > progresses. With TDD, you are running the tests after every change 
> > > > (generally you test every time you build), so as soon as you've 
> > > > broken something you find out about it. This means that the bug is 
> > > > guaranteed to live in the last change you made, which is a smaller 
> > > > sample and fresher-in-your-mind than changes you made weeks ago. 
> > > > Regarding your complexity concern, generally the process isn't 
> > > > complex (it's actually very simple) and it _simplifies_ 
> > > > development once you learn how to do it. The most complex part is 
> > > > figuring out how to integrate testing into the CMake project, and 
> > > > that's only complicated because CMake is complicated. Regarding 
> > > > tools, there are already testing tools available for Vala, 
> > > > including GLib (so we don't have to maintain anything). Anyway, 
> > > > testing tools don't take a terribly long time to learn.
> > > >
> > > > >  Can we focus on the half done things before adding new 
> > > > > projects? Granite is not ready, documentation is missing, not to 
> > > > > talk about the bugs that survived Luna release ... 
> > > >
> > > > TDD is more valuable the sooner you start implementing it. Even if 
> > > > you didn't write tests for old code and only started TDD with new 
> > > > code (and existing defects), you would be doing yourself a huge 
> > > > favor. I'm not suggesting that everyone stop what they're doing 
> > > > and go back and test every line of code (although it would be a 
> > > > good thing to chip away at over time), but practicing TDD on _new_ 
> > > > code can't hurt that much, can it?
> > > >
> > > > With that in mind, are there any arguments against TDD that 
> > > > outweigh its merits?
> > > >
> > > > Thanks again for your questions! :)
> > > > Craig
> > > >
> > > >
> > > > On Mon, Aug 19, 2013 at 12:32 PM, Albert Palacios 
> > > > <optimisme@xxxxxxxxx> wrote:
> > > > > Hi Craig and Gufran,
> > > > >
> > > > > I don't agree with TDD, and making a committee. Can you prove 
> > > > > that there will be huge benefits in time/resources? Can you prove 
> > > > > that there will be less bugs? (looks like that if tests are not 
> > > > > right, bugs will populate equally). Can you prove that creating, 
> > > > > modifying and fixing code is going to be easier?
> > > > >
> > > > > From what's been said, looks like there will be an extra effort 
> > > > > on development, adding complexity and more tools to know (not to 
> > > > > say maintain).  Can we focus on the half done things before 
> > > > > adding new projects? Granite is not ready, documentation is 
> > > > > missing, not to talk about the bugs that survived Luna release 
> > > > > ... 
> > > > >
> > > > >
> > > > > On Mon, Aug 19, 2013 at 7:27 PM, Gufran <dogabhai@xxxxxxxxx> 
> > > > > wrote:
> > > > > > Count me in.
> > > > > > I cant really put forward any point on how should we proceed but 
> > > > > > I'd definitely love to be with the team.
> > > > > > Yes, a testing committee is a good idea, maybe something 
> > > > > > independent of dev community. We can write scripts to automate 
> > > > > > tests, and we can do that in any language (Python for example), 
> > > > > > so it would be like a Python coupling between software and tests.
> > > > > > Just my two cents :)
> > > > > >
> > > > > >
> > > > > >
> > > > > > On Mon, Aug 19, 2013 at 10:43 PM, Craig <weberc2@xxxxxxxxx> 
> > > > > > wrote:
> > > > > > > This is cool and important, but I don't think it should stop 
> > > > > > > the discussion on test driven development. Perhaps this could 
> > > > > > > be a separate thread? It doesn't sound as though anyone is 
> > > > > > > opposed to TDD, so can we confirm that? And if no one is 
> > > > > > > opposed, how can we proceed? Can we start some kind of a 
> > > > > > > "testing committee" to help determine what testing steps are 
> > > > > > > needed, what framework to use, and how to integrate testing 
> > > > > > > into the existing project structure (i.e., using CMake)?
> > > > > > >
> > > > > > > Does this sound like a good plan? Thoughts?
> > > > > > >
> > > > > > >
> > > > > > > On Mon, Aug 19, 2013 at 12:05 PM, David Gomes 
> > > > > > > <david@xxxxxxxxxxxxxxxx> wrote:
> > > > > > > > I'll work on it, so far we only have this I 
> > > > > > > > made: https://dl.dropboxusercontent.com/u/19899464/reviewstutorial.html
> > > > > > > >
> > > > > > > >
> > > > > > > > On Mon, Aug 19, 2013 at 6:01 PM, Albert Palacios 
> > > > > > > > <optimisme@xxxxxxxxx> wrote:
> > > > > > > > > Hi Munchor,
> > > > > > > > >
> > > > > > > > > A contribution / bug fixing step by step guide is needed at 
> > > > > > > > > the developers site. There was a .pdf before the new site 
> > > > > > > > > change, but now it is impossible to find. 
> > > > > > > > >
> > > > > > > > > The problem with the old guide is that it encouraged to 
> > > > > > > > > create your own branch instead of using the 
> > > > > > > > > ~elementary-dev-community one (this is totally new for me). 
> > > > > > > > > Obviously, bazaar guides doesn't teach you on using the 
> > > > > > > > > "elementary-dev-community".
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > On Mon, Aug 19, 2013 at 6:57 PM, David Gomes 
> > > > > > > > > <david@xxxxxxxxxxxxxxxx> wrote:
> > > > > > > > > > I always tell people if they make their branches owned by 
> > > > > > > > > > ~elementary-dev-community I will volunteer to fix the code 
> > > > > > > > > > style myself. I have all the free time and the will to do 
> > > > > > > > > > it, just people always make their branches owned by 
> > > > > > > > > > themselves.
> > > > > > > > > >
> > > > > > > > > > ~David "Munchor" Gomes
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > On Mon, Aug 19, 2013 at 4:29 PM, Albert Palacios Jimenez 
> > > > > > > > > > <optimisme@xxxxxxxxx> wrote:
> > > > > > > > > > > Hi,
> > > > > > > > > > >
> > > > > > > > > > > Before talking about testing, and advanced development 
> > > > > > > > > > > techniques for teams with resources, there is one easy and 
> > > > > > > > > > > simple thing we can do to accelerate development.
> > > > > > > > > > >
> > > > > > > > > > > Sometimes (very often), bugs are stopped due spaces not 
> > > > > > > > > > > following the "code style" guidelines. Adding a "code 
> > > > > > > > > > > style" validator script before compiling, we can prevent 
> > > > > > > > > > > uploads with spaces at the end of the lines ...  and save 
> > > > > > > > > > > a lot of time.
> > > > > > > > > > >
> > > > > > > > > > > For example, just executing the next line before compiling:
> > > > > > > > > > >
> > > > > > > > > > > find . -type f -print0 | xargs -0 sed -i 's/[ \t]*$//'
> > > > > > > > > > >
> > > > > > > > > > > We will remove every "white space" at the end of any line, 
> > > > > > > > > > > including new lines with tab spaces.
> > > > > > > > > > >
> > > > > > > > > > > This can sound stupid, but it is absurd to block bug fixes 
> > > > > > > > > > > during several days due white spaces at the end of lines.
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > --
> > > > > > > > > > > Mailing list: 
> > > > > > > > > > > https://launchpad.net/~elementary-dev-community
> > > > > > > > > > > Post to     : elementary-dev-community@xxxxxxxxxxxxxxxxxxx
> > > > > > > > > > > Unsubscribe : 
> > > > > > > > > > > https://launchpad.net/~elementary-dev-community
> > > > > > > > > > > More help   : https://help.launchpad.net/ListHelp
> > > > > > > >
> > > > > > > >
> > > > > > > > --
> > > > > > > > Mailing list: https://launchpad.net/~elementary-dev-community
> > > > > > > > Post to     : elementary-dev-community@xxxxxxxxxxxxxxxxxxx
> > > > > > > > Unsubscribe : https://launchpad.net/~elementary-dev-community
> > > > > > > > More help   : https://help.launchpad.net/ListHelp
> > > > > > >
> > > > > > >
> > > > > > > --
> > > > > > > Mailing list: https://launchpad.net/~elementary-dev-community
> > > > > > > Post to     : elementary-dev-community@xxxxxxxxxxxxxxxxxxx
> > > > > > > Unsubscribe : https://launchpad.net/~elementary-dev-community
> > > > > > > More help   : https://help.launchpad.net/ListHelp
> > > > > >
> > > > > >
> > > > > > --
> > > > > > Mailing list: https://launchpad.net/~elementary-dev-community
> > > > > > Post to     : elementary-dev-community@xxxxxxxxxxxxxxxxxxx
> > > > > > Unsubscribe : https://launchpad.net/~elementary-dev-community
> > > > > > More help   : https://help.launchpad.net/ListHelp
> > >
> > >
> > > --
> > > Mailing list: https://launchpad.net/~elementary-dev-community
> > > Post to     : elementary-dev-community@xxxxxxxxxxxxxxxxxxx
> > > Unsubscribe : https://launchpad.net/~elementary-dev-community
> > > More help   : https://help.launchpad.net/ListHelp
> >
> >
> >
> > --
> > Alex Lourie