cmd/jujud: factor out zookeeper test fixture

cmd/jujud: factor out zookeeper test fixture

The machine agent wants this stuff too.

https://code.launchpad.net/~rogpeppe/juju-core/jujud-zkfixture/+merge/111060

(do not edit description out of merge proposal)

[rog, 2012-06-19 16:53:22]

Please take a look.

[niemeyer, 2012-06-19 17:25:24]

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go
File cmd/jujud/suite_test.go (right):

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go#newc...
cmd/jujud/suite_test.go:24: type zkFixture struct {
Why isn't this following the standard conventions? FooSuite, SetUpSuite,
TearDownSuite?

[roger.peppe, 2012-06-19 19:57:00]

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go
File cmd/jujud/suite_test.go (right):

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go#newc...
cmd/jujud/suite_test.go:24: type zkFixture struct {
On 2012/06/19 17:25:24, niemeyer wrote:
> Why isn't this following the standard conventions? FooSuite, SetUpSuite,
> TearDownSuite?

Because it is not a suite: it has no tests. Hence I thought the usual Suite
methods would be misleading. I can use them if you would like.

[fwereade, 2012-06-19 22:49:03]

LGTM

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go
File cmd/jujud/suite_test.go (right):

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/suite_test.go#newc...
cmd/jujud/suite_test.go:24: type zkFixture struct {
On 2012/06/19 19:57:00, roger.peppe wrote:
> On 2012/06/19 17:25:24, niemeyer wrote:
> > Why isn't this following the standard conventions? FooSuite, SetUpSuite,
> > TearDownSuite?
> 
> Because it is not a suite: it has no tests. Hence I thought the usual Suite
> methods would be misleading. I can use them if you would like.

Agree rog; I think there's a valid distinction here. IMO it's reasonable to have
fixtures either/both for suites and for individual tests, and to insist on
either SetUpSuite or SetUpTest implies usage restrictions that do, or at least
*should*, not actually exist.

[dave_cheney.net, 2012-06-20 05:49:48]

LGTM.

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/provisioning_test.go
File cmd/jujud/provisioning_test.go (right):

https://codereview.appspot.com/6312051/diff/1005/cmd/jujud/provisioning_test....
cmd/jujud/provisioning_test.go:14: zkFixture
Nice use of embedding.

[rog, 2012-06-20 12:35:46]

On 19 June 2012 23:49,  <fwereade@gmail.com> wrote:
> Agree rog; I think there's a valid distinction here. IMO it's reasonable
> to have fixtures either/both for suites and for individual tests, and to
> insist on either SetUpSuite or SetUpTest implies usage restrictions that
> do, or at least *should*, not actually exist.

this is what i wanted to say, but more elegantly put.

[niemeyer, 2012-06-20 12:43:47]

On 2012/06/20 12:35:46, rog wrote:
> On 19 June 2012 23:49,  <mailto:fwereade@gmail.com> wrote:
> > Agree rog; I think there's a valid distinction here. IMO it's reasonable
> > to have fixtures either/both for suites and for individual tests, and to
> > insist on either SetUpSuite or SetUpTest implies usage restrictions that
> > do, or at least *should*, not actually exist.
> 
> this is what i wanted to say, but more elegantly put.

LoggingSuite, MgoSuite, HTTPSuite..

We can't have a convention every day.

[rog, 2012-06-20 13:07:10]

On 20 June 2012 13:43,  <n13m3y3r@gmail.com> wrote:
> On 2012/06/20 12:35:46, rog wrote:
>
>> On 19 June 2012 23:49,  <mailto:fwereade@gmail.com> wrote:
>> > Agree rog; I think there's a valid distinction here. IMO it's
>
> reasonable
>>
>> > to have fixtures either/both for suites and for individual tests,
>
> and to
>>
>> > insist on either SetUpSuite or SetUpTest implies usage restrictions
>
> that
>>
>> > do, or at least *should*, not actually exist.
>
>
>> this is what i wanted to say, but more elegantly put.
>
>
> LoggingSuite, MgoSuite, HTTPSuite..
>
> We can't have a convention every day.

I'm inclining towards the view that those are a mistake, coming from
the OO view of "is-a" relationships rather than "has-a" relationships.

The idea behind LoggingSuite, for example, is that by including it, the
including type somehow "becomes" a logging suite. I'm starting to
think that's a misguided view, and that perhaps those things should
change.

But for the time being, I'm ok to use those conventions.

[niemeyer, 2012-06-20 13:09:27]

On 2012/06/20 13:07:10, rog wrote:
> I'm inclining towards the view that those are a mistake, coming from
> the OO view of "is-a" relationships rather than "has-a" relationships.
> 
> The idea behind LoggingSuite, for example, is that by including it, the
> including type somehow "becomes" a logging suite. I'm starting to
> think that's a misguided view, and that perhaps those things should
> change.
> 
> But for the time being, I'm ok to use those conventions.

Quite possibly. I'm happy to discuss this in another forum. My point in this
review is only that we have a convention right now, and it should either be
followed or debated and changed consistently in a follow up.

[rog, 2012-06-20 15:19:05]

Please take a look.

[niemeyer, 2012-06-20 15:48:03]

LGTM, thanks.

[rog, 2012-06-20 16:49:37]

*** Submitted:

cmd/jujud: factor out zookeeper test fixture

The machine agent wants this stuff too.

R=niemeyer, roger.peppe, fwereade, dfc
CC=
https://codereview.appspot.com/6312051