Commits
-------

fc108e06 completely removed pimple as submodule

Discussion
----------

Pimple was not completely removed as a submodule

See https://github.com/fabpot/Silex/issues/158#issuecomment-4498365

Commits
-------

4a5c10e1 Grammar
e1264470 Fix URL to Symfony API doc
7a176af4 Cookbook recipe on how to use PdoSessionStorage

Discussion
----------

Cookbook recipe on how to use PdoSessionStorage

Related to this issue #254

---------------------------------------------------------------------------

by igorw at 2012-03-09T16:23:11Z

👍

Commits
-------

6445b9e9 Add Routes Middlewares

Discussion
----------

add "after($callback)" to Controller in order to mimic a "Route Middleware" feature in Silex

Hi !

As discussed in #262 and in my previous similar PR (https://github.com/fabpot/Silex/pull/270) with @GromNaN, here is another implementation of the Routes Middlewares feature, shipped with its Unit Test.

The feature addition is still 100% backward compatible, but this time it may be more "Silex style" compliant - and it's "100% no performance loss" :-)

With this PR the Silex ```Controller``` class has a new ```before()``` method, which registers a callback to trigger when the Route is matched, just before the route callback itself.
This ```before()``` method can be called more than one time on a single Route definition, allowing multiple secondary callbacks (a.k.a. "Routes Middlewares") to be executed before the Route main callback.

Would this feature be eligible for Silex integration, like this ?

Olivier

---------------------------------------------------------------------------

by DrBenton at 2012-03-01T22:16:05Z

* The method is now named ```addMiddleware```, which may be more explicit for people used to "a la Sinatra" framewoks like _Node.js Express_  or _Slim_.
* I no longer use the Silex ControllerResolver, but a dedicated Application protected service instead, fired just before the Application "before" process (I could have added a new listener on the ```Kernel```, but since we already have one for the right event for the "before" process, I used this one for simpler code).
* I also added some checks on the ```$request``` in the Unit Test.

---------------------------------------------------------------------------

by fabpot at 2012-03-02T11:40:30Z

Can you add some information about this new feature in the documentation (with some typical usages)? Thanks for your work on this feature.

---------------------------------------------------------------------------

by DrBenton at 2012-03-02T12:35:44Z

**Done !** (the method is now named ```middleware```, and I've removed the useless comment)

I've tried my best for the feature documentation, but unfortunately my english is fairly poor... :-/

At this point, the Middlewares have access to the Request, but if I'm not wrong it can't be used for redirecting, for instance. Should I handle ```Response``` returns from middlewares, as in ```before``` filter ?

---------------------------------------------------------------------------

by GromNaN at 2012-03-02T12:39:00Z

I guess you can throw an exception (access denied for example) and manage this exception with the error handlers.

---------------------------------------------------------------------------

by DrBenton at 2012-03-02T12:44:38Z

Yes, in the sample usage of the documentation I've used exactly that ```Exception``` process.
But I don't know if this is ok for everyone ?

---------------------------------------------------------------------------

by igorw at 2012-03-02T15:08:29Z

Very nice feature. But it should really be possible to return a response from the handler, like you can do with before() filters.

---------------------------------------------------------------------------

by DrBenton at 2012-03-02T21:15:41Z

All right, I think I know how to handle Responses, thanks to the code of the before() filter ; I'll push soon a new commit with the last code syle fixes and this feature !

---------------------------------------------------------------------------

by DrBenton at 2012-03-03T11:34:44Z

Ok, I've fixed the code style problems, and handled the Responses from middlewares.
What do you think of it ?

---------------------------------------------------------------------------

by DrBenton at 2012-03-05T13:11:56Z

So... Are we OK for this feature, gentlemen ? :-)
Do you want me to push a squashed commit ?

---------------------------------------------------------------------------

by igorw at 2012-03-05T22:24:53Z

Apart from the minor comments I made I am quite happy with the implementation as it is. I guess when those are fixed it can be merged.

---------------------------------------------------------------------------

by DrBenton at 2012-03-05T22:58:23Z

Sorry for all these "code style" errors, I'm still not enough used to Symfony 2 coding standards - but I'll get them ! :-)

Well, all these errors should be fixed now. I also swapped  the order of execution with the Silex global ```before``` filters, as you asked it: the global filters are now triggered **before** the middlewares of the matched route.

---------------------------------------------------------------------------

by igorw at 2012-03-05T23:10:35Z

I'm not so sure about checking for `instanceof Response`. Maybe we should enforce a middleware to return either null or a Response to prevent unexpected behaviour.

---------------------------------------------------------------------------

by DrBenton at 2012-03-06T14:37:42Z

Well, I just copied the behaviour of the ```before``` filter :
```
if ($ret instanceof Response) {
    $event->setResponse($ret);
}```
(in Application.php, line 237)

Since route middlewares act as this ```before``` filter (in a "specific for each route" way), I thought we could copy the code of this ```before``` functionnality ?

---------------------------------------------------------------------------

by DrBenton at 2012-03-08T14:34:13Z

So ? Do you want me to change the ```instanceof Response```, or can we keep the current "copy of before filter" implementation ?

---------------------------------------------------------------------------

by igorw at 2012-03-08T15:57:00Z

IMO we should add something like this:

    if ($response instanceof Response) {
        ...
    } elseif (null !== $response) {
        throw new \RuntimeException('Before filter (or middleware for route foo) returned an invalid response value. Must return null or an instance of Response.');
    }

---------------------------------------------------------------------------

by DrBenton at 2012-03-08T16:26:54Z

Hum, I see. Sounds good to me, I'll push it soon with the documentation fixes :-)

---------------------------------------------------------------------------

by igorw at 2012-03-08T16:42:48Z

And tests please ;-)

---------------------------------------------------------------------------

by DrBenton at 2012-03-08T22:15:11Z

Here it is, sir ! :-)

---------------------------------------------------------------------------

by DrBenton at 2012-03-10T19:45:59Z

Shall I push in a new PR these 13 commits in a single squashed one ?

---------------------------------------------------------------------------

by igorw at 2012-03-10T19:51:53Z

👍

---------------------------------------------------------------------------

by stof at 2012-03-10T19:52:43Z

No need to create a new PR for this. Simply squash your commits and force the push to the branch

---------------------------------------------------------------------------

by DrBenton at 2012-03-10T21:59:12Z

All right, here is the squashed commit ! :-)

Commits
-------

8d30526d [composer] remove version and type
9d88f817 [composer] add branch-alias

Discussion
----------

[composer] add branch-alias

This means we get a 1.0-dev version that can be installed as `1.0.*` or `>=1.0-dev`

---------------------------------------------------------------------------

by fabpot at 2012-02-27T13:57:11Z

I think we also need to remove the version entry (ping @Seldaek).

---------------------------------------------------------------------------

by igorw at 2012-02-27T14:01:17Z

Good point, done.

Commits
-------

9516a002 Revert composer.json to use 2.1 versions again instead of `dev-master`.

Discussion
----------

Revert composer.json to use 2.1 versions again instead of `dev-master`

Commits
-------

8e7c0a4b Applied fixes requested by igorw.
d7df7eb8 Added Translating Validation Messages to cookbook.

Discussion
----------

Added Translating Validation Messages to Cookbook.

Commits
-------

9a624c47 Add a link to third party providers page in the doc

Discussion
----------

Add a link to third party providers page in the doc

---------------------------------------------------------------------------

by igorw at 2012-02-25T17:44:53Z

👍

Commits
-------

ee1793c7 [session-update] bring vendors up to date
77429a59 [session-update] Update all docs references to 2.1
dbc55d25 [session-update] add note to CHANGELOG
143ca105 [session-update] update to 2.1 session refactoring

Discussion
----------

Session update

Commits
-------

c78353e2 Give useful error message when accessing request outside of request scope

Discussion
----------

Give useful error message when accessing request outside of request scope

Commits
-------

2a17fc96 [composer] update composer dependencies

Discussion
----------

[composer] update composer dependencies

---------------------------------------------------------------------------

by igorw at 2012-02-20T20:56:36Z

This is kind of urgent, since it breaks silex for everyone who is using it with composer. If you could merge this ASAP, would be great.

---------------------------------------------------------------------------

by stof at 2012-02-20T22:21:26Z

@igorw see my suggestion on doctrine/DoctrineBundle#24 I think it is what composer will force us to do instead of using master as main branch to be able to keep a sane versionning

Commits
-------

81dc9911 [cookbook] Fix typo {Ceated => Created}

Discussion
----------

[cookbook] Fix typo {Ceated => Created}