Rewrite/refactor parts of routing, at least gen()

[david]

Especially gen() is awfully broken because it had to evolve and change direction numerous times (e.g. #434, #444) - we can rewrite it to make it more maintainable, faster, simpler, etc etc etc etc. Right now, it's more or less a house of cards prone to fall over if we change the most simple things. Maybe we can optimize other aspects while we're at it.

[david]

#630 describes another issue.

Basically, we should refactor the routing so that any necessary encoding/escaping of values is done _afterwards_ by the concrete implementation. In case of the WebRouting?, that'd be rawurlencode().

I had the idea that the returned $path from parent::gen() (which we should rename since the signatures and return values differ) is an object with __toString(), and each of the $usedParams is a reference to the respective chunk in that object. Then, a rawurlencode call can be made afterwards. Also, a chunk could be some AgaviLiteralRoutingFragment or such object, which do not get encoded.

That would even, more or less, maintain BC.

Oh and... the array('AgaviToolkit', 'isNotArray') callback to array_filter is horribly slow; is_numeric does the job ~5 times faster, but is not the same thing of course.

The goal for this routing refactoring should also be that we can allow passing arrays and objects as routing gen args.

[david]

More stuff to be done:

• factor out code from gen() methods
• allow $ro->gen('+rss') which would generate the current routes plus rss
• move Routing::gen() to Routing::doGen() so subclasses can overwrite that one independently

[felix@…]

a partial patch against 0.11 that enables generation of '+rss' style routes

[felix@…]

I noted that the solution for emulation '+rss' style routes that was quoted on irc has some major flaws: It does not honor get params that are not handled by routes, making you loose the query string. That's a bad thing (tm) and so I decided to develop a patch against 0.11.

Have fun