Trouble with Routing.

[James Curran]

In my Controller, I have two actions (of importance here):
TroupeController.Display(int id)
TroupeController.Display(string name)

and I have (for now) just one routing rule:

PatternRoute pt = new PatternRoute("troupebyname", "/Troupe/
<name>.rails")
.DefaultForController().Is<TroupeController>().DefaultForAction().Is("Display")
.Restrict("name").ValidRegex("[a-zA-Z1-9]+");
RoutingModuleEx.Engine.Add(pt);

Now, if I were to try to use this routing:
http://localhost:20322/Troupe/papermill.rails

One would think (or at least, hope) that this would be routed to the
second Display action. It is however, being routed to the first
(where, naturally, it fails).

What am I doing wrong, and how to I fix it?

[James Curran]

I guess I should add that I will eventually be adding:

pt = new PatternRoute("troupebyid", "/Troupe/<id>.rails")
.DefaultForController().Is<TroupeController>()
.DefaultForAction().Is("Display")
.Restrict("id").ValidInteger;
RoutingModuleEx.Engine.Add(pt);

to route to the first Action list above (as well as a bunch of other
routes), but I just trying to get the one working first.

[James Curran]

With further experimentation, it gets odder. Using the standard URL
form:

http://localhost:20322/Troupe/Display.rails?name=12miles

it gets routed to the correct action, except now the parameter "name"
has the value "Display".

And
http://localhost:20322/Troupe/papermill.rails?name=xxxx

routes "correctly" (i.e. it calls
TroupeController.Display("papermill"); )

[hammett]

Trunk version? Really?!

--
Cheers,
hammett
http://hammett.castleproject.org/

[James Curran]

On Sep 23, 4:20 pm, hammett <hamm...@gmail.com> wrote:
> Trunk version? Really?!
>

Probably not the absolute latest, but to should be less than a month
old. (I'll update when I get home)

[hammett]

Maybe it's time for another route overhaul.

--
Cheers,
hammett
http://hammett.castleproject.org/

[Ayende Rahien]

In that case, what about just using system.web.routes?

[hammett]

I'd consider supporting it. Not sure I'd like depending on it...

[Ken Egozi]

making the API similar might be good for PR though

--
Ken Egozi.
http://www.kenegozi.com/blog
http://www.musicglue.com
http://www.castleproject.org
http://www.gotfriends.co.il

[hammett]

And would also kill our identity.

[Ayende Rahien]

Is it the routing API that makes MR identity?

[hammett]

The whole thing, the whole belief system. I wouldnt consider asp.net
mvc's routing perfect and might consider some of their decisions
questionable. Why would I copy their api then?

On Tue, Sep 23, 2008 at 3:02 PM, Ayende Rahien <aye...@ayende.com> wrote:
> Is it the routing API that makes MR identity?

[Ayende Rahien]

I haven't actually used routing for anything serious, so I don't really consider this to be an interesting part or MR.
As such, I don't have an issue with using this if this make our life easier. If this is a problem, then no, we need to use something that works better.

[hammett]

I guess donjon was the main source of requirements of the current form
of MR routing.
I admit it's not enough, though, we need more use cases.

[Alex Henderson]

My biggest concern would be portability - Monorail on mono is actually a really good story and was one of the key selling points for the work I'm doing for a client at the moment (they favor a Linux/mono/MySql based platform for their SaaS branch of the product) - locking in to the MS 3.5 SP1 routing requires either Mono to implement routing themselves, or to maintain a seperate implementaiton of routing for mono in the Castle source/my product (unless the source is available for System.Web.Routing ??) - neither seems a very attractice situation - compared to just having a best of breed routing engine, which could perhaps include an SystemWebRoutingPatternAdaptor to make it easy for developers to use System.Web.Routing patterns in Monorail?

Cheers,

 - Alex

[hammett]

Very good point, I dont think they are planning to release the routing
code as they did for the mvc.

[Victor Kornov]

Won't it be available on source servers like the rest of FW? But that will be next to useless in such situation considering license problems.

I guess that could help with internal projects though :)

[Alex Henderson]

If I was going to bother doing that I might as well just use the System.Web.Routing assembly which I suspect would probably just work as a drop-in replacement.

[Ken Egozi]

.net 2.0 is a major sell point currently.

And I did not mean using the ms-routing as is.
and also not limiting us.
but rather to try and use the same basic syntax for routes.

The idea is that people with knowledge in MR routing will be able to approach MS-route guys, and help them out more easily - selling MR as a smarter option on the way.

But like ayende, I don't consider it an interesting enough feature to be passionate about.

[James Curran]

To update on the original problem. I was running the truck from early
July (I download & build regularly, but forget to copy it to my
website bin directory).
But even with the latest, the same problems exist. So, I started
single-stepping through the code, to see where things go awry.

First problem: I had, 'new PatternRoute("troupebyname", "/Troupe/
<name>.rails") ' but it wants to be 'new PatternRoute("troupebyname",
"/Troupe/
[name].rails")'. That's probably a bug in the code, as in other
places it accepts either character, but at one spot, it only accepts
"[".

But, that didn't solve the problem, so I continued further. What I
found as in /trunk/Components/Binder/Castle.Components.Binder/
DefaultConverter.cs at about line 33, we have:
else if (inputType == desiredType)
{
exactMatch = true;
}

when what we actually want is:
else if (inputType == desiredType)
{
exactMatch = true;
return true;
}

Which bring up the question... Why isn't that bring the whole system
down?

(I created a patch for that, but since it's just a one-liner someone
may want to just fix it directly)

[hammett]

After you patched the tests still pass?

--
Cheers,
hammett
http://hammett.castleproject.org/

[James Curran]

Normally, I build with testing off (figuring if it's checked into the
truck, they pass) (also when I started doing this, I didn't have
nUnit on that machine)

Tried building with tests turned on. It failed.
Removed my change and build again. Still failed. Apparently still
some problem with my testing setup. will investigate.

[James Curran]

On Wed, Sep 24, 2008 at 12:23 PM, hammett <ham...@gmail.com> wrote:
>
> After you patched the tests still pass?

Tests do still pass... but that's mainly because that method is never
called by any of the unit tests. I've written a set to test it, but
it's also being called in SmartDispatchController, and since the
method is never called at all in the unit tests, that means there's a
least one path through SDC that's no being tested either. I'll try to
work out a test for that also.

--
Truth,
James