Method for clearing handler cache.

[gcirne]

This is needed when adding a model dynamically after the rest handler has been cached. In this situation, the model would not be reflected in the handler and would return 404.

[slnode]

Can one of the admins verify this patch? To accept patch and trigger a build add comment ".ok\W+to\W+test."

[bajtos]

@slnode ok to test

[slnode]

Test FAILed. To trigger a build add comment - ".test\W+please"
Refer to this link for build results: http://ci.strongloop.com/job/loopback/937/

[gcirne]

Hi @bajtos, apparently the jenkins build failed. But the Travis build is passing. Anything I can do to help fix it?

[bajtos]

Could you please rebase your changes on top of the current master? It looks like your patch is based on loopback@1.6.1.

Hint:

$ git fetch upstream # upstream = strongloop/loopback
$ git rebase upstream/master
$ git push -f

[gcirne]

I believe it is already based on current master.

$ git log
commit 1424e83c586e5f96f100f9cc4f6dd1fca09b62cd
Author: Guilherme Cirne <gcirne@gmail.com>
Date:   Mon Feb 17 14:27:35 2014 -0300

    Method for clearing handler cache.

commit e56574fd13265e2f725a7376d5ff71ebaea1bac0
Merge: 1dc4a3f 46b579d
Author: Raymond Feng <enjoyjava@gmail.com>
Date:   Fri Feb 14 11:36:17 2014 -0800

    Merge pull request #181 from strongloop/feature/user-token-rel

    Make sure User/AccessToken relations are set up by default

[gcirne]

$ git remote -v
origin  git@github.com:globocom/loopback.git (fetch)
origin  git@github.com:globocom/loopback.git (push)
upstream    https://github.com/strongloop/loopback.git (fetch)
upstream    https://github.com/strongloop/loopback.git (push)
$ git fetch upstream
$ git checkout clear-handler-cache
$ git rebase upstream/master
Current branch clear-handler-cache is up to date.

[bajtos]

Then it's a bug of our Jenkins setup :(

@rmg could you please investigate? Looks like this is the same issue we were discussing over the e-mail recently. Is there a way how to reset the Jenkins state to get this pull request built correctly?

[bajtos]

Back to the patch itself. First of all, thank you @gcirne for submitting it!

AFAIK app.handler is intentionally kept undocumented, LB applications should not need to worry about the implementation details of handler caching.

If I recall correctly, we were discussing this problem with @ritch and/or @raymondfeng and the plan was to add an event when a new model is registered. The handler cache can listen on this event and invalidate REST handler when a new model is added.

@gcirne Could you please describe your use case? How does it happen that you are adding/modifying models after the first REST request was handled?

[gcirne]

Sure. We basically want to add a new model dynamically by POSTing a JSON
Schema (http://json-schema.org/) which describes the new model.

We have a JsonSchema model with an afterSave hook registered. When we POST
a JSON Schema to /api/json-schemas the afterSave hook is executed and
registers the new model. At this point, we need to clear the REST handler
from the cache so that we can access the new model over the REST api.

The event solution you describe sounds good to me and fits perfectly with
our use case. Do you want me to modify this PR? I'm new to LB internals but
from what I've seen, this isn't a really big change.

Guilherme Machado Cirne
gcirne@gmail.com

On Mon, Feb 17, 2014 at 4:12 PM, Miroslav Bajtoš
notifications@github.comwrote:

Back to the patch itself. First of all, thank you @gcirnehttps://github.com/gcirnefor submitting it!

AFAIK app.handler is intentionally kept undocumented, LB applications
should not need to worry about the implementation details of handler
caching.

If I recall correctly, we were discussing this problem with @ritchhttps://github.com/ritchand/or
@raymondfeng https://github.com/raymondfeng and the plan was to add an
event when a new model is registered. The handler cache can listen on this
event and invalidate REST handler when a new model is added.

@gcirne https://github.com/gcirne Could you please describe your use
case? How does it happen that you are adding/modifying models after the
first REST request was handled?

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/187#issuecomment-35313789
.

[rmg]

test please

[rmg]

Sorry, @gcirne, Jenkins seems to be quite confused at the moment. Easiest way to trigger a rebuild and have it actually build the right thing is to push another commit. :-(

[slnode]

Test FAILed. To trigger a build add comment - ".test\W+please"
Refer to this link for build results: http://ci.strongloop.com/job/loopback/942/

[gcirne]

Hmm, failed again (but Travis is fine). :(

[slnode]

Test FAILed. To trigger a build add comment - ".test\W+please"
Refer to this link for build results: http://ci.strongloop.com/job/loopback/943/

[rmg]

Jenkins is still building the wrong thing.. :-/

Whoever commits the merge will just have to make sure they test locally when they do it, I guess.

[bajtos]

The event solution you describe sounds good to me and fits perfectly with
our use case. Do you want me to modify this PR? I'm new to LB internals but
from what I've seen, this isn't a really big change.

That would be awesome!

[gcirne]

@bajtos ok, I'll try to get it done today.

[ritch]

Sorry, but as @bajtos said this is exposing internals shouldn't be exposed. We might drastically refactor these in 2.0.

However clearing the handlers on an attached event sounds good. In that PR you could include this method if it is private.

[gcirne]

Hey @ritch, I'm not really sure I agree that the current implementation is exposing internals. The clearHandlerCache function exists to abstract away the internals.

I was thinking of making RestAdapter listen to the attached event and then call app.clearHandlerCache('rest'). But if this function is "private", how would I go about that?