Feature/udp #111

jrconlin · 2015-07-10T21:33:23Z

Fixes #106.

Add UDP Wake support for SimplePush using router
** Adds timeout to boot connections that specify UDP wake.
Fix some flake8 issues
convert tests that use ....assert_called() to use eq_(....called, True) to fix travis incompatibilities
Fix missing labels for some args.
convert some worrisome " is " tests to " == "
bumped pyasn to 0.1.8 to fix install issue.

* Added label to APNS args * TODO: Tests * Added first go on idle timeouts for websocket connections * TODO: is there a path for the UDP wake server or just go to that host/port?

* start testing

Rewrote UDP as a wake function for SimplePush. There's potentially a bit of wake abstraction that could be made in the future, but that's best done when there's a future requirement.

* Fixing some formatting * added router_conf for simplepush (since it now has stuff, sorta)

* Added edge case for check_idle * fix to correct timeout to be sec not millisec * fix to some improper is usage * restore some code from test tweaks.

* convert assert_called to eq_(...called, True) * removed websocket router init test ** Because we don't do routing registrations over websockets. ** Why, yes, I had forgotten that and spent an hour trying to wire it back in.

ghost · 2015-07-13T18:51:13Z

autopush/db.py

@@ -247,7 +247,7 @@ def register_user(self, data):
                for key, value in result["Attributes"].items():
                    try:
                        r[key] = self.table._dynamizer.decode(value)
-                    except AttributeError:
+                    except (TypeError, AttributeError):


What does the TypeError catch?

If value is not proper javascript, it throws a TypeError. If table or _dynamizer aren't initialized, they throw Attribute error. In either case, just take the value as specified and don't be clever.

* change idle_timeout to udp_timeout * add udp_server as arg, pass UDP wake args as urlencoded dict

bbangert · 2015-07-17T23:13:21Z

autopush/logging.py

@@ -76,7 +76,7 @@ def stdout(message):
    else:
        sys.stdout.write("       %s\n" % message['message'])
    return
-    """
+    # """


This seems unintentional?

Actually, sorta intentional. It lets me activate the "human readable logging" by adding a # to the start of line 73 instead of to both lines. The triple quote stops on that line regardless of the starting hash.

bbangert · 2015-07-17T23:39:01Z

I'd really like to see the wakeup split out of the router/websocket into its own small object with a few methods that implements a Wake interface (and hooked onto the completion of routing). I won't be surprised if in the future we may end up doing something similar to what Google does, where we actually have all active mobile devices connect directly to us, but when 'gone idle', they drop, and we wake them with GCM/APNS/etc. In which case we can easily grow several wakeup methods, and having each one be a custom one-off in this style will impair us.

jrconlin · 2015-07-17T23:45:56Z

Fair. I considered this a "one off" mostly because that's all we at at the moment, and the alternate bridge mechanisms handled their own idle periods. Having a wake interface makes sense. Do we expect there to be other side event interfaces (Has data/ Priority/ etc.)?

bbangert · 2015-07-17T23:48:59Z

Well, the reason I could see copying Google's approach, is that we might find GCM/APNS lacking in some way for optimal message delivery/latency, so FF on mobile when its in-use would actually hold open a connection directly to us, but drop it as soon as the app is backgrounded/killed.

* Fix typo in default collapse key for gcm * make wake references more generic

Looks like i need to figure out what's goofy with my vim config

jrconlin · 2015-07-23T20:45:56Z

Note: this may require refactoring if PR #115

Conflicts: requirements.txt

…to feature/udp Conflicts: autopush/tests/test_endpoint.py autopush/tests/test_router.py autopush/tests/test_websocket.py

bbangert · 2015-09-10T22:13:05Z

autopush/router/simple.py

@@ -70,6 +78,12 @@ def route_notification(self, notification, uaid_data):
        # Determine if they're connected at the moment
        node_id = uaid_data.get("node_id")
        uaid = uaid_data["uaid"]
+        self.udp = None
+        try:
+            self.udp = json.loads(uaid_data["udp"])


The common path for 99% of clients is going to be no UDP, exceptions are expensive in Python so we're essentially slow-pathing the common case by adding an exception to it. A 'if udp in uaid_data' test would be fast.

Also, uaid_data is a dict-like item out of DynamoDB. DynamoDB items can store structured nested dicts, so in theory we don't need to JSON encode it on the way in/out like this.

* fixed r's

…ature/udp Conflicts: autopush/websocket.py

bbangert · 2015-09-14T20:33:47Z

CHANGELOG.rst

@@ -138,6 +138,9 @@ Features
 * Add Sphinx docs with ReadTheDocs publishing. Issue #98.
  This change also includes a slight Metrics refactoring with a IMetrics
  interface, and renames MetricSink -> SinkMetrics for naming consistency.
+* Add UDP Wake support. Issue #106
+  Some devices which use SimplePush routing offer a feature to wake on
+  a carrier provided UDP ping. See issue #106 for details.


Needs to be moved to a 1.7.0 section now. Sorry bout that.

Heh, tempted to move it to ∞.0.2 at this point.

…ature/udp

bbangert · 2015-09-14T22:33:04Z

r+ pending deferToThread change.

bbangert · 2015-09-14T22:35:13Z

autopush/router/simple.py

+            if self.udp is not None and "server" in self.conf:
+                # Attempt to send off the UDP wake request.
+                try:
+                    yield requests.post(


Er, I dunno why this works in the tests. yield can only be done on results that are deferred, requests.post is a blocking call that doesn't return a deferred. It needs a deferToThread in front.

codecov-io · 2015-09-14T22:40:38Z

Current coverage is `99.95%`

Merging #111 into master will not affect coverage as of a51ef81

@@            master    #111   diff @@
======================================
  Files           27      27       
  Stmts         4934    4992    +58
  Branches         0       0       
  Methods          0       0       
======================================
+ Hit           4932    4990    +58
  Partial          0       0       
  Missed           2       2

Review entire Coverage Diff as of a51ef81

Powered by Codecov. Updated on successful CI builds.

bbangert · 2015-09-14T23:05:01Z

r+

Feature/udp

jrconlin and others added 7 commits July 6, 2015 16:40

initial UDP router

edd60eb

* Added label to APNS args * TODO: Tests * Added first go on idle timeouts for websocket connections * TODO: is there a path for the UDP wake server or just go to that host/port?

WIP: Added udp router

e3190e2

* start testing

WIP: First round tests

5a28b08

UDP Rewrite - not the router

3dbc969

Rewrote UDP as a wake function for SimplePush. There's potentially a bit of wake abstraction that could be made in the future, but that's best done when there's a future requirement.

WIP Adding tests

6f3da92

* Fixing some formatting * added router_conf for simplepush (since it now has stuff, sorta)

100% coverage

fb0289d

* Added edge case for check_idle * fix to correct timeout to be sec not millisec * fix to some improper is usage * restore some code from test tweaks.

Fix test

d8bc1a2

* convert assert_called to eq_(...called, True) * removed websocket router init test ** Because we don't do routing registrations over websockets. ** Why, yes, I had forgotten that and spent an hour trying to wire it back in.

ghost reviewed Jul 13, 2015
View reviewed changes

r's

0532c99

* change idle_timeout to udp_timeout * add udp_server as arg, pass UDP wake args as urlencoded dict

bbangert reviewed Jul 17, 2015
View reviewed changes

jrconlin added 2 commits July 23, 2015 11:21

Move UDP to interface

13f8236

* Fix typo in default collapse key for gcm * make wake references more generic

Flake8 fixes

4d3f829

Looks like i need to figure out what's goofy with my vim config

jrconlin added 2 commits July 24, 2015 11:16

Merge branch 'master' into feature/udp

684d25e

Conflicts: requirements.txt

Merge branch 'feature/udp' of github.com:mozilla-services/autopush in…

39c482d

…to feature/udp Conflicts: autopush/tests/test_endpoint.py autopush/tests/test_router.py autopush/tests/test_websocket.py

jrconlin self-assigned this Jul 27, 2015

jrconlin added the enhancement label Jul 27, 2015

bbangert reviewed Sep 10, 2015
View reviewed changes

jrconlin added 4 commits September 11, 2015 14:02

Removed waker, using TimeoutMixin

a57bba7

* fixed r's

Merge branch 'master' of github.com:mozilla-services/autopush into fe…

6b703e6

…ature/udp Conflicts: autopush/websocket.py

remove more waker detrius

9beea44

flake8 fixes

612d5e2

bbangert modified the milestones: 1.7.0, 1.6.0 Sep 14, 2015

bbangert reviewed Sep 14, 2015
View reviewed changes

jrconlin added 2 commits September 14, 2015 15:10

Merge branch 'master' of github.com:mozilla-services/autopush into fe…

4fc09ea

…ature/udp

r's and made requests.post yield

3ff89ba

bbangert reviewed Sep 14, 2015
View reviewed changes

deferToThread for requests.post

1aa9c52

bbangert added a commit that referenced this pull request Sep 14, 2015

Merge pull request #111 from mozilla-services/feature/udp

1e406c0

Feature/udp

bbangert merged commit 1e406c0 into master Sep 14, 2015

bbangert deleted the feature/udp branch September 14, 2015 23:23

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feature/udp #111

Feature/udp #111

jrconlin commented Jul 10, 2015

ghost Jul 13, 2015

jrconlin Jul 13, 2015

bbangert Jul 17, 2015

jrconlin Jul 17, 2015

bbangert commented Jul 17, 2015

jrconlin commented Jul 17, 2015

bbangert commented Jul 17, 2015

jrconlin commented Jul 23, 2015

bbangert Sep 10, 2015

bbangert Sep 10, 2015

bbangert Sep 14, 2015

jrconlin Sep 14, 2015

bbangert commented Sep 14, 2015

bbangert Sep 14, 2015

codecov-io commented Sep 14, 2015

bbangert commented Sep 14, 2015

Feature/udp #111

Feature/udp #111

Conversation

jrconlin commented Jul 10, 2015

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bbangert commented Jul 17, 2015

jrconlin commented Jul 17, 2015

bbangert commented Jul 17, 2015

jrconlin commented Jul 23, 2015

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bbangert commented Sep 14, 2015

Choose a reason for hiding this comment

codecov-io commented Sep 14, 2015

Current coverage is 99.95%

bbangert commented Sep 14, 2015

Current coverage is `99.95%`