Office Hours — Today, July 31

← Thursday, July 26

Mark M.
has entered the room
Mark M.
turned on guest access
Jul 31
4:00 PM
Frank S.
has entered the room
Mark M.
howdy, Frank!
how can I help you today?
Frank S.
hey mark
1 sec. I have to change rooms real quick
I'm trying to animate a view pre 11 skd. And I'm having a real problem with combining TransitionAnimations with offsetLeftandright
Terence M.
has entered the room
Mark M.
howdy, Terence!
Frank: I have not used offsetLeftAndRight
Terence M.
Howdy do!
Frank S.
aww pickels
Mark M.
what is your overall objective? what are you using offsetLeftAndRight() for?
4:05 PM
Mark M.
and what's a skd?
Frank S.
View paste (80 more lines) 
    private void closeMenu(float startingX) {

        Log.d(TAG, "close startingX " + startingX);

        TranslateAnimation animation = new TranslateAnimation(startingX - mOffset, -1 * mOffset, 0,
                0);
        animation.setDuration(mAnimationDuration);

        animation.setAnimationListener(new AnimationListener() {

            @Override
            public void onAnimationStart(Animation animation) {
                mIsOpen = false;
                mContent.setOnTouchListener(null);
            }
...
**sdk
Mark M.
setFillAfter() is almost always the wrong answer for whatever you're trying to do
mostly because it does not change the actual touch positions
it just changes where pixels get drawn
usually, you need to do something in onAnimationEnd() that actually makes your animation effect permanent, by adjusting the layout
Frank S.
ok. So that might actually work because I'm losing track of my pixels somewhere
Mark M.
for example, if you are sliding something off the screen, make its visibility be GONE in onAnimationEnd()
Frank S.
Yeah I tried to reset the offset, but it made the view jump to weird positions
Mark M.
my desperately-needing-a-refresh Animation chapter covers that
is this some sort of sliding menu?
Frank S.
Trying to make a nice lib for the G+ and youtube sliding menu
Mark M.
if you haven't done so already, Cyril Mottier has a multiple-post blog series on how he implemented one of those
let me take a question from Terence, and I'll be back with you shortly
Terence: do you have a question?
4:10 PM
Frank S.
yep. I got it to work perfectly with the new sdk 11 tools, but that doesn't work pre 3.0
Mark M.
OK
if either of you have a question, chime in
Terence M.
I have a couple of questions referring to Tutorial 17, the one I was working on yesterday when I sent the mail
Mark M.
ah, that mail
OK, I'm with you
go ahead
Terence M.
yeah I'm a real trouble maker ;)
Step 4 mentions needing to modify things to use sendWakefulWork() rather than start service, however for the DownloadCompleteReceiver the code listed looks the same as what I already have pre-modification
I checked the version current in the Repo and it's the same
I modified the line to be: WakefulIntentService.sendWakefulWork(context, DownloadInstallService.class);
Mark M.
which is correct
Terence M.
everything seems to work, and the download completes ok and installs but the notification about the download remains in the bar even after you click it
Mark M.
I am not sure how I messed up my copy
Terence M.
I've only been working in the emulator since it involves changing the clock so maybe it's just that, but I thought the two might be related
Mark M.
in your InstallReceiver, make sure you have setAutoCancel(true)
4:15 PM
Mark M.
in the chain of calls to the NotificationCompat.Builder
Tom
has entered the room
Mark M.
howdy, Tom!
Tom: do you have a question?
Tom
hi Mark
Yes, I'm having problems with bind/unbind and wondering whether I should use startService instead.
The first problem. My GUI calls bind and waits for onServiceConnected() to be called before proceeding (e.g. cause it doesn't even have access to preferences until it has its handle to the service). In my service's onBind I send a message to defer doing my initialization so I can return very quickly, but even that is not enough to get my on onServiceConnected called - I think I'll have to send the message with a delay.
Terence M.
I think that may be in the next tutorial, I don't see any mention of that here in #17. Is that something that will be added in the notification tutorial coming up or have I messed up here?
Mark M.
IMHO, your onBind() should be doing nothing but returning the Binder
Tom
The second problem. When the user presses home button, my activites onStop gets called and I unbind (I think I have no choice - that onStop might be the last call my GUI gets). But now my service shuts down (including killing all active network connections) which is rather premature given that the user just pressed home (not back). So, I'm going to add something to my service to stick around another ~20 seconds after the unbind, but I feel like I'm fighting against the bind/unbind mechanism again.
Mark M.
Terence: no, you're right -- I was looking at the T-18 code
4:20 PM
Mark M.
Tom: what specifically are you trying to gain from binding that you could not accomplish via the command pattern and startService()?
4:20 PM
Tom
To tell you the truth, I can't remember why I chose to use bind/unbind rather then start.
Mark M.
regardless, rather than calling unbindService() directly from onStop(), you could do the delay logic there
use postDelayed() to post a Runnable to be invoked after 20 seconds
if onStart() is called, cancel the Runnable
if the Runnable gets to execute, have it do the unbindService()
Tom
But if onStop has been called, and I delay before calling unbind, then my activity could well get killed and I never call unbind.
Mark M.
if by "killed" you mean "destroyed", you could also unbindService() and cancel the Runnable in onDestroy()
if by "killed" you mean "the process is terminated", your service is gone then too
Tom
Yes, I meant destroyed.
I thought there was no guarantee that onDestroy would be called.
Mark M.
there isn't, but if it is not called, your process was terminated
in which case, the issue is moot
Tom
Ah, good point.
Mark M.
moreover, your process can be terminated even if you switched this to the command pattern and were doing the 20 second delay in the service itself
4:25 PM
Tom
That resolves my issue#2 with bind/unbind, but that still leaves issue#1 and has me still wondering whether there is any disadvantage to switching to StartService instead.
Mark M.
personally, I prefer startService() to bindService()
so long as you know when to shut down the service
that being said, shouldn't your initialization be in onCreate(), rather than onBind()?
Tom
I wanted to move as much init as I could after onCreate and onBind in the hopes of allowing the service to get started and return from onBind as fast as possible.
So that the app/gui could get its handle to the service.
Mark M.
if onServiceConnected() is not being called at all, you have bigger problems, and I haven't the foggiest notion what they'd be
Tom
It is being called, but it takes longer then I want.
Mark M.
Tom: let me take some questions from the others, and I'll circle back to you
Frank: do you have another question?
Tom
Sure. Thanks.
Mark M.
Terence: do you have another question?
4:30 PM
Frank S.
Nope. Just trying to get everything working again without fillenabled
Terence M.
Not at the moment. Running a second time seemed to clear things up but some other stuff has gone off the rails so I'm going to have to poke more and see if I missed something
Mark M.
Tom: OK, back to you
the only reason I know of for onServiceConnected() to be delayed would be if you are tying up the main application thread
Tom
View paste 
I don't think I'm doing that: the service postpones all its initialazation.  onServiceConnected is being called, but I need some info from the service in order to populate the GUI, and the extra time required to wait for onServiceConnected to be called means that populating my GUI gets delayed.  I find myself doing some strange things, e.g. my app loads its own temporary copy of preferences (the service's copy is the 'main' copy) rather then wait for the service to be ready.
Mark M.
define "postpones all its initialization"
are you using a background thread?
Tom
Not using a background thread, but the onBind does almost nothing other then post a delayed message (which later triggers more initialization).
Mark M.
define "post a delayed message"
postDelayed() is a method on View; your Service does not have a View
Tom
Just a sec - I'm checking the code.
Mark M.
oh, it's also on Handler
4:35 PM
Mark M.
if you are using postDelayed() on Handler, that work in the Runnable is being done on the main application thread
Tom
I'm sending an empty message to a handler.
Mark M.
same thing -- handleMessage() is called on the main application thread
Tom
Yes, the work is being done on the app thread - but after the delay which allows the binding to be completed.
Mark M.
why would you even *want* the work to be done on the main application thread?
Tom
Because there are quite a few interactions between the GUI and the service and it simplifies those interactions if they are in the same thread.
Mark M.
frankly, you are scaring me with each passing entry in this chat... :-)
Tom
All I/O is done in its own threads.
Well, frankly, this whole architecture has become more complicated then it should be - certainly something needs to change.
Mark M.
regardless, if onServiceConnected() is being delayed, I believe reasonably strongly that it is because of work you are doing on the main application thread
to put it another way, I have no idea what else would delay it
Tom
Ok, I accept that diagnosis - I just haven't figured out how. But that is probably moot as I think I need to make bigger changes.
Mark M.
with regards to switching to startService(), that's certainly possible, but you will need to choose how you want to get info from the service back to an activity
in terms of standard Android stuff, I'm a fan of LocalBroadcastManager from the Android Support package
4:40 PM
Mark M.
your service can send local broadcasts from a background thread, and the activities can pick them up in the main application thread
Tom
Yes, I have started using LocalBroadcastmanager where possible.
Mark M.
Otto, a library from Jake Wharton and the guys from Square, is another alternative, though I haven't tried it yet
or, use regular broadcasts, a Messenger tied to your activity's Handler, etc.
Tom
Okay, I'll take a look at that. I've also heard about this strict mode and been thinking I need to start using that.
Mark M.
StrictMode certainly will yell at you if you stray from the straight-and-narrow with respect to work on the main application thread
Tom
I have one more - much simpler - question, but perhaps you want to check with the others before I ask it.
Mark M.
Frank, Terence: either of you have a question?
Terence M.
nay, just soaking up info at this point
Mark M.
OK, Tom, go ahead
Tom
I believe that, if the system needs to load one of my receivers, it must load (though not initialize) the entire app. So a large APK with many activites take away from the efficiency of a tiny receiver.
Or does the system cache receivers?
4:45 PM
Mark M.
um, to some extent, neither :-)
4:45 PM
Tom
Does the system open my APK and just load the receiver from it?
Mark M.
if you have no process, and a broadcast comes that matches something in your manifest, Android will fork the zygote process, then instantiate any ContentProviders of yours, then instantiate your Application, then it will create your BroadcastReceiver and call onReceive()
notably, it will not load any classes not touched by those things
hence, your activities and such should not impact matters
now, if you have tons of initialization work in onCreate() of a custom Application subclass, *that* price you will pay on each process start
which is one of the reasons I prefer static data members/singletons, as they will only get initialized when needed
Tom
Okay, I do have a custom App subclass, but it is rather light-weight.
Thanks Mark. Bye.
4:50 PM
Mark M.
if anyone has any questions, chime in
Terence M.
I can't find the reference at the moment but I think you mentioned that there's no "good" way to get a notification from the DownloadManager for only a download that you started
Mark M.
near as I can tell, you get all of them, which bugs the crap out of me
Terence M.
or something to that effect, at any rate
Mark M.
but it's been a bit since I looked at this issue
Terence M.
is there a workaround mentioned in the book about how to hook into that in some way to sort of "fake" it?
Mark M.
fake what?
Terence M.
that may be the wrong choice of words.. to start a download and be able to tell when you get a notification if it was one that you started or not
4:55 PM
Mark M.
first, you'd keep track of your own pending downloads somewhere
then, you'd use DownloadManager.Request to check the status of any pending downloads, and deal with those that are now completed
or, use the extras on the ACTION_DOWNLOAD_COMPLETE broadcast, like EXTRA_DOWNLOAD_ID, and compare that to your downloads
Terence M.
ahh ok, cool. I was afraid you had to check to see if the file you thought you were downloading was there and complete somehow or some such
I thank you for the info sir, and the clarification on the sendWakefulWork() line.
Mark M.
yeah, somehow I lost the one in the receiver, though the activity had it
repairs will go out in tomorrow's Version 4.0 of the book
5:00 PM
Mark M.
OK, that's a wrap for today's chat
next one is Thursday, 10am Eastern
Terence M.
have a good evening, sir!
Mark M.
have a pleasant day, all!
Frank S.
has left the room
Terence M.
has left the room
Tom
has left the room
Mark M.
turned off guest access

← Thursday, July 26

 

Office Hours

People in this transcript

  • Frank Sposaro
  • Mark Murphy
  • Terence Martin
  • Tom