Age | Commit message (Collapse) | Author |
|
Bug: 73007132
Test: NumberAttributesConverterTest, PhoneLookupInfoConsolidatorTest
PiperOrigin-RevId: 185545712
Change-Id: I228d8c4e1b6327e38057f73aad63bb7048704d49
|
|
Bug: 73077158
Test: CallLogEntryTextTest, GlidePhotoManagerImplTest, PhoneLookupInfoConsolidatorTest
PiperOrigin-RevId: 185017362
Change-Id: I113472482da2213d17a847054272a22249edc578
|
|
These are frequently used attributes of numbers that we can compute once at parse-time.
Also did some general cleanup of DialerPhoneNumberUtil:
-Removed unused Future version of parse()
-Remove formatToValidE164 now that the new fields are available
-Inlined normalizeNumber()
Bug: 72563861
Test: existing
PiperOrigin-RevId: 183720128
Change-Id: I702dc265360e590439c5352c493ae8a858f36812
|
|
It turns out the storing the libphonenumber representation of the number is not particularly useful because even formatting these objects cannot be done on the main thread. Rather than propagate the requirement of using PhoneNumberUtil (and background threads by extension) in the call log UI, we now just store a dialer-normalized version of the number which contains all information required by the UI in a way that allows us to avoid any background work in the UI code.
Bug: 72563861
Test: existing
PiperOrigin-RevId: 183463907
Change-Id: I4bdadaccb7a84033b3c72c54fe3833064f587ee3
|
|
CallLogPhoto.getPhotoUri() returns a URI to a drawable so it will be easier to transition into glide. Meanwhile ContactPhotoManager will just show the drawable directly.
Bug: 70989547
Test: Unit tests
PiperOrigin-RevId: 183163818
Change-Id: I4ee4ff98782e35d2be03dfe14f8bf3dfd6ded074
|
|
This is an optimization to reduce "popping" in the new call log. Since we have to do the PhoneLookup anyway to update the UI, we may as well write the result to PhoneLookupHistory so that the next time the AnnotatedCallLog is refreshed, the updated results can be retrieved from PhoneLookupHistory.
I also updated RealtimeRowProcessorTest to use FakePhoneLookup rather than the real Cp2PhoneLookup since RealtimeRowProcessor no longer uses Cp2PhoneLookup directly (it was updated to use the general-purpose PhoneLookup in a previous CL but I didn't update the test at that time).
Test: unit
PiperOrigin-RevId: 182974567
Change-Id: I813e9d69f802ca08757238290fdfcf58e78b3592
|
|
There's a problem with the existing implementation of RealtimeRowProcessor; when CP2 information is not present, data from other sources can potentially be erased.
This CL fixes the problem by fetching the latest data from all sources, instead of just CP2. This requires being able to fetch PhoneLookup info without a Call, using only a number, so I changed PhoneLookup#lookup to accept a DialerPhoneNumber rather than a Call.
(The reason that it accepted a Call was to support CNAP so we'll need a revised solution for that later.)
There is a potential concern with performance in RealtimeRowProcessor due to performing a full [Composite]PhoneLookup vs. a CP2 lookup, because the full lookup includes network requests. However, it's anticipated that the real time lookups will very rarely have changes to apply so this might be OK as-is. If not, a mitigation strategy could be improving the performance of CompositePhoneLookup#lookup by short-circutiing slower sources when faster, higher priority sources have already completed.
A follow-up CL will write the result of RealtimeRowProcessor queries to PhoneLookupHistory to further reduce how frequently real time updates need to be applied.
Bug: 72229553
Test: existing unit
PiperOrigin-RevId: 182839130
Change-Id: I8cb26827b4f4dc4702fb416234c8938179cd5ac5
|
|
This is an optimization to reduce popping in the new call log. Currently when Cp2LocalPhoneLookup determines a number to be "incomplete" (because it is an invalid number and there are too many invalid numbers in the call log to efficiently bulk update) we clear the existing data, which has been populated in PhoneLookupHistory (for example, from InCallUi). This means that we will show the number initially when displaying the call log, and then when the query completes we will "pop in" the new information.
This change makes it so that we don't clear the existing data from PhoneLookupHistory, and just add the "incomplete" bit. The result of this is that we immediately display the available information when initially displaying the call log (even though it may be out of date). When the query completes, the row will be updated with the most recent information; in most cases this is likely to be the same as the information used to initially display the row, and no update will need to be applied.
Additoinal changes to support this functionality:
-RealtimeRowProcessor is now just responsible for returning an updated row, and NewCallLogView holder will compare the result to the originally displayed row and only update the UI if there are differences.
-NewCallLogFragment now calls clears the RealtimeRowProcessor's cache and notifies data set changed during onResume. This is to account for the fact that AnnotatedCallLog no longer contains the complete set of information necessary to show the call log; there may be changes we need to show which can't be detected by the cursor loader. We now show those potential changes in onResume.
Additional notes:
-If there is real-time data that changes after onResume it won't be detected but there shouldn't be such cases; changes made to contact information from dialer are always done through contact cards which pause the fragment.
-This change has the effect that whatever information was written to PhoneLookupHistory during the previous invocation of InCallUi will always be (initially) shown. For example, if the contact name for number "123" is "Joe" when the call comes in, we'll write "Joe" to PhoneLookupHistory. If the user changes Joe's name to "Jane", the UI will pop from "Joe" to "Jane" until PhoneLookupHistory is updated (which is currently only done from InCallUi). If this turns out to be a problem it could be mitigated by writing updated results to PhoneLookupHistory from the UI.
Test: unit, manual
PiperOrigin-RevId: 182277145
Change-Id: I3d9916b7747390ff956f399fe84b26d578e5a07f
|
|
PhoneLookupInfoConsolidator is designed for the following two purposes.
(1) Different sub-messages in a PhoneLookupInfo proto can contain information for the same purpose. For example, all of cp2_local_info, cp2_remote_info, and people_api_info have the information for a contact's name. PhoneLookupInfoConsolidator defines the rules that determine which sub-message should be used to display the name in the UI. This is the same as PhoneLookupSelector.
(2) Avoid mixing info from different sub-messages when we are supposed to stick with only one sub-message. For example, if a PhoneLookupInfo proto has both cp2_local_info and cp2_remote_info but only cp2_remote_info has a photo URI, PhoneLookupInfoConsolidator should return an *empty* photo URI as cp2_local_info has higher priority and we should not use cp2_remote_info's photo URI to display the contact's photo. This is what PhoneLookupSelector is unable to do.
Bug: 71763594
Test: PhoneLookupInfoConsolidatorTest
PiperOrigin-RevId: 182236013
Change-Id: If19cdc1a9e076f3ebc8f9e2901f050b519e273f2
|
|
Otherwise changes to "incomplete" rows won't take effect until a new adapter is created.
Test: unit and manual
PiperOrigin-RevId: 181823087
Change-Id: I24e1b1b465c8d37cf794312b88b6cdd3ad394b5d
|
|
This allow extra information from PhoneLookup to be more easily added. Only PhoneLookupSelector and the proto will be affected for new attributes.
Test: Unit tests.
PiperOrigin-RevId: 181675568
Change-Id: I4e0bc1c6005b58a9b684b030b55bea6223af9ce3
|
|
PhoneLookupInfo.cp2_local_info.
To support remote CP2 contacts, there will be a new PhoneLookup ("Cp2RemotePhoneLookup") and a new field in proto PhoneLookupInfo ("cp2_remote_info").
In proto PhoneLookupInfo, cp2_local_info and cp2_remote_info will be of the same type ("Cp2Info").
Bug: 71763594
Test: Existing tests
PiperOrigin-RevId: 181405798
Change-Id: I6c43b486229d4e9ae7b55c579d9c9997a2884c80
|
|
This is for performing work inside of the call log's RecyclerView, when the view holder is bound. Most of the time, this should be a no-op but there are possible edge cases where the call log data cannot be updated efficiently through the standard batch mechanism.
One example of this is when there are too many invalid numbers in the call log; the CP2 information for invalid numbers cannot be efficiently batch updated so we fetch this information at display time. (Note that we do handle up to 5 invalid numbers in the batch update mechanism, but if there are more than that we fallback to this realtime processing.)
Test: unit, manual
PiperOrigin-RevId: 181400016
Change-Id: Iea6b380742e757b48d19f319fe46dc5fae837604
|