SyncEngine (CloudKit) fetching unwanted zones

[mahee96]

Hi Team, Thanks for the great library which also provides CloudKit syncing.

I have 1 question though, because I have a case where I wanted to keep the data from individual device types to be separately synced though under same container since all are under same app bundle.

    _ = try SyncEngine(
                    for: dbPool,
                    tables: TestTable.self,
                    defaultZone: zone,                // this will be "iPhone" for iPhone and "Mac" for Mac device 
                    startImmediately: true
                )

in this case, when I tried to first freshly load up my app on iPhone and SyncEngine synced data into CK db, and I tried to install same app on Mac (designed for iPad mode) on Mac, the data from iPhone zone was also merged into Mac app's db

is this intentional, or is there something I am missing here. is Zone isolation for fetch+write not possible?

[mbrandonw]

Hi @mahee96, we did not really have your use case in mind when building SQLiteData. The defaultZone simply represents the zone we put records into when they are created. It does not prevent records from being sync'd between devices, though that is a possibility we could consider.

Can you explain more why you are separating records by device type?

[mahee96]

my use case was limited and was mostly exploratory for device specific sync. ex: I wanted to have cloudKit isolation during syncing based on device type purely for isolating environments.

lets say an app which runs on iOS can go to zone = iPhone and same app when ran on macOS zone = macOS, I can pass a dynamically generated value for "defaultZone" as u have pointed out correctly.

which is how it works currently matching the snippet I showed, so creation/upload works as expected (data goes into respective zones).

Now regarding download/fetch part, instead of fetching the active/default zone passed to engine, I got data from other zones or all zones too.

Use case Flow:

1. App is installed on iOS
2. SyncEngine uses zone = iPhone, sync starts immediately.
3. Zone gets created on CK database for zone = "iPhone" on the provided container.
4. Data/records gets uploaded.
5. now I also install the same app [destination = Mac("Designed for iPad")] simultaneously without any code modifications onto a MacOS.
6. App launches and SyncEngine auto-fetches data because startImmediately = true, but uses zone = "MacDesignedForIpad"...
7. Now I see that, data uploaded into "iPhone" zone by iPhone install is now also present on the app UI running on Mac, despite no data was generated by Mac app installation yet. (meaning iCloud sync worked).

But it didn't adhere to strict zoning, that is, it pulled unwanted zone's data.

I checked cloudKit db and it showed properly 2 zone created for same container as expected, but data was almost same on both zones though I was expecting an empty zone, coz MacOS app installation didn't do any useful work yet (it didn't generate its own data)

8. Now I relaunched my iOS app, and it had Mac zone contents...

I felt it did a cross zone sync and seemed to be a bug coz, it felt it was polluting data now, by merging both zones without consent.

I am new to cloudKit too in general so if my understanding of this zone fetch was wrong and it was working as expected, then we don't have any issue. That's why I asked it here.

I could have done this in a different way too, by creating multiple containers completely isolated for device type.

but since zoning was already available, I tried to use and keep it in same container and not pollute my CK db entries.

this way I could keep cross-device-sync as on-demand facility or give user a chance to keep iOS and macOS data completely separate without change in DB Schema to account for device variant types.

[stephencelis]

@mahee96 Zones provide a way of bucketing records, but CloudKit sync engines attempt to sync everything in a container regardless of zone, and I don't think we want to change this behavior.

Instead, you should be able to quarantine things to their own CloudKit containers. You could have a "Mac" container and an "iPhone" container and introduce code that chooses the appropriate container for the device.

[mahee96]

Thanks for the explanation, yeah, agree that the engine should mimic CK API behavior, thanks for quickly getting back on this, much appreciated!