Create FakeHealthService for unit testing

[jmichas]

It would be cool to create a fake inmemory storage version of the IHealthService for testing.
Android has a library dedicated to this https://developer.android.com/health-and-fitness/health-connect/test/unit-tests

We could do the same thing here.

Also, it would be cool to have a small android and ios app that could be used to populate data on each device, HealthKit/Health Connect for automated UI tests. I guess on Android you could use the toolkit apk, but I have tried to drive it with UI tests yet.

I do a lot of UI tests and the biggest pain is trying to drive the iOS UI to enter in step data because as some of you know a request for permissions can only be checked by reading and getting results. So, my tests all start with a wiped device and my UI test step goes into Health and enters a single step so when my app requests access, the Health dialog shows and I can toggle the switch, but at any time later the only "check" you can do to ensure that the permissions have not been revoked is to do a read op.

Anyway, a small app with documented UI that could be driven by Appium would be super convenient since you wouldn't have to deal with UI changes on different OS versions.

Just noting this really for future work, maybe others will see a benefit, depends on how you test I guess.

[Saurus119]

Hello @jmichas ,
In the big refactor that I am working on, you can direcly write/read metrics about health/workouts to the stores? Why would you need another MockService class?

In your tests you can crete sub class that will be mock of our interfaces and you can specify what it should return no ? You can define your own logic if you need anything extra or just setup what data it should return directly?

So instead of "Public nuget" in your test you use your MockIHealth which has same method and you define your return types?

[jmichas]

@Saurus119
So, for me, my app reads data written by other sources, ie Fitbit, Samsung Health, Google Fit, etc.
To test my app in the final QA it is on device and I test it actually generating steps and ensure that my app is reading them at an interval (on android) and when data changes (on iOS).

But prior to that testing I have a number of automated UI tests in Appium that test other critical functions like user reg, login, purchases, etc. I can't really test the steps part on a simulator or emulator. So what I was thinking was more of a separate app that could be set to write data at an interval for any health metric more than just a mock of the service itself. My intent is to integration test the actual calls to the health service and make sure there isn't something weird happening.

This may not be the place for it. I have actually already sketched out some of the things I would need in a test health data writer and may create that and release it to the public when its done.

Also, I know you are working on some massive changes. I have a bunch of stuff that I have added to support the android side of things, it just happens that android is what I am working on now, Maui.Health ios does not yet expose enough for me to use it the way I have implemented my own health service for ios. But I am on a deadline so I can't put the time into bringing it all together into Maui.Health yet.

[Kebechet]

@jmichas

This may not be the place for it. I have actually already sketched out some of the things I would need in a test health data writer and may create that and release it to the public when its done.

When u will release the testing project then let us know and at least we will link it into our README.
Because I think this wont be a part of our library.

Also, I know you are working on some massive changes.

Yes we are planning to implement quite a lot of stuff and after that do some final polishing of the code so that it is ready at least for production testing. After that is done and you have some customized functionality you would like to push to our package, feel free to create a PR.

Maui.Health ios does not yet expose enough for me to use it the way I have implemented my own health service for ios

For example whats things do you have implemented and are not a part of this library yet?

[jmichas]

@Kebechet

For example whats things do you have implemented and are not a part of this library yet?

For iOS I start a long running HKObserverQuery with a complex compound predicate watching for changes to step records. When that fires the HKObserverQueryUpdateHandler I create a HKAnchoredObjectQuery that gets the previously saved anchor and returns all records after transforming them into our internal custom record type. I am not sure, but I don't think Maui.Health handles this scenario yet.

On the Android side, what I have experienced is that there was no way to handle any rate limiting exceptions since that is a common problem. Much of my Android work is around flows that allow more user interactions if necessary. For instance I don't want to just popup the Play Store Health Connect page if you need to download it, I would like to check if that needs to be done, then I can choose to show the user something and then they can click a button to start the process. I see why things were done this way because they appear to follow the Android docs examples. Anyway, I figure I will wait until the big rework is done and then when I have time I can submit a PR for my stuff. If I submit it now, it will likely be hard to deal with and end up feeling like doing double the work.

Also, I noticed that Xamarin.AndroidX.Health.Connect.ConnectClient 1.1.0 was finally released. Hopefully that fixes some stuff.

[Saurus119]

-Ios: If I understand you correctly, what you need is to read out for exampel for Steps From/To. And in another reading you want to point to your latest reading? You can do it statticaly ia IhealthTime which we use to now in for the timewindow where you can read any "health data" which you wish.

For example:
You can read out <10AM - 11 AM>, you start your app you know last reading was at 11 you can query from <11:Now>. So I am not sure if any extra logic is needed for properly read the data? Internally in your app you can hold state when you did last reading?

Updating of lib version will be done by @Kebechet (to the latest stable) soon, as he is going to do some cleanups and stabilization before we publish this big refactor as "stable" version.

[Kebechet]

@jmichas

but I don't think Maui.Health handles this scenario yet.

As @Saurus119 mentioned we have a way to specify datetime ranges. But does Android have similar observers like u mentioned for iOS? Because we would like to expose just things that are common as this will be abstraction above both platforms

For instance I don't want to just popup the Play Store Health Connect page if you need to download it, I would like to check if that needs to be done, then I can choose to show the user something and then they can click a button to start the process

This certainly makes sense to have it implemented.

Anyway, I figure I will wait until the big rework is done and then when I have time I can submit a PR for my stuff. If I submit it now, it will likely be hard to deal with and end up feeling like doing double the work.

Exactly, thank you. I will try to do some work on it today. @Saurus119 PR got merged and now the library needs some polishing and as you mentioned we should bump the Android dependency to the stable one.

[jmichas]

Ok, so I know the date range queries exist and for Android that was the first thing I tried to use.
The one major issue I found with having a moving timeframe window to look up records is that we don't know when the records are written by 3rd parties. On Android testing with a phone recording steps with Google Fit to Health Connect, I found that steps were recoded into GF but not written to HC until after I had tried to read sometimes.

So, for instance, I read every 15 minutes starting from 00:00-00:15m some steps are recorded by fit, when I read that range at 00:15m the steps haven't been written by GF into HC so it comes back with zero records/zero steps. At 00:30, I read again but my window is 00:15-00:30, at this point GF has written the records to HC BUT they are dated when the steps were recorded by GF so they are written into the 00:00-00:15 window. The result is missed data on my end and I can never know to look back to a previous window.

This is where HKAnchoredQuery and Health Connects GetChanges/ChangeToken comes in.
I re implemented my Android version using client 1.1.0 and exposing a new method in Maui.Health service GetHealthDataChangesAsync(GetChangesRequest request, ct). This method handles getting a token if the request doesn't have one and then reading all records involved with the change, which may be multiple reads requests, see docs for "change token hasMore".

The result is that my moving window scenario is solved, I get any records that I haven't "synced" before when passing in the change token. My service then just manages persisting the change token.

This actually works great and moves Android and iOS in more alignment. The HKAnchoredQuery does essentially the same thing, you manage the "anchor" and pass that in to the query and it retrieves all the changes since the last anchor.

I have this all working in my fork of Maui.Health, it involved a number of changes to extensions methods because of the new type the change sets in android bring in UpsertionChange.

I have only stubbed out the iOS side of it for now, but when I have more time I will come back to it.

So with Android I am just polling using the change token every X minutes, with iOS I start my observer query and it runs a handler when changes occur, the handler creates and manages the anchor query. iOS is more efficient but it doesn't appear that Android has any type of pub/sub events to hook into so we can only poll.

[Saurus119]

@jmichas we’ve merged the big PR 👍

Just to make sure I understand you correctly: you’re able to add functionality so that when the app starts, it listens for changes in the stores. As you mentioned, this is efficient on iOS, but Android doesn’t provide the same mechanism.

Would you be able to create a PR where this is implemented behind a single public method that collects everything internally? Our goal is to have a unified API that works the same way on both platforms. From the user’s perspective, there should be no need to distinguish between iOS and Android—both platforms should call the same method, and the “magic” should happen under the hood. The internal implementations can differ as long as the observable behavior stays the same.

If my understading is right and you are able to create this for both platforms you can try add method which can collect this metrics, but are you able to do this for any metric that our lib provides? For example: Steps/Burned Kcal and so on? So we are not bouded only to one health metric.

[jmichas]

@Saurus119 This is great news. At some point I will try and pull in the new stuff into my fork so I can merge all my work Ive been doing and see how it all lines up.

So, yeah, on iOS you can fire up the query to watch the HealthKit database for whatever metric you want. When it reports a change you then can call it with the anchored query to get all the changes that have occurred since the last time you called with an anchor.

The anchored query lines up perfectly with the android changetoken/changeset that they expose in Health Connect, so that will be an easy thing to do as I have the android version of that working very well, the only thing to add is a little more around updates/deletes as I only do reads so I am not concerned with deleted data.

The observer thing is tricky. What I was thinking, and since I also use Shiny libs JobManager, is that for android we can fake the idea of the observer by starting a worker in the background that polls for changes using the changtoken. Then, if it finds any, it can just return them in an event or or observable or something. We can make it so the iOS version manages the observer, anchor and anchored query internally so you just tell it to start with a request that has the metric type you want to watch and your app just listens for changes to be emitted. This way we could surface a single api to both platforms that effectively returns data in the same way.

I will have more time in the beginning of February to work on all this stuff. I have most of the parts written as parts of my custom services for my app so I just need to pull it all together into something more cleaned up and probably better thought out.

[Saurus119]

Sounds good! @jmichas

• Deletion, yeah, may be tricky but I don´t suppose normal people go into stores and delete data manually (maybe weight and stuff like that), but most of scenarios should be fine as I don´t expect many users to modify them in the history.

• If we are able to create functionality that allows listen for changes (events), should be fantastic. The most "important" from my side that I would like to be able to watch easily for multiple "health data types", so yeah something like: ListenForHealthDataUpdate(<list<StepsDto, BUrnedKcalDto,.. and so on>) should do the work, so we can easily watch multiple data types and be able react on their changes.

• Have you had a chance to look for example into "ios" acitivty like "Running", I found out limitations that if you start any activity for example on Watch it runs only in the memory (IOS), so you have no way how to detect "active" workout from the app itself +collect metrics like HearthBeat it really send with low frequency (every 5 minutes one sample). Here I didn´t find a solution for example how to "mirror" current HearthRate from your active IOS workout to the app via our nuget.
  Here as I was digging around, everything is send to the "health DB" only after you finish the activity because of watch battery (they don´t send samples of data every sec and so on).