iOS StateProvider crash

[chris-praxis]

Doing Https.request, got error:
Alamofire/SessionDelegate.swift:49: Fatal error: StateProvider is nil for task 1.

Relevant Stack:

#4	0x00000001a28e5b54 in assertionFailure ()
#5	0x000000010610bc2c in SessionDelegate.request<Alamofire.DataRequest>(for:as:) at /Users/chris/Projects/Mantis/mantis_laser_trainer_app/platforms/ios/Pods/Alamofire/Source/Core/SessionDelegate.swift:49
#6	0x000000010610f40c in SessionDelegate.urlSession(_:dataTask:didReceive:) at /Users/chris/Projects/Mantis/mantis_laser_trainer_app/platforms/ios/Pods/Alamofire/Source/Core/SessionDelegate.swift:254

Code:

    func request<R: Request>(for task: URLSessionTask, as type: R.Type) -> R? {
        guard let provider = stateProvider else {
            assertionFailure("StateProvider is nil for task \(task.taskIdentifier).")
            return nil
        }

        return provider.request(for: task) as? R
    }

[farfromrefug]

@chris-praxis thanks could you give details live version. Also can you share a reproducible request example ?

[chris-praxis]

CNR unfortunately. I'm using a custom API to pull a large blob of JSON, so not sure how to let you test. I only hit this stateProvider assert once. Sometimes response data is cut off (SyntaxError: Unexpected end of JSON input)... I don't know why that would occur.

To summarize my recent comments (ChrisA) in the #android channel...
I'm unable to get an HTTP implementation that pulls this large JSON blob consistently without error on both platforms. If I use core http, I get out-of-memory (after first time success) on Android. If I use 'https' latest v4, I get incorrect JSON parsing on iOS. If I use 'https' v5, I intermittently get unexpected end of JSON input or once I got stateProvider assert on iOS. I have not yet tested Downloader because it seems like the wrong tool for the job, requiring app to save payload to file and then load from file in order to parse.

[farfromrefug]

@chris-praxis what s your json size ?
Did you enable large heap on android?

Your case is special and might need a special.implementation which is out of the scope.of both core and this plugin.
When using really large files you should stream them and maybe even not load them entirely in memory.

I think you should investigate how this should be implemented even natively

Also you can use this plugin just like downloader to save file to disk (stream with progress)

[chris-praxis]

I know it's ridiculous, but I don't know how to get the size of the buffer prior to parsing. Regardless, this is a growing set of user data, so there's really no upper bound.
I have not enabled large heap.
It would be nice of core/http supported a streaming callback feature for handling data in chunks.
I cannot use Downloader because I need to POST query parameters for our custom API.

My workaround ATM is to use @nativescript-community/https on Android, but it's buggy on iOS so I'm still using the old core http there. Because this works makes me think there is a bug in core/http, possibly a memory leak. But there's no reason one of my users couldn't accumulate more data, so I think this is not a real solution.

So I need a streaming solution that includes POST. Probably the ideal is to download data in chunks/pages from the server, possibly as needed, but currently I don't think that's supported, at least not at our API level.

[farfromrefug]

@chris-praxis
1/ we want help.you without a reproductible exemple and detailed spécification about what you are trying to achieve
2/ we won't add edge case features if not needed by many users
3/ you need to investigate how this is done by others. I believe your approach is not the right one
4/ we can't help.you if you just say ios is buggy. Create an issue about it with details.and an example to reproduce

[chris-praxis]

1. Each time user performs an exercise/drill in my app, a session file is stored and uploaded to the server. After a user logs in, app sends POST to query those sessions saved on server. The content is JSON, and there's no size limit because a user can always do more sessions.
2. Understood, but I'm not finding any NS solution to stream large-ish downloads via POST.
3. I think a better approach is to have a paging feature in our API so that we can download sessions in manageable batches. I'm asking our server guy about this possibility.
4. I understand, it's difficult partly because results are inconsistent, and it's not trivial to take a complex app and make a simple test case for a bug report.

For this assert, which I only got once, I hoped you might check your code and find why 'stateProvider' would go away.

As for "I get incorrect JSON parsing on iOS," sometimes when I go to parse the returned content, 'toJSON' fails, other times it succeeds but produces invalid JSON, other times it seems to work okay. Now that I'm trying to repro of course it's not happening. It's as if the buffer is sometimes corrupted. I haven't seen these problems on Android. I doubt there's anything special about my app that's causing this.

[farfromrefug]

@chris-praxis don't only look at nativescript for a solution. The true power of Nativescript is that it allows you to port any code easily. This is why I told you to investigate how others do that.

As for ios buggy https. There could be edge cases still having issues. But for the rest https is used in production in tens of app all using api and json parsing without issue.
So.again without specific examples we can reproduce with we can't help.you.
And it will be the same with N.
You can also, and it would help a lot, try to debug those issues yourself.