[iOS] Synchronized tables, separated requests and network logic and other...

[chrusart]

Hi,

this PR came from crashes we still get (some description):
wcandillon/react-native-img-cache#95

...but evolved.

Whole RNFetchBlobNetwork class actually do two things:

• keeping tasks references to cancel them, change their progress config etc.,
• session and requests logic.

I moved session and requests logic into separate file, now it's more clean what do what.
NSMapTable seemed to be used incorrectly or just it's (weak) features not used (it seems from log NSLog(@"object released by ARC."); that it should or was used), anyway, it's now instantiated with NSMapTableWeakMemory option for objects, so requests will be released, default in NSMapTable is to use strong references.

sharedInstance is used instead tables and dictionaries instances created on load. This pattern is used all around but please give me a feedback why previous is better solution, maybe I missed something.

Other things done in this PR when implementing:

• checkExpiredSessions didn’t find anywhere, removed from header
• __block storage type not needed, variables not used in blocks in scope
• extern storage type not used in .m files
• nullability type in RNFetchBlobNetwork, should be checked and applied for method arguments in other classes
• unimplemented + (void) enableProgressReport:(NSString *) taskId, + (void) enableUploadProgress:(NSString *) taskId, - (void) sendRequest method declarations removed
  unused fileTaskCompletionHandler and dataTaskCompletionHandler properties and synthesizers removed
• other small warning messages fixes
• every single request is created on new RNFetchBlobNetwork instance so taskQueue.maxConcurrentOperationCount seems not work as expected -> made taskQueue instantiated per RNFetchBlobNetwork instance.

I can't find any place where actually something is added to expirationTable.

Does HTTPMaximumConnectionsPerHost = 10 have any impact as new session is created for every request (except defaultSessionConfiguration)? In docs it says that it applies to all sessions '...based on this configuration', what actually this means, the same instance, identifier or some comparison???

I'm aware of few fixes that are in 0.10.9 not included here.

TODO:

• subclass RCTEventEmitter to send events
• expired task logic?

This is WIP, works for now for us, but didn't test everything and any feedback or testing would be great.

[chrusart]

I added it at the end of the day quickly, fixing sharedInstance access now.

[chrusart]

What about applying:
http://google.github.io/styleguide/objcguide.html
I agree with that code guideline mostly.

[chrusart]

https://github.com/flatfox-ag/react-native-fetch-blob/tree/exception_fixes 2 days already in production, no exceptions yet.

This branch have all these fixes:
#619
#499

and all changes from:
https://github.com/flatfox-ag/react-native-fetch-blob/tree/synchronized

DOES NOT HAVE wkh237:0.11.0 commits listed in this PR!

feel free to:
"react-native-fetch-blob": "github:flatfox-ag/react-native-fetch-blob#exception_fixes",

As we use it in production I removed WIP note.
Stuff in TODO will come in different PRs I guess.

[chrusart]

Week without crashes on production, fixes confirmed :)

[lll000111]

FYI

react-native-fetch-blob has a new maintained location. Check out the README.

Please consider submitting your PR there, this repository is unmaintained.