fix duration can not be paresd

[theScrabi]

this will fix the duration error that came up recently:
#176

[Stypox]

Before reading GitHub notifications I delved into playerArgs too, and I found out how to fix this bug by myself. Only then I saw you had fixed it already :-).

By the way playerArgs["player_response"] contains many interesting things (among which "videoDetails object, formats...). What about replacing all suitable functions in the YoutubeStreamExtractor with player_response-based functions? As you said it is simple and fast to parse json, and player_response seems to have the same format in "youtube.com/get_video_info" so we could remove some duplicate code for age restricted videos. But this is for another pr. :-D

[Stypox]

Oh, by skimming through YoutubeStreamExtractor I also found out why NewPipe always fetches location-dependent strings instead of the chosen ones. It is because downloader.download(...) is called only with the url and not with the localization parameter. But this is also for another pr.

[Stypox]

Sorry if I do another review, but I didn't see you had fixed sts too. ;-)
Btw if you think there is something wrong in the way I do reviews please tell me, since I haven't done many of them.

[theScrabi]

This:

The F****** extractor is literally untestable at the moment. Everytime you run it other tests fail. We need to fix that otherwise our whole CI is BullS****.

[theScrabi]

Or there is a S***load of changes on the youtube webpage coming around, and sometimes the extractor hits those servers with the new site.

13 failed. This is a catastrophe.

[Stypox]

🤔 Many of the errors are caused by ReCaptcha...

[theScrabi]

thinking Many of the errors are caused by ReCaptcha...

uff. So how are we going to tell the the CI?
Maybe we "bruteforced" youtube from our CI so often that they made Travis use ReCaptcha.

[Stypox]

What about creating a custom test runner that, for example, ignores ReCaptcha exceptions and only generates warnings for classes where more than 60% of test functions fail?

This happened to me once, and I wasn't able to reproduce it afterwards. No ReCaptcha exception was thrown, but the downloaded page lacked content anyway. With the custom runner you would just be warned that YoutubePlaylistExtractorRest$HugePlaylist couldn't be run.

Another option would be to wait some time before every network request: this way youtube should not create robot-verification problems.

[theScrabi]

Yes :D, the idea with a custom test runner sounds good. Was it possible to "re run" the test if it failed? For that we could check in the setup function of the test class weather the request went through well.

[Stypox]

Now all seems to be ok. I ran all the YouTube tests on my pc until I got a random error, and they passed 10 times in a row (probably my pc is slower than the build one, so it overloads Youtube servers much less). This means that the code is all good, we just need to add a custom test runner in another pr just after this one.

The things I pointed out in this review are just small ones, then the pr is ready :-)

[Stypox]

I think we're good now. Thank you :-)