Currently all the models sit in the 'models.py' file. I think we should reorganize models into separate files under a 'models' folder.

The API wrapper is no longer usable as aniwatch now has a recaptcha on their login page. Are there any plans to work around this?

Should we prioritize the merging of the experimental branch into master? The experimental branch is very far ahead of the master branch. I think we should focus on testing before we move on to adding new features. What do you think?

Should we start a new project on this?

just try to download demon slayer episode 1 sd, the only ts file i can play is intro, other chunk (part) gave error,

The experimental branch is broken due to importing issues. This is strange because my fork, which has basically the same code is working just fine,

API requests are being blocked by Cloudfare after the recent commit. I have double checked my username, user ID and auth token (called 'X-AUTH' token internally). Any ideas?

Currently, not passing in a value for the 'file_name' argument in the 'download' method just names the file as 'download.ts'. This needs to change as not only is it non-descriptive, it can also lead to issues when downloading more than one file as the new file just overwrites the old one since they have the same name.

be my friend c:

Currently we use the Session object as a context manager. This closes the Session immediately after every request, essentially sidestepping any performance benefits that using session objects can provide through connection pooling. This needs to change.

The requests should thus be made using try-except blocks.

The download completes without issues but the resultant .mp4 file is unplayable.

My guess

Skiping the first chunk (intro) causes this issue.

The 'headers 'and 'cookies' attributes are only used internally by data models to make requests, therefore they should be prefixed with double underscores.

Some of the methods in the main Sakurajima class return raw JSON responses. We should change this so that they return True if the operation was successful and False if some error occurred.

Recently we changed the way downloads are handled, ffmpeg is now required for downloads to work, I think we have made the API less accessible to new comers by introducing this pretty big requirement, The default media player on windows is able to play '.ts' files. Can we look into making converting to mp4 optional so that ffmpeg can be made an optional requirement to improve accessibility.

Kindly setup GIthub Pages to host documentation for the API. I think this will be easiest way for us to provide documentation, if you know of some other way kindly suggest that too. I am not too experienced with this sort of thing so any input will be appreciated.

The 'requirements.txt' file needs to be updated to reflect the newly added dependencies.

New dependencies:
pycryptodome==3.9.7
m3u8==0.6.0

We should look into adding the option to download the videos to a given directory instead of just saving it in the same folder as the script calling it.

Marking an already 'watched' episode as 'watched' again marks it as 'unwatched'.

Currently, we use from Sakurajima.models import * to import models into the api.py file. This leads to some unwanted import like AES and datetime.

Yesterday,I was working on documenting the Anime class. To be honest, I felt kind of limited using markdown for documentation, I felt that the formatting using vanilla markdown is just not satisfactory. I was looking into this and found out about a tool called Sphinx that can be used to document python modules.

Here is a video briefly showing what all this is about.

What do you think?

Currently the get_user_friends() method returns a raw JSON response. We need to make a data model class for this. Also I think that get_friends() is a better name for this method.

I have recently made quite a lot of changes to the API, but I have not talked about any of them. I apologize for this, I am making this thread to discuss recent changes, what I changed and my rationale behind it.

• Changes

  • All the code that deals with downloading episodes has been shifted to the modules in utils package. See #27
  • New classes namely Downloader, MultiThreadDownloader, ChunkDownloader, StatefulSegmentList, ChunkMerger,
    FFmpegMerger and ChunkRemover have been introduced.
  • Concurrency is now achieved using threads instead of processes. Since our use case is I/O limited, threads are a better
    choice due to the lower overhead.
  • The number of threads that are used for downloading can now be limited using the MultiThreadDownloader class.
  • Single threaded downloads now show a progress bar. I plan to extend this to multi threaded downloads as well.
  • Single threaded downloads can now be resumed using Kurisu I was unsure of where to put this so I made a separate repo
    for this. I will update this post with a link to that repo when I finish setting it up. (I have tested this and it works). Here is the
    link

All the new downloading changes are implemented using the download_using_downloader() method. The hope is that in the future we can simply rename this method to download() and remove the old download() method.

We need to have a page that lists all the data models that the API returns. Details page for each model can be created later.

Can you please add an experimental branch so that we can test out things before merging into the master branch.

We are sorely lacking in the documentation department. This will become an even greater issue when we finally decide to merge the experimental branch into master. With the introduction of data models, we are greatly increasing the complexity of the library.
We need to work on some form of documentation.

Currently, all the code that implements download functionality lives in the Episode class, The Episode class was originally meant to be a data model class, as things went on and we developed the download functionality the class became really bloated.

Why should we abstract downloading

• It will debloat the Episode class
• It will increase modularity
• It will make it possible to further extend downloading (like implementing progress bars) easier.

What do you think.

Trying to get an episode by its number using get_episode_by_number() method throws an error if there are no matching episodes.

We can instantiate a requests.Session() object in the Sakurajima class, set the headers and cookies and just pass that down to the data models instead of passing down cookies and headers. This will not only result in cleaner code but can also increase performance.