WIP: SenderPool pop_updates #49

HBcao commented 2026年07月13日 08:34:34 +02:00

• Add method SenderPool::pop_updates().
• Re-added the MessageBox and ChatHashes that I deleted in my last PR.
• Make the process_socket_updates function asynchronous
• I only did a simple test of events.Raw and events.NewMessage. The test scripts need to be added (I don't really like writing them).

By the way, should we remove unused PySender python bindings? By doing this, we can remove the dependency on grammers-tl-types.

- Add method `SenderPool::pop_updates()`. - Re-added the MessageBox and ChatHashes that I deleted in my last PR. - Make the `process_socket_updates` function asynchronous - I only did a simple test of `events.Raw` and `events.NewMessage`. The test scripts need to be added (I don't really like writing them). By the way, should we remove unused `PySender` python bindings? By doing this, we can remove the dependency on `grammers-tl-types`.

HBcao commented 2026年07月14日 04:52:00 +02:00

@Lonami wrote in #49 (comment):

Thanks for the huge amount of work once again.

But from experience update handling is very tricky, and I would really prefer if there was a single source of truth that was able to handle update ordering and gaps correctly. That being grammers-session.

So currently I haven't reviewed any file under message_box/, because I would prefer if we could instead bridge most functionality to grammers. Or even have grammers give us the processed updates in order instead of duplicating that functionality in Telethon.

We would still need to duplicate some of the logic to Python (perhaps a good chunk of Client::next_raw; depends who performs the requests) because we still need a way for Python to initiate the process to getDifference but we would still be reusing a lot more.

I don't have in-depth knowledge of updates, so I can only make as few changes as possible to the existing code in v2. I shouldn't have removed the message_box in the last PR, which makes this PR seem like it did a lot of work.

@Lonami wrote in https://codeberg.org/Lonami/Telethon/pulls/49#issuecomment-19138382: > Thanks for the huge amount of work once again. > > But from experience update handling is very tricky, and I would really prefer if there was a single source of truth that was able to handle update ordering and gaps correctly. That being grammers-session. > > So currently I haven't reviewed any file under `message_box/`, because I would prefer if we could instead bridge most functionality to grammers. Or even have grammers give us the processed updates in order instead of duplicating that functionality in Telethon. > > We would still need to duplicate some of the logic to Python (perhaps a good chunk of [`Client::next_raw`](https://codeberg.org/Lonami/grammers/src/commit/c34290345a4636ff689cec312e655419e5972609/grammers-client/src/client/updates.rs#L105); depends who performs the requests) because we still need a way for Python to initiate the process to `getDifference` but we would still be reusing a lot more. I don't have in-depth knowledge of updates, so I can only make as few changes as possible to the existing code in v2. I shouldn't have removed the message_box in the last PR, which makes this PR seem like it did a lot of work.

Lonami commented 2026年07月14日 22:36:23 +02:00

All the more reason to try to reuse more from grammers then.

Telethon v1 barely did any proper update handling for the longest time. Then came grammers, then Telethon v2. But Telethon v2 stopped receiving changes from me while grammers' update handling continued to evolve.

So I think we should use as much from grammers as possible. This means no Python implementation of the message box (unless we ported it anew from grammers, but that would mean maintaining the logic twice). So rather, I suggest another bridge to grammers' MessageBox.

All the more reason to try to reuse more from grammers then. Telethon v1 barely did any proper update handling for the longest time. Then came grammers, then Telethon v2. But Telethon v2 stopped receiving changes from me while grammers' update handling continued to evolve. So I think we should use as much from grammers as possible. This means no Python implementation of the message box (unless we ported it anew from grammers, but that would mean maintaining the logic twice). So rather, I suggest another bridge to grammers' MessageBox.

HBcao commented 2026年07月15日 03:15:02 +02:00

@Lonami wrote in #49 (comment):

All the more reason to try to reuse more from grammers then.

Telethon v1 barely did any proper update handling for the longest time. Then came grammers, then Telethon v2. But Telethon v2 stopped receiving changes from me while grammers' update handling continued to evolve.

So I think we should use as much from grammers as possible. This means no Python implementation of the message box (unless we ported it anew from grammers, but that would mean maintaining the logic twice). So rather, I suggest another bridge to grammers' MessageBox.

Okay, bridge MessageBoxes and UpdateStream.

Should I commit directly under this PR or create a new PR?

@Lonami wrote in https://codeberg.org/Lonami/Telethon/pulls/49#issuecomment-19229252: > All the more reason to try to reuse more from grammers then. > > Telethon v1 barely did any proper update handling for the longest time. Then came grammers, then Telethon v2. But Telethon v2 stopped receiving changes from me while grammers' update handling continued to evolve. > > So I think we should use as much from grammers as possible. This means no Python implementation of the message box (unless we ported it anew from grammers, but that would mean maintaining the logic twice). So rather, I suggest another bridge to grammers' MessageBox. Okay, bridge MessageBoxes and UpdateStream. Should I commit directly under this PR or create a new PR?