Use Sequence instead of Union[List[…], Tuple[…]]

[clot27]

Working on #3350 with this

closes #3350

[Viicos]

As mentioned in the related issue, using Sequence can be problematic with str (see python/typing#256, I still haven't figured how to handle this case properly, so you might have some ideas!)

[clot27]

As mentioned in the related issue, using Sequence can be problematic with str (see python/typing#256, I still haven't figured how to handle this case properly, so you might have some ideas!)

as mentioned in the issue itself, using Sequence[something other than str] already excludes str :)

[Viicos]

As mentioned in the related issue, using Sequence can be problematic with str (see python/typing#256, I still haven't figured how to handle this case properly, so you might have some ideas!)

as mentioned in the issue itself, using Sequence[something other than str] already excludes str :)

Indeed but I see that there's a few places where Sequence[str] is used in this PR, so maybe this would require isinstance checks as stated in the issue comment you mentioned :)

This typing issue really is a problem in my opinion, as you can see in the typing issue mentioned this caused quite some trouble in production environments for some people :/

[Bibo-Joshi]

Indeed but I see that there's a few places where Sequence[str] is used in this PR, so maybe this would require isinstance checks as stated in the issue comment you mentioned :)

The meaning #3350 (comment) is that we have to make sure not to pass strings as list of chars to TG, e.g. for message.reply_text(text="Hi"), we must tell TG that the text is "Hi", not ["H", "i"].

In the cases of the parameters allowed_updates, custom_emoji_ids and options (the params currently typed as Sequence[str], such a check will have the following effect: If a user passes allowed_updates="message", we will pass the value "message" to TG and TG will probably complain because it expects a JSON array and not a string. So while a type checker will not complain, there will be exceptions at runtime, so I doubt that this will lead to hard-to-debug edge cases.

This typing issue really is a problem in my opinion, as you can see in the typing issue mentioned this caused quite some trouble in production environments for some people :/

You are right that providing a precise type annotation for this doesn't seem possible atm. Personally I'd still be okay with using Sequence[str] because

• it's a general limitation of Pynthon that we can do little about
• passing a false value to such a parameter will likely be easy to detect (see arguments above)
• so far only 3 parameters are affected …

[Viicos]

Yes I agree Sequence[str] should still be used as this is a limitation of the language itself. The way I'm doing it currently is using isinstance as a runtime check, but even without this check as you said TG will probably complain so this isn't really important.

(Some people suggested the idea of using a Diff type annotation (e.g. Diff[Sequence[str], str]), imo Annotated could also be used in the future when it will be correctly supported).

[Bibo-Joshi]

Nice job! :) Thanks for the PR! Harshil in fact already covered two comments that I had in the pipeline xD

[Bibo-Joshi]

Thank you very much for your work on this nice PR @clot27 🙂