Small update: Apparently pyright still thinks that PTb is not very type complete:

{
        "exportedSymbolCounts": {
            "withKnownType": 415,
            "withAmbiguousType": 30,
            "withUnknownType": 294
        },
        "otherSymbolCounts": {
            "withKnownType": 1815,
            "withAmbiguousType": 702,
            "withUnknownType": 566
        },
        "missingFunctionDocStringCount": 25,
        "missingClassDocStringCount": 0,
        "missingDefaultParamCount": 107,
        "completenessScore": 0.5615696887686062,
}

Full json output.

Update:

{
        "exportedSymbolCounts": {
            "withKnownType": 426,
            "withAmbiguousType": 277,
            "withUnknownType": 36
        },
        "otherSymbolCounts": {
            "withKnownType": 2292,
            "withAmbiguousType": 671,
            "withUnknownType": 120
        },
        "missingFunctionDocStringCount": 25,
        "missingClassDocStringCount": 0,
        "missingDefaultParamCount": 107,
        "completenessScore": 0.5764546684709067,
}

So at least a rather large jump from "unknown" to "ambiguous - even though the completeness increased only ~1.5 points 😄
I found out that as soon as anything in a class A is unknown/ambiguous (annotation of param/return value/attr, base class), that whole class A is of unknown/ambiguous type and using this class A in some other class B as type annotation will hence also make B unknown/ambiguous.

Also I apparently messed up the tests big time

{
        "exportedSymbolCounts": {
            "withKnownType": 705,
            "withAmbiguousType": 0,
            "withUnknownType": 34
        },
        "otherSymbolCounts": {
            "withKnownType": 3180,
            "withAmbiguousType": 0,
            "withUnknownType": 102
        },
        "missingFunctionDocStringCount": 23,
        "missingClassDocStringCount": 0,
        "missingDefaultParamCount": 147,
        "completenessScore": 0.9539918809201624,
}

😎

Last update for today messe up the tests, but got to 96%:

Symbols exported by "telegram": 734
  With known type: 707
  With ambiguous type: 0
  With unknown type: 27
    (Ignoring unknown types imported from other packages)
  Functions without docstring: 13
  Functions without default param: 0
  Classes without docstring: 0

Other symbols referenced but not exported by "telegram": 2496
  With known type: 2408
  With ambiguous type: 1
  With unknown type: 87

Type completeness score: 96.3%

IISC most of what's left boils down to the generic nature of Application and CallbackContext. I'll see how far I'll get with that …

The GH Actions thingy doesn't quite work yet and sphinx is acting up while documenting ContextTypes.DEFAULT_TYPES. but type completeness is at 100% now 🥳
If anyone's using VSCode + pylance, I'd be happy to know if this improves autocompletion etc noticable :)

Okay so I tested it and it's just as good as before, except now the autocompletion of context.{user, chat,bot}_data works better. Also sadly it still shows None for something like update.message.poll.id, though I guess we can't fix that without a TypeGuard(?).

Anyway running pyright on my machine still found some unknown types, sending a pastebin in dev chat

@harshil21 Nice :)
I doubt that TypeGuard can help with that. To know in which handlers update.message.poll is None and in which it's a Message, we'd have to make Update and Message generic classes and couple that with the handlers & filters - (nearly) impossible

which Python version are did you run pyright on? did you use the --ignoreexternal flag?

Is it worth to also modify bot method parameters to Optional[...] as well? Even the newer mypy version complains of implicit optionals now

You mean e.g. parse_mode: Optional[str] = None instead of parse_mode: str = None? Does mypy copmlain about that when you use PTB or only when developing PTB?

You mean e.g. parse_mode: Optional[str] = None instead of parse_mode: str = None?

yes

Does mypy copmlain about that when you use PTB or only when developing PTB?

pylance complains about that while developing PTB. I mentioned mypy since they made a similar change with implicit_optional

pylance complains about that while developing PTB.

So this changed? #3367 (comment)

If it has: Fair enough, let's do it. But I vote to do that in a different PR - we can reopen #3367 for that.