Allow missing __match_args__ to accept a single positional arg

[brandtbucher]

It is useful to be able to match against both the type and the value of a target. Currently, the only way to do this is to manually assign self to some attribute on the returned proxy, or use a guard.

How about this: if __match_args__ is None or is missing from the proxy (as it is from object, by default), we allow a single positional argument to ~test for equality~ continue matching against the proxy itself. As a result, all of these examples will just work by default:

from numbers import Real

specials = {"spam", "eggs", "cheese"}

match x:
    case bool(True):  # Doesn't match 1 or 1.0.
        ...
    case Real(42):
        ...
    case set(.specials):  # Doesn't match frozensets.
        ...

Currently, they must be spelled as:

match x:
    case bool(real=1):
        ...
    case Real() if x == 42:
        ...
    case set() if x == specials:
        ...

If a proxy wants to disable this behavior, it should just assign __match_args__ = [].

[Tobias-Kohn]

This looks very reasonable to me. I particularly like that this makes all these special cases with built-in objects and types just work. +1 from me.

[gvanrossum]

Yeah, I like this (in fact this is what I have often assumed would work). While it seems redundant to be able to write

match x:
  case int(i): ...

it's far from redundant as a subpattern:

match p:
  case Point(int(x), int(y)): # A point of two integers
    ...

[brandtbucher]

I just realized that this makes patterns like tuple([1, 2, 3]) work naturally too.

[gvanrossum]

I had hoped you had already implemented this, but it doesn't seem so:

>>> match "abc":
...   case str(x): print(x)
... 
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: too many positional matches in pattern

Also:

I just realized that this makes patterns like tuple([1, 2, 3]) work naturally too.

And this is actually an argument for allowing (...) for sequence patterns, since then we could write it as the much more satisfying tuple((1, 2, 3)).

[brandtbucher]

This is implemented now. I've only had time to test some simple cases so far, so let me know if you notice any issues!