Improve error message when st.builds() is missing required uninferrable arguments

[sobolevn]

Right now hypothesis ignores @no_type_check decorator applied to functions and types.

For example:

from hypothesis import strategies as st, given

class NotTypes:
    def __init__(self, arg: str) -> None:
        self.arg = arg

@given(st.builds(NotTypes))
def test_no_type_check(instance):
    assert isinstance(instance, NotTypes)

This passes as expected. But, consider this:

from hypothesis import strategies as st, given
from typing import no_type_check

@no_type_check
class NotTypes:
    def __init__(self, arg: str) -> None:
        self.arg = arg

@given(st.builds(NotTypes))
def test_no_type_check(instance):
    assert isinstance(instance, NotTypes)

This example fails with unclear error:

_________________________________ test_no_type_check _________________________________

    @given(st.builds(NotTypes))
>   def test_no_type_check(instance):

ex.py:10: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:470: in run
    self._run()
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:870: in _run
    self.reuse_existing_examples()
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:548: in reuse_existing_examples
    data = self.cached_test_function(existing)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:1051: in cached_test_function
    self.test_function(data)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:208: in test_function
    self.__stoppable_test_function(data)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/engine.py:184: in __stoppable_test_function
    self._test_function(data)
.venv/lib/python3.9/site-packages/hypothesis/core.py:725: in _execute_once_for_engine
    escalate_hypothesis_internal_error()
.venv/lib/python3.9/site-packages/hypothesis/core.py:699: in _execute_once_for_engine
    result = self.execute_once(data)
.venv/lib/python3.9/site-packages/hypothesis/core.py:637: in execute_once
    result = self.test_runner(data, run)
.venv/lib/python3.9/site-packages/hypothesis/executors.py:47: in default_new_style_executor
    return function(data)
.venv/lib/python3.9/site-packages/hypothesis/core.py:596: in run
    args, kwargs = data.draw(self.search_strategy)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:878: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/collections.py:58: in do_draw
    return tuple(data.draw(e) for e in self.element_strategies)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/collections.py:58: in <genexpr>
    return tuple(data.draw(e) for e in self.element_strategies)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/strategies.py:818: in do_draw
    result = self.pack(data.draw(self.mapped_strategy))
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/lazy.py:164: in do_draw
    return data.draw(self.wrapped_strategy)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/strategies.py:818: in do_draw
    result = self.pack(data.draw(self.mapped_strategy))
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/collections.py:58: in do_draw
    return tuple(data.draw(e) for e in self.element_strategies)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/collections.py:58: in <genexpr>
    return tuple(data.draw(e) for e in self.element_strategies)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/lazy.py:164: in do_draw
    return data.draw(self.wrapped_strategy)
.venv/lib/python3.9/site-packages/hypothesis/internal/conjecture/data.py:874: in draw
    return strategy.do_draw(self)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = builds(NotTypes), data = ConjectureData(VALID, 0 bytes, frozen)

    def do_draw(self, data):
        try:
>           return self.target(
                *(data.draw(a) for a in self.args),
                **{k: data.draw(v) for k, v in self.kwargs.items()},
            )
E           TypeError: __init__() missing 1 required positional argument: 'arg'

.venv/lib/python3.9/site-packages/hypothesis/strategies/_internal/core.py:777: TypeError

I think it happens because get_type_hints return {} for types like NotTypes.
That's why we can't create a correct amount of arguments for __init__.

Probably, we can at least improve the error message to soething like:

Auto resolving types marked with `@no_type_check` is not supported. Explicit `register_type_strategy()` is required.

What do you think?

Context: https://bugs.python.org/issue46571

[Zac-HD]

This isn't specific to @no_type_check; it's a generic error message for any case where st.builds() is missing a strategy for a required unannotated argument.

I think it makes sense here to catch TypeError, check if our args and kwargs can't satisfy the required args for the target, and if so raise a TypeError(...) from None with a much more specific error message. That message might or might not have a variant for "you used @no_type_check" 🙂