I'm trying to use assertions to show some invariants (mostly in testing) Thus i want to write something like the following:
values = [ range(10) ]
expected_values = [ range(10) ]
map (lambda x: assert x[0] == x[1] ,zip( [ run_function(i) for i in values ], expected_values))
If I use this with unittest.assertEqual this works perfectly fine , but if I want to write this with an assertion it just fails. Is there a way to fix this?
From the documentation:
Note that functions created with lambda forms cannot contain statements.
So no, you cannot use the assert statement in a lambda expression.
assert statement. You are achieving a similar effect by throwing the error, but that doesn't invalidate what I said.
Unfortunately, assert is a statement and Pythons limited lambdas don't allow that in them. They also restrict things like print.
You can use a generator expression here though.
assert all(x[0] == x[1] for x in zip( [run_function(i) for i in values ], expected_values))
I personally think that the following would be more readable
assert all(run_function(i) == j for i,j in zip(inputs, expected_values))
Actually you can:
assertion_raiser = lambda: (_ for _ in ()).throw(AssertionError("My Lambda CAN raise an assertion!"))
Here is some validation:
try:
assertion_raiser()
except AssertionError:
print("assertion caught")
Worked in my application, which was similar:
def _lambda_assert(fn, *args, **kwargs):
def _assert(fn_out):
if isinstance(fn_out, (list, tuple)):
value, fail_msg = fn_out
assert value, fail_msg
else:
assert fn_out # value only
return lambda: _assert(fn(*args, **kwargs))
# return lambda x: _assert(fn(*args, **kwargs)) # for pytest; need dummy positional
Usage example:
import numpy as np
def _test_precision(size, atol=1e-2):
x = np.random.randn(size)
y = np.random.randn(size)
absdiff = np.abs(x - y)
fail_msg = f"absdiff: {absdiff}; atol={atol}"
return np.allclose(x, y, atol=atol), fail_msg
lambda_assert = _lambda_assert(_test_precision, 20, atol=1)
value and fail_msg are derived is highly flexible. For example, I actually fetched both from yet another function: value, fail_msg = _fn(), where _fn() is arbitrarily complex. It was among the few things I had to craft to hack around PyTest's various requirements.One simple solution is to define a auxiliary function that raises an assert given an input boolean value:
def assert_(cond): assert cond
values = [ range(10) ]
expected_values = [ range(10) ]
map (lambda x: assert_(x[0] == x[1]) ,zip( [ run_function(i) for i in values ], expected_values))
