13 swarm_learning = SwarmLearning(X, y)
14 swarm_learning.train()
15 accuracy = swarm_learning.evaluate(X, y)
16 assert accuracy >= 0.5, "The accuracy of the model is low"17
18if __name__ == '__main__':
19 test_swarm_learning()12
13 quantum_ai = QuantumAI(X, y)
14 accuracy = quantum_ai.evaluate(X, y)
15 assert accuracy >= 0.5, "The accuracy of the model is low"16
17if __name__ == '__main__':
18 test_quantum_ai()12
13 federated_learning = FederatedLearning(X, y)
14 accuracy = federated_learning.train()
15 assert accuracy >= 0.5, "The accuracy of the model is low"16
17if __name__ == '__main__':
18 test_federated_learning()13 edge_ai = EdgeAI(X, y)
14 edge_ai.train()
15 accuracy = edge_ai.evaluate(X, y)
16 assert accuracy >= 0.5, "The accuracy of the model is low"17
18if __name__ == '__main__':
19 test_edge_ai()13 automl = AutoML(X, y)
14 automl.train()
15 accuracy = automl.evaluate(X, y)
16 assert accuracy >= 0.5, "The accuracy of the model is low"17
18if __name__ == '__main__':
19 test_automl()Usage of assert statement in application logic is discouraged. assert is removed with compiling to optimized byte code. Consider raising an exception instead. Ideally, assert statement should be used only in tests.
Python has an option to compile the optimized bytecode and create the respective .pyo files by using the options -O and -OO. When used, these basic optimizations are done:
- All the assert statements are removed
- All docstrings are removed (when -OO is selected)
- Value of the
__debug__built-in variable is set toFalse
It is recommended not to use assert in non-test files. A better way for internal self-checks is to check explicitly and raise respective error using an if statement.
Tip: Make sure test_patterns are defined in .deepsource.toml to avoid false-positives. Please check the documentation to know more.
Consider this code snippet:
def read_secret(self):
assert self.is_admin, "You are unauthorized to read this"
return self._secret
If python is run with the -O flag, the check for self.is_admin is
completely ignored, which can cause secrets to be leaked.
This is how you can ensure the code always works:
def read_secret(self):
if not self.is_admin:
raise AssertionError("You are unauthorized to read this")
return self._secret
Here's a more detailed example. Consider the following script foo.py:
import sys
def run():
assert len(sys.argv) == 5 # Insecure, statement will be removed when compiled to optimized byte code
print("Argument variables are: ", sys.argv)
run()
When optimization is disabled:
$ python foo.py 1 2 3 4 5
Traceback (most recent call last):
File "foo.py", line 7, in <module>
run()
File "foo.py", line 4, in run
assert len(sys.argv) == 5 # Insecure, statement will be removed when compiled to optimized byte code
AssertionError
When optimization is enabled:
$ python -O foo.pyo 1 2 3 4 5 6
Argument variables are: ['foo.pyo', '1', '2', '3', '4', '5', '6']
Here, all the internal self-checks using the assert statements are removed, as we can see. Therefore, there's a chance for an application to behave strangely in this case. It is better do raise the Exception explicitly:
import sys
def run():
if not len(sys.argv) == 5:
raise ValueError
print("Argument variables are: ", sys.argv)
run()
Note: During autofix, DeepSource will change the assert statements to if statements raising AssertionError.
This is done to replicate the existing behavior.
References:
- -O flag in python
- -OO flag in python
- Discuss blogpost - Using assert outside tests