Don't exit 0 with linter errors (#26)

* Cosmetic: Make linter happy

* Fix test safe log test

Co-authored-by: Lukas Wutschitz <lukas.wutschitz@microsoft.com>

.github/workflows/test-package.yml

@@ -34,7 +34,7 @@ jobs:
         # stop the build if there are Python syntax errors or undefined names
         flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
         # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
-        flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
+        flake8 . --count --max-complexity=10 --max-line-length=127 --statistics
     - name: Test with pytest
       run: |
         pytest

prv_accountant/accountant.py

@@ -4,8 +4,6 @@
 import numpy as np
 from typing import Tuple
 
-from scipy import optimize
-
 from .other_accountants import RDP
 from . import discretisers
 from . import composers
@@ -55,7 +53,7 @@ class Accountant:
             mesh_size = 2*eps_error / np.sqrt(2*max_compositions*np.log(2/eta0))
 
         prv = privacy_random_variables.PoissonSubsampledGaussianMechanism(sampling_probability, noise_multiplier)
-        
+
         rdp = RDP(prv=prv, delta=self.delta_error/4)
         L = self.eps_error + rdp.compute_epsilon(max_compositions)[2]
         rdp = RDP(prv=prv, delta=self.delta_error/8/max_compositions)

prv_accountant/discrete_privacy_random_variable.py

@@ -34,7 +34,7 @@ class DiscretePrivacyRandomVariable:
             if i <= 0:
                 raise RuntimeError("Cannot compute epsilon")
             return np.log((d1[i-1]-delta_target)/d2[i-1])
-       
+
         eps_upper = find_epsilon(delta - delta_error) + epsilon_error
         eps_lower = find_epsilon(delta + delta_error) - epsilon_error
         eps_estimate = find_epsilon(delta)

prv_accountant/other_accountants.py

@@ -28,7 +28,7 @@ class RDP:
         Compute bounds on epsilon.
 
         This function is based on Google's TF Privacy:
-        https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py 
+        https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py
         """
         rdp_steps = self.rdp*num_compositions
         orders_vec = np.atleast_1d(self.orders)

prv_accountant/privacy_random_variables.py

@@ -17,8 +17,8 @@ M_PI = np.pi
 def log(x):
     valid = (x > 0)
     x_is_0 = (x == 0)
-    return np.where(valid, np.log(np.where(valid, x, 1)), 
-        np.where(x_is_0, -np.inf, np.nan))
+    return np.where(valid, np.log(np.where(valid, x, 1)),
+                    np.where(x_is_0, -np.inf, np.nan))
 
 
 class PrivacyRandomVariable(ABC):
@@ -43,7 +43,7 @@ class PrivacyRandomVariable(ABC):
         """
         raise NotImplementedError()
 
-        
+
 class PrivacyRandomVariableTruncated:
     def __init__(self, prv, t_min: float, t_max: float) -> None:
         self.prv = prv
@@ -102,7 +102,7 @@ class PoissonSubsampledGaussianMechanism(PrivacyRandomVariable):
     def cdf(self, t):
         sigma = self.sigma
         p = self.p
-        z = np.where(t>0, log((p-1)/p + exp(t)/p), log((p-1)/p + exp(t)/p))
+        z = np.where(t > 0, log((p-1)/p + exp(t)/p), log((p-1)/p + exp(t)/p))
         return np.where(t > log(1 - p), (
                 (1.0/2.0) * p * (-erfc(np.double((1.0/4.0)*M_SQRT2*(2*pow(sigma, 2)*z - 1)/sigma))) -
                 1.0/2.0*(p - 1) * (-erfc(np.double((1.0/4.0)*M_SQRT2*(2*pow(sigma, 2)*z + 1)/sigma))) + 1
@@ -114,8 +114,8 @@ class PoissonSubsampledGaussianMechanism(PrivacyRandomVariable):
     def rdp(self, alpha: float) -> float:
         """
         Compute RDP of this mechanism of order alpha
-        
-        Based on Google's TF Privacy: https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py 
+
+        Based on Google's TF Privacy: https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py  # noqa: E501
         """
         if self.p == 0:
             return 0
@@ -130,7 +130,7 @@ class PoissonSubsampledGaussianMechanism(PrivacyRandomVariable):
 
 
 # The following code is based on Google's TF Privacy
-# https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py 
+# https://github.com/tensorflow/privacy/blob/master/tensorflow_privacy/privacy/analysis/rdp_accountant.py
 
 def _compute_log_a(q, sigma, alpha):
     """Compute log(A_alpha) for any positive finite alpha."""

tests/test_privacy_random_variables.py

@@ -1,7 +1,6 @@
 # Copyright (c) Microsoft Corporation.
 # Licensed under the MIT License.
 
-from typing import AsyncContextManager
 import scipy
 import math
 import pytest
@@ -10,9 +9,10 @@ import sys
 
 from prv_accountant.privacy_random_variables import PoissonSubsampledGaussianMechanism, log, PrivacyRandomVariableTruncated
 
+
 def test_safe_log():
-    assert np.isnan(log(-1)) == True
-    assert np.isneginf(log(0)) == True
+    assert np.isnan(log(-1)) == True  # noqa: E712
+    assert np.isneginf(log(0)) == True  # noqa: E712
     assert log(1) == pytest.approx(0)