Tests for Python 3.1's treatment of negated imaginary literals.
authorMark Dickinson <dickinsm@gmail.com>
Wed, 30 Jun 2010 10:55:51 +0000 (10:55 +0000)
committerMark Dickinson <dickinsm@gmail.com>
Wed, 30 Jun 2010 10:55:51 +0000 (10:55 +0000)
Lib/test/test_complex.py

index a84e9495bbdd07293074e1f0172cf25705bc98ae..f28bb6902c30ab1b65deb17371341825e6212e1c 100644 (file)
@@ -407,6 +407,25 @@ class ComplexTest(unittest.TestCase):
             self.assertEquals(atan2(z1.imag, -1.), atan2(0., -1.))
             self.assertEquals(atan2(z2.imag, -1.), atan2(-0., -1.))
 
+    @unittest.skipUnless(float.__getformat__("double").startswith("IEEE"),
+                         "test requires IEEE 754 doubles")
+    def test_negated_imaginary_literal(self):
+        z0 = -0j
+        z1 = -7j
+        z2 = -1e1000j
+        # This behaviour is actually incorrect: the real part of a negated
+        # imaginary literal should be -0.0, not 0.0.  It's fixed in Python 3.2.
+        # However, the behaviour is already out in the wild in Python 2.x and
+        # Python <= 3.1.2, and it would be too disruptive to change it in a
+        # bugfix release, so we call it a 'feature' of Python 3.1, and test to
+        # ensure that the behaviour remains consistent across 3.1.x releases.
+        self.assertFloatsAreIdentical(z0.real, 0.0)
+        self.assertFloatsAreIdentical(z0.imag, -0.0)
+        self.assertFloatsAreIdentical(z1.real, 0.0)
+        self.assertFloatsAreIdentical(z1.imag, -7.0)
+        self.assertFloatsAreIdentical(z2.real, 0.0)
+        self.assertFloatsAreIdentical(z2.imag, -INF)
+
     @unittest.skipUnless(float.__getformat__("double").startswith("IEEE"),
                          "test requires IEEE 754 doubles")
     def test_overflow(self):