date and datetime comparison: when we don't know how to

tim-one · tim-one · commit 8d81a012efa7 · 2003-01-24T22:36:34.000Z
compare against "the other" argument, we raise TypeError,
in order to prevent comparison from falling back to the
default (and worse than useless, in this case) comparison
by object address.

That's fine so far as it goes, but leaves no way for
another date/datetime object to make itself comparable
to our objects.  For example, it leaves Marc-Andre no way
to teach mxDateTime dates how to compare against Python
dates.

Discussion on Python-Dev raised a number of impractical
ideas, and the simple one implemented here:  when we don't
know how to compare against "the other" argument, we raise
TypeError *unless* the other object has a timetuple attr.
In that case, we return NotImplemented instead, and Python
will give the other object a shot at handling the
comparison then.

Note that comparisons of time and timedelta objects still
suffer the original problem, though.
diff --git a/Doc/lib/libdatetime.tex b/Doc/lib/libdatetime.tex
@@ -381,6 +381,14 @@ \subsection{\class{date} Objects \label{datetime-date}}
     comparison of date to date, where date1 is considered less than
     date2 when date1 precedes date2 in time.  In other words,
     date1 < date2 if and only if date1.toordinal() < date2.toordinal().
+    \note{In order to stop comparison from falling back to the default
+          scheme of comparing object addresses, date comparison
+          normally raises \exception{TypeError} if the other comparand
+          isn't also a \class{date} object.  However, \code{NotImplemented}
+          is returned instead if the other comparand has a
+          \method{timetuple} attribute.  This hook gives other kinds of
+          date objects a chance at implementing mixed-type comparison.}
+
 
   \item
     hash, use as dict key
@@ -711,6 +719,14 @@ \subsection{\class{datetime} Objects \label{datetime-datetime}}
     are compared.  If both comparands are aware and have different
     \member{tzinfo} members, the comparands are first adjusted by
     subtracting their UTC offsets (obtained from \code{self.utcoffset()}).
+    \note{In order to stop comparison from falling back to the default
+          scheme of comparing object addresses, datetime comparison
+          normally raises \exception{TypeError} if the other comparand
+          isn't also a \class{datetime} object.  However,
+          \code{NotImplemented} is returned instead if the other comparand
+          has a \method{timetuple} attribute.  This hook gives other
+          kinds of date objects a chance at implementing mixed-type
+          comparison.}
 
   \item
     hash, use as dict key
diff --git a/Lib/test/test_datetime.py b/Lib/test/test_datetime.py
@@ -880,6 +880,44 @@ def test_compare(self):
             self.assertRaises(TypeError, lambda: badarg > t1)
             self.assertRaises(TypeError, lambda: badarg >= t1)
 
+    def test_mixed_compare(self):
+        our = self.theclass(2000, 4, 5)
+        self.assertRaises(TypeError, cmp, our, 1)
+        self.assertRaises(TypeError, cmp, 1, our)
+
+        class AnotherDateTimeClass(object):
+            def __cmp__(self, other):
+                # Return "equal" so calling this can't be confused with
+                # compare-by-address (which never says "equal" for distinct
+                # objects).
+                return 0
+
+        # This still errors, because date and datetime comparison raise
+        # TypeError instead of NotImplemented when they don't know what to
+        # do, in order to stop comparison from falling back to the default
+        # compare-by-address.
+        their = AnotherDateTimeClass()
+        self.assertRaises(TypeError, cmp, our, their)
+        # Oops:  The next stab raises TypeError in the C implementation,
+        # but not in the Python implementation of datetime.  The difference
+        # is due to that the Python implementation defines __cmp__ but
+        # the C implementation defines tp_richcompare.  This is more pain
+        # to fix than it's worth, so commenting out the test.
+        # self.assertEqual(cmp(their, our), 0)
+
+        # But date and datetime comparison return NotImplemented instead if the
+        # other object has a timetuple attr.  This gives the other object a
+        # chance to do the comparison.
+        class Comparable(AnotherDateTimeClass):
+            def timetuple(self):
+                return ()
+
+        their = Comparable()
+        self.assertEqual(cmp(our, their), 0)
+        self.assertEqual(cmp(their, our), 0)
+        self.failUnless(our == their)
+        self.failUnless(their == our)
+
     def test_bool(self):
         # All dates are considered true.
         self.failUnless(self.theclass.min)
diff --git a/Misc/NEWS b/Misc/NEWS
@@ -100,6 +100,14 @@ Extension modules
   useful behavior when the optional tinzo argument was specified.  See
   also SF bug report <http://www.python.org/sf/660872>.
 
+  date and datetime comparison:  In order to prevent comparison from
+  falling back to the default compare-object-addresses strategy, these
+  raised TypeError whenever they didn't understand the other object type.
+  They still do, except when the other object has a "timetuple" attribute,
+  in which case they return NotImplemented now.  This gives other
+  datetime objects (e.g., mxDateTime) a chance to intercept the
+  comparison.
+
   The constructors building a datetime from a timestamp could raise
   ValueError if the platform C localtime()/gmtime() inserted "leap
   seconds".  Leap seconds are ignored now.  On such platforms, it's
diff --git a/Modules/datetimemodule.c b/Modules/datetimemodule.c
@@ -2437,8 +2437,15 @@ date_richcompare(PyDateTime_Date *self, PyObject *other, int op)
 	int diff;
 
 	if (! PyDate_Check(other)) {
+		if (PyObject_HasAttrString(other, "timetuple")) {
+			/* A hook for other kinds of date objects. */
+			Py_INCREF(Py_NotImplemented);
+			return Py_NotImplemented;
+		}
+		/* Stop this from falling back to address comparison. */
 		PyErr_Format(PyExc_TypeError,
-			     "can't compare date to %s instance",
+			     "can't compare '%s' to '%s'",
+			     self->ob_type->tp_name,
 			     other->ob_type->tp_name);
 		return NULL;
 	}
@@ -4018,6 +4025,11 @@ datetime_richcompare(PyDateTime_DateTime *self, PyObject *other, int op)
 	int offset1, offset2;
 
 	if (! PyDateTime_Check(other)) {
+		if (PyObject_HasAttrString(other, "timetuple")) {
+			/* A hook for other kinds of datetime objects. */
+			Py_INCREF(Py_NotImplemented);
+			return Py_NotImplemented;
+		}
 		/* Stop this from falling back to address comparison. */
 		PyErr_Format(PyExc_TypeError,
 			     "can't compare '%s' to '%s'",
