Add time component awareness to TimeVariable.

biolab · sstanovnik · Jun 10, 2016 · Jun 17, 2016 · Jun 17, 2016 · Jun 22, 2016
commit 118a8d1c0b983eed9bf31d3cc84ca717ad4fedc8
diff --git a/Orange/data/variable.py b/Orange/data/variable.py
@@ -639,6 +639,8 @@ def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
         # None if no timezone, pytz object if any timezone
         self.timezone = None
+        # True if the time component exists (and will be displayed), parsing in column_to_datetime
+        self.has_time_component = True
 
     @staticmethod
     def column_looks_like_time(column):
@@ -725,11 +727,23 @@ def column_to_datetime(self, column):
         # .apply(str) because parsing a float discards the fractional part for some reason
         # utc=True: make timezone aware
         # .values: return a DatetimeIndex so we can actually localize to UTC
-        return pd.to_datetime(column.apply(str).values, errors='raise', exact=True, utc=True,
-                              infer_datetime_format=True, **kwargs)
+        result = pd.to_datetime(column.apply(str).values, errors='raise', exact=True, utc=True,
+                                infer_datetime_format=True, **kwargs)
+
+        # determine whether we should display the time part (HH:MM:SS.MS)
+        # only display it when hours, minutes and seconds are all 0 in all cases
+        # this is the most robust way as it doesn't depend on regexes
+        # (we might not even have strings as inputs) and offloads any parsing to pandas
+        # addition works because everything is non-negative
+        self.has_time_component = (result.hour + result.minute + result.second + result.microsecond).sum() != 0
+        return result
 
     def repr_val(self, val):
-        return str(val.tz_convert(self.timezone)) if val is not pd.NaT else "?"
+        if val is pd.NaT:
+            return "?"
+        else:
+            tzval = val.tz_convert(self.timezone)
+            return str(tzval) if self.has_time_component else tzval.strftime("%Y-%m-%d")
 
     str_val = repr_val
 

diff --git a/Orange/tests/test_variable.py b/Orange/tests/test_variable.py
@@ -290,16 +290,16 @@ class TestTimeVariable(VariableTest):
         ('2015-10-12 14:13', 1444659180, '2015-10-12 14:13:00'),
         ('20151012T141311', 1444659191, '2015-10-12 14:13:11'),
         ('20151012141311', 1444659191, '2015-10-12 14:13:11'),
-        ('2015-10-12', 1444608000, '2015-10-12 00:00:00'),
-        ('20151012', 1444608000, '2015-10-12 00:00:00'),
-        ('2015-10', 1443657600, '2015-10-01 00:00:00'),
-        ('2015', 1420070400, '2015-01-01 00:00:00'),
+        ('2015-10-12', 1444608000, '2015-10-12'),
+        ('20151012', 1444608000, '2015-10-12'),
+        ('2015-10', 1443657600, '2015-10-01'),
+        ('2015', 1420070400, '2015-01-01'),
         ('01:01:01.01', TODAY_TIMESTAMP + 3661.01, TODAY_STRING + ' 01:01:01.010000'),
         ('01:01:01', TODAY_TIMESTAMP + 3661, TODAY_STRING + ' 01:01:01'),
         ('01:01', TODAY_TIMESTAMP + 3660, TODAY_STRING + ' 01:01:00'),
-        ('1970-01-01 00:00:00', 0, '1970-01-01 00:00:00'),
+        ('1970-01-01 00:00:00', 0, '1970-01-01'),
         ('1969-12-31 23:59:59', -1, '1969-12-31 23:59:59'),
-        ('1900-01-01', -2208988800, '1900-01-01 00:00:00'),
+        ('1900-01-01', -2208988800, '1900-01-01'),
         ('nan', np.nan, '?'),
         (1444651991.81, 1444651991.81, '2015-10-12 12:13:11.810000'),
     ]
@@ -353,7 +353,7 @@ def test_readwrite_timevariable(self):
 
         table = CSVReader(input_csv).read()
         self.assertIsInstance(table.domain['Date'], TimeVariable)
-        self.assertEqual(table.domain['Date'].repr_val(table['Date'].iloc[0]), '1920-12-12 00:00:00')
+        self.assertEqual(table.domain['Date'].repr_val(table['Date'].iloc[0]), '1920-12-12')
         # Dates before 1970 are negative
         self.assertTrue((table['Date'] < '1970-01-01').all())
         self.assertTrue(all(table.domain['Date'].to_val(inst['Date']) < 0 for idx, inst in table.iterrows()))