from datetime import date import dateutil import numpy as np import pytest import pandas as pd from pandas import ( DataFrame, DatetimeIndex, Index, Timestamp, date_range, offsets, ) import pandas._testing as tm class TestDatetimeIndex: def test_time_loc(self): # GH8667 from datetime import time from pandas._libs.index import _SIZE_CUTOFF ns = _SIZE_CUTOFF + np.array([-100, 100], dtype=np.int64) key = time(15, 11, 30) start = key.hour * 3600 + key.minute * 60 + key.second step = 24 * 3600 for n in ns: idx = date_range("2014-11-26", periods=n, freq="S") ts = pd.Series(np.random.randn(n), index=idx) i = np.arange(start, n, step) tm.assert_numpy_array_equal(ts.index.get_loc(key), i, check_dtype=False) tm.assert_series_equal(ts[key], ts.iloc[i]) left, right = ts.copy(), ts.copy() left[key] *= -10 right.iloc[i] *= -10 tm.assert_series_equal(left, right) def test_time_overflow_for_32bit_machines(self): # GH8943. On some machines NumPy defaults to np.int32 (for example, # 32-bit Linux machines). In the function _generate_regular_range # found in tseries/index.py, `periods` gets multiplied by `strides` # (which has value 1e9) and since the max value for np.int32 is ~2e9, # and since those machines won't promote np.int32 to np.int64, we get # overflow. periods = np.int_(1000) idx1 = date_range(start="2000", periods=periods, freq="S") assert len(idx1) == periods idx2 = date_range(end="2000", periods=periods, freq="S") assert len(idx2) == periods def test_nat(self): assert DatetimeIndex([np.nan])[0] is pd.NaT def test_week_of_month_frequency(self): # GH 5348: "ValueError: Could not evaluate WOM-1SUN" shouldn't raise d1 = date(2002, 9, 1) d2 = date(2013, 10, 27) d3 = date(2012, 9, 30) idx1 = DatetimeIndex([d1, d2]) idx2 = DatetimeIndex([d3]) result_append = idx1.append(idx2) expected = DatetimeIndex([d1, d2, d3]) tm.assert_index_equal(result_append, expected) result_union = idx1.union(idx2) expected = DatetimeIndex([d1, d3, d2]) tm.assert_index_equal(result_union, expected) # GH 5115 result = date_range("2013-1-1", periods=4, freq="WOM-1SAT") dates = ["2013-01-05", "2013-02-02", "2013-03-02", "2013-04-06"] expected = DatetimeIndex(dates, freq="WOM-1SAT") tm.assert_index_equal(result, expected) def test_stringified_slice_with_tz(self): # GH#2658 start = "2013-01-07" idx = date_range(start=start, freq="1d", periods=10, tz="US/Eastern") df = DataFrame(np.arange(10), index=idx) df["2013-01-14 23:44:34.437768-05:00":] # no exception here def test_append_nondatetimeindex(self): rng = date_range("1/1/2000", periods=10) idx = Index(["a", "b", "c", "d"]) result = rng.append(idx) assert isinstance(result[0], Timestamp) def test_iteration_preserves_tz(self): # see gh-8890 index = date_range("2012-01-01", periods=3, freq="H", tz="US/Eastern") for i, ts in enumerate(index): result = ts expected = index[i] assert result == expected index = date_range( "2012-01-01", periods=3, freq="H", tz=dateutil.tz.tzoffset(None, -28800) ) for i, ts in enumerate(index): result = ts expected = index[i] assert result._repr_base == expected._repr_base assert result == expected # 9100 index = DatetimeIndex( ["2014-12-01 03:32:39.987000-08:00", "2014-12-01 04:12:34.987000-08:00"] ) for i, ts in enumerate(index): result = ts expected = index[i] assert result._repr_base == expected._repr_base assert result == expected @pytest.mark.parametrize("periods", [0, 9999, 10000, 10001]) def test_iteration_over_chunksize(self, periods): # GH21012 index = date_range("2000-01-01 00:00:00", periods=periods, freq="min") num = 0 for stamp in index: assert index[num] == stamp num += 1 assert num == len(index) def test_misc_coverage(self): rng = date_range("1/1/2000", periods=5) result = rng.groupby(rng.day) assert isinstance(list(result.values())[0][0], Timestamp) def test_string_index_series_name_converted(self): # #1644 df = DataFrame(np.random.randn(10, 4), index=date_range("1/1/2000", periods=10)) result = df.loc["1/3/2000"] assert result.name == df.index[2] result = df.T["1/3/2000"] assert result.name == df.index[2] def test_groupby_function_tuple_1677(self): df = DataFrame(np.random.rand(100), index=date_range("1/1/2000", periods=100)) monthly_group = df.groupby(lambda x: (x.year, x.month)) result = monthly_group.mean() assert isinstance(result.index[0], tuple) def test_isin(self): index = tm.makeDateIndex(4) result = index.isin(index) assert result.all() result = index.isin(list(index)) assert result.all() tm.assert_almost_equal( index.isin([index[2], 5]), np.array([False, False, True, False]) ) def assert_index_parameters(self, index): assert index.freq == "40960N" assert index.inferred_freq == "40960N" def test_ns_index(self): nsamples = 400 ns = int(1e9 / 24414) dtstart = np.datetime64("2012-09-20T00:00:00") dt = dtstart + np.arange(nsamples) * np.timedelta64(ns, "ns") freq = ns * offsets.Nano() index = DatetimeIndex(dt, freq=freq, name="time") self.assert_index_parameters(index) new_index = date_range(start=index[0], end=index[-1], freq=index.freq) self.assert_index_parameters(new_index) def test_asarray_tz_naive(self): # This shouldn't produce a warning. idx = date_range("2000", periods=2) # M8[ns] by default result = np.asarray(idx) expected = np.array(["2000-01-01", "2000-01-02"], dtype="M8[ns]") tm.assert_numpy_array_equal(result, expected) # optionally, object result = np.asarray(idx, dtype=object) expected = np.array([Timestamp("2000-01-01"), Timestamp("2000-01-02")]) tm.assert_numpy_array_equal(result, expected) def test_asarray_tz_aware(self): tz = "US/Central" idx = date_range("2000", periods=2, tz=tz) expected = np.array(["2000-01-01T06", "2000-01-02T06"], dtype="M8[ns]") result = np.asarray(idx, dtype="datetime64[ns]") tm.assert_numpy_array_equal(result, expected) # Old behavior with no warning result = np.asarray(idx, dtype="M8[ns]") tm.assert_numpy_array_equal(result, expected) # Future behavior with no warning expected = np.array( [Timestamp("2000-01-01", tz=tz), Timestamp("2000-01-02", tz=tz)] ) result = np.asarray(idx, dtype=object) tm.assert_numpy_array_equal(result, expected)