Add "unit"-parameter to date_range, enhance iso time parser to us (#9885

)

* Add "unit"-parameter to date_range, enhance iso time parser to microseconds
* add microsecond to test_parse_iso8601_like
* fix comment on microsecond
* add whats-new.rst entry
* add test for microsecond

doc/whats-new.rst

@@ -24,7 +24,9 @@ New Features
 - Better support wrapping additional array types (e.g. ``cupy`` or ``jax``) by calling generalized
   duck array operations throughout more xarray methods. (:issue:`7848`, :pull:`9798`).
   By `Sam Levang <https://github.com/slevang>`_.
-
+- Add ``unit`` - keyword argument to :py:func:`date_range` and ``microsecond`` parsing to
+  iso8601-parser (:pull:`9885`).
+  By `Kai Mühlbauer <https://github.com/kmuehlbauer>`_.
 
 Breaking changes
 ~~~~~~~~~~~~~~~~

xarray/coding/cftime_offsets.py

@@ -68,7 +68,12 @@
 from xarray.core.utils import attempt_import, emit_user_level_warning
 
 if TYPE_CHECKING:
-    from xarray.core.types import InclusiveOptions, Self, TypeAlias
+    from xarray.core.types import (
+        InclusiveOptions,
+        PDDatetimeUnitOptions,
+        Self,
+        TypeAlias,
+    )
 
 
 DayOption: TypeAlias = Literal["start", "end"]
@@ -971,7 +976,6 @@ def cftime_range(
         Include boundaries; whether to set each bound as closed or open.
 
         .. versionadded:: 2023.02.0
-
     calendar : str, default: "standard"
         Calendar type for the datetimes.
 
@@ -1180,6 +1184,7 @@ def date_range(
     normalize=False,
     name=None,
     inclusive: InclusiveOptions = "both",
+    unit: PDDatetimeUnitOptions = "ns",
     calendar="standard",
     use_cftime=None,
 ):
@@ -1210,6 +1215,10 @@ def date_range(
         Include boundaries; whether to set each bound as closed or open.
 
         .. versionadded:: 2023.02.0
+    unit : {"s", "ms", "us", "ns"}, default "ns"
+        Specify the desired resolution of the result.
+
+        .. versionadded:: 2024.12.0
     calendar : str, default: "standard"
         Calendar type for the datetimes.
     use_cftime : boolean, optional
@@ -1245,6 +1254,7 @@ def date_range(
                 normalize=normalize,
                 name=name,
                 inclusive=inclusive,
+                unit=unit,
             )
         except pd.errors.OutOfBoundsDatetime as err:
             if use_cftime is False:

xarray/coding/cftimeindex.py

@@ -92,14 +92,15 @@ def trailing_optional(xs):
     return xs[0] + optional(trailing_optional(xs[1:]))
 
 
-def build_pattern(date_sep=r"\-", datetime_sep=r"T", time_sep=r"\:"):
+def build_pattern(date_sep=r"\-", datetime_sep=r"T", time_sep=r"\:", micro_sep=r"."):
     pieces = [
         (None, "year", r"\d{4}"),
         (date_sep, "month", r"\d{2}"),
         (date_sep, "day", r"\d{2}"),
         (datetime_sep, "hour", r"\d{2}"),
         (time_sep, "minute", r"\d{2}"),
         (time_sep, "second", r"\d{2}"),
+        (micro_sep, "microsecond", r"\d{1,6}"),
     ]
     pattern_list = []
     for sep, name, sub_pattern in pieces:
@@ -131,11 +132,12 @@ def _parse_iso8601_with_reso(date_type, timestr):
     result = parse_iso8601_like(timestr)
     replace = {}
 
-    for attr in ["year", "month", "day", "hour", "minute", "second"]:
+    for attr in ["year", "month", "day", "hour", "minute", "second", "microsecond"]:
         value = result.get(attr, None)
         if value is not None:
-            # Note ISO8601 conventions allow for fractional seconds.
-            # TODO: Consider adding support for sub-second resolution?
+            if attr == "microsecond":
+                # convert match string into valid microsecond value
+                value = 10 ** (6 - len(value)) * int(value)
             replace[attr] = int(value)
             resolution = attr
     return default.replace(**replace), resolution

xarray/core/types.py

@@ -249,6 +249,7 @@ def copy(
     "Y", "M", "W", "D", "h", "m", "s", "ms", "us", "μs", "ns", "ps", "fs", "as", None
 ]
 NPDatetimeUnitOptions = Literal["D", "h", "m", "s", "ms", "us", "ns"]
+PDDatetimeUnitOptions = Literal["s", "ms", "us", "ns"]
 
 QueryEngineOptions = Literal["python", "numexpr", None]
 QueryParserOptions = Literal["pandas", "python"]

xarray/tests/test_cftimeindex.py

@@ -35,9 +35,23 @@
     standard_or_gregorian = "standard"
 
 
-def date_dict(year=None, month=None, day=None, hour=None, minute=None, second=None):
+def date_dict(
+    year=None,
+    month=None,
+    day=None,
+    hour=None,
+    minute=None,
+    second=None,
+    microsecond=None,
+):
     return dict(
-        year=year, month=month, day=day, hour=hour, minute=minute, second=second
+        year=year,
+        month=month,
+        day=day,
+        hour=hour,
+        minute=minute,
+        second=second,
+        microsecond=microsecond,
     )
 
 
@@ -86,6 +100,30 @@ def date_dict(year=None, month=None, day=None, hour=None, minute=None, second=No
             year="1999", month="01", day="01", hour="12", minute="34", second="56"
         ),
     ),
+    "microsecond-1": (
+        "19990101T123456.123456",
+        date_dict(
+            year="1999",
+            month="01",
+            day="01",
+            hour="12",
+            minute="34",
+            second="56",
+            microsecond="123456",
+        ),
+    ),
+    "microsecond-2": (
+        "19990101T123456.1",
+        date_dict(
+            year="1999",
+            month="01",
+            day="01",
+            hour="12",
+            minute="34",
+            second="56",
+            microsecond="1",
+        ),
+    ),
 }
 
 
@@ -98,9 +136,12 @@ def test_parse_iso8601_like(string, expected):
     result = parse_iso8601_like(string)
     assert result == expected
 
-    with pytest.raises(ValueError):
-        parse_iso8601_like(string + "3")
-        parse_iso8601_like(string + ".3")
+    if result["microsecond"] is None:
+        with pytest.raises(ValueError):
+            parse_iso8601_like(string + "3")
+    if result["second"] is None:
+        with pytest.raises(ValueError):
+            parse_iso8601_like(string + ".3")
 
 
 _CFTIME_CALENDARS = [
@@ -301,6 +342,7 @@ def test_cftimeindex_days_in_month_accessor(index):
         ("19990202T01", (1999, 2, 2, 1), "hour"),
         ("19990202T0101", (1999, 2, 2, 1, 1), "minute"),
         ("19990202T010156", (1999, 2, 2, 1, 1, 56), "second"),
+        ("19990202T010156.123456", (1999, 2, 2, 1, 1, 56, 123456), "microsecond"),
     ],
 )
 def test_parse_iso8601_with_reso(date_type, string, date_args, reso):