Merge remote-tracking branch 'origin/develop' into chore/multiple-fix…

…es-ci-pre-commit

pyproject.toml

@@ -41,8 +41,9 @@ classifiers = [
 
 dynamic = [ "version" ]
 dependencies = [
+  "aniso8601",
   "pyyaml",
-  "tomli",  # Only needed before 3.11
+  "tomli",     # Only needed before 3.11
   "tqdm",
 ]
 

src/anemoi/utils/dates.py

@@ -8,11 +8,12 @@
 
 import calendar
 import datetime
+import re
 
-from .hindcasts import HindcastDatesTimes
+import aniso8601
 
 
-def normalise_frequency(frequency) -> int:
+def normalise_frequency(frequency):
     if isinstance(frequency, int):
         return frequency
     assert isinstance(frequency, str), (type(frequency), frequency)
@@ -22,7 +23,7 @@ def normalise_frequency(frequency) -> int:
     return {"h": v, "d": v * 24}[unit]
 
 
-def no_time_zone(date) -> datetime.datetime:
+def _no_time_zone(date) -> datetime.datetime:
     """Remove time zone information from a date.
 
     Parameters
@@ -40,32 +41,200 @@ def no_time_zone(date) -> datetime.datetime:
 
 
 # this function is use in anemoi-datasets
-def as_datetime(date) -> datetime.datetime:
+def as_datetime(date, keep_time_zone=False) -> datetime.datetime:
     """Convert a date to a datetime object, removing any time zone information.
 
     Parameters
     ----------
     date : datetime.date or datetime.datetime or str
         The date to convert.
+    keep_time_zone : bool, optional
+        If True, the time zone information is kept, by default False.
 
     Returns
     -------
     datetime.datetime
         The datetime object.
     """
 
+    tidy = _no_time_zone if not keep_time_zone else lambda x: x
+
     if isinstance(date, datetime.datetime):
-        return no_time_zone(date)
+        return tidy(date)
 
     if isinstance(date, datetime.date):
-        return no_time_zone(datetime.datetime(date.year, date.month, date.day))
+        return tidy(datetime.datetime(date.year, date.month, date.day))
 
     if isinstance(date, str):
-        return no_time_zone(datetime.datetime.fromisoformat(date))
+        return tidy(datetime.datetime.fromisoformat(date))
 
     raise ValueError(f"Invalid date type: {type(date)}")
 
 
+def _as_datetime_list(date, default_increment):
+    if isinstance(date, (list, tuple)):
+        for d in date:
+            yield from _as_datetime_list(d, default_increment)
+
+    if isinstance(date, str):
+        # Check for ISO format
+        try:
+            start, end = aniso8601.parse_interval(date)
+            while start <= end:
+                yield as_datetime(start)
+                start += default_increment
+
+            return
+
+        except aniso8601.exceptions.ISOFormatError:
+            pass
+
+        try:
+            intervals = aniso8601.parse_repeating_interval(date)
+            for date in intervals:
+                yield as_datetime(date)
+            return
+        except aniso8601.exceptions.ISOFormatError:
+            pass
+
+    yield as_datetime(date)
+
+
+def as_datetime_list(date, default_increment=1):
+    default_increment = frequency_to_timedelta(default_increment)
+    return list(_as_datetime_list(date, default_increment))
+
+
+def frequency_to_timedelta(frequency) -> datetime.timedelta:
+    """Convert a frequency to a timedelta object.
+
+    Parameters
+    ----------
+    frequency : int or str or datetime.timedelta
+        The frequency to convert. If an integer, it is assumed to be in hours. If a string, it can be in the format:
+
+        - "1h" for 1 hour
+        - "1d" for 1 day
+        - "1m" for 1 minute
+        - "1s" for 1 second
+        - "1:30" for 1 hour and 30 minutes
+        - "1:30:10" for 1 hour, 30 minutes and 10 seconds
+        - "PT10M" for 10 minutes (ISO8601)
+
+        If a timedelta object is provided, it is returned as is.
+
+    Returns
+    -------
+    datetime.timedelta
+        The timedelta object.
+
+    Raises
+    ------
+    ValueError
+        Exception raised if the frequency cannot be converted to a timedelta.
+    """
+
+    if isinstance(frequency, datetime.timedelta):
+        return frequency
+
+    if isinstance(frequency, int):
+        return datetime.timedelta(hours=frequency)
+
+    assert isinstance(frequency, str), (type(frequency), frequency)
+
+    try:
+        return frequency_to_timedelta(int(frequency))
+    except ValueError:
+        pass
+
+    if re.match(r"^\d+[hdms]$", frequency, re.IGNORECASE):
+        unit = frequency[-1].lower()
+        v = int(frequency[:-1])
+        unit = {"h": "hours", "d": "days", "s": "seconds", "m": "minutes"}[unit]
+        return datetime.timedelta(**{unit: v})
+
+    m = frequency.split(":")
+    if len(m) == 2:
+        return datetime.timedelta(hours=int(m[0]), minutes=int(m[1]))
+
+    if len(m) == 3:
+        return datetime.timedelta(hours=int(m[0]), minutes=int(m[1]), seconds=int(m[2]))
+
+    # ISO8601
+    try:
+        return aniso8601.parse_duration(frequency)
+    except aniso8601.exceptions.ISOFormatError:
+        pass
+
+    raise ValueError(f"Cannot convert frequency {frequency} to timedelta")
+
+
+def frequency_to_string(frequency) -> str:
+    """Convert a frequency (i.e. a datetime.timedelta) to a string.
+
+    Parameters
+    ----------
+    frequency : datetime.timedelta
+        The frequency to convert.
+
+    Returns
+    -------
+    str
+        A string representation of the frequency.
+    """
+
+    frequency = frequency_to_timedelta(frequency)
+
+    total_seconds = frequency.total_seconds()
+    assert int(total_seconds) == total_seconds, total_seconds
+    total_seconds = int(total_seconds)
+
+    seconds = total_seconds
+
+    days = seconds // (24 * 3600)
+    seconds %= 24 * 3600
+    hours = seconds // 3600
+    seconds %= 3600
+    minutes = seconds // 60
+    seconds %= 60
+
+    if days > 0 and hours == 0 and minutes == 0 and seconds == 0:
+        return f"{days}d"
+
+    if days == 0 and hours > 0 and minutes == 0 and seconds == 0:
+        return f"{hours}h"
+
+    if days == 0 and hours == 0 and minutes > 0 and seconds == 0:
+        return f"{minutes}m"
+
+    if days == 0 and hours == 0 and minutes == 0 and seconds > 0:
+        return f"{seconds}s"
+
+    if days > 0:
+        return f"{total_seconds}s"
+
+    return str(frequency)
+
+
+def frequency_to_seconds(frequency) -> int:
+    """Convert a frequency to seconds.
+
+    Parameters
+    ----------
+    frequency : _type_
+        _description_
+
+    Returns
+    -------
+    int
+        Number of seconds.
+    """
+
+    result = frequency_to_timedelta(frequency).total_seconds()
+    assert int(result) == result, result
+    return int(result)
+
+
 DOW = {
     "monday": 0,
     "tuesday": 1,
@@ -142,7 +311,7 @@ def __init__(self, start, end, increment=24, *, day_of_month=None, day_of_week=N
         """
         self.start = as_datetime(start)
         self.end = as_datetime(end)
-        self.increment = datetime.timedelta(hours=increment)
+        self.increment = frequency_to_timedelta(increment)
         self.day_of_month = _make_day(day_of_month)
         self.day_of_week = _make_week(day_of_week)
         self.calendar_months = _make_months(calendar_months)
@@ -270,15 +439,16 @@ def datetimes_factory(*args, **kwargs):
         name = kwargs.get("name")
 
         if name == "hindcast":
+            from .hindcasts import HindcastDatesTimes
+
             reference_dates = kwargs["reference_dates"]
             reference_dates = datetimes_factory(reference_dates)
             years = kwargs["years"]
             return HindcastDatesTimes(reference_dates=reference_dates, years=years)
 
         kwargs = kwargs.copy()
         if "frequency" in kwargs:
-            freq = kwargs.pop("frequency")
-            kwargs["increment"] = normalise_frequency(freq)
+            kwargs["increment"] = kwargs.pop("frequency")
         return DateTimes(**kwargs)
 
     if not any((isinstance(x, dict) or isinstance(x, list)) for x in args):
@@ -294,3 +464,8 @@ def datetimes_factory(*args, **kwargs):
             return datetimes_factory(*a)
 
     return ConcatDateTimes(*[datetimes_factory(a) for a in args])
+
+
+if __name__ == "__main__":
+    print(as_datetime_list("R10/2023-01-01T00:00:00Z/P1D"))
+    print(as_datetime_list("2007-03-01T13:00:00/2008-05-11T15:30:00", "200h"))

src/anemoi/utils/humanize.py

@@ -15,6 +15,8 @@
 import warnings
 from collections import defaultdict
 
+from anemoi.utils.dates import as_datetime
+
 
 def bytes_to_human(n: float) -> str:
     """Convert a number of bytes to a human readable string
@@ -629,3 +631,60 @@ def shorten_list(lst, max_length=5) -> list:
         if isinstance(lst, tuple):
             return tuple(result)
         return result
+
+
+def _compress_dates(dates):
+    dates = sorted(dates)
+    if len(dates) < 3:
+        yield dates
+        return
+
+    prev = first = dates.pop(0)
+    curr = dates.pop(0)
+    delta = curr - prev
+    while curr - prev == delta:
+        prev = curr
+        if not dates:
+            break
+        curr = dates.pop(0)
+
+    yield (first, prev, delta)
+    if dates:
+        yield from _compress_dates([curr] + dates)
+
+
+def compress_dates(dates) -> str:
+    """Compress a list of dates into a human-readable format.
+
+    Parameters
+    ----------
+    dates : list
+        A list of dates, as datetime objects or strings.
+
+    Returns
+    -------
+    str
+        A human-readable string representing the compressed dates.
+    """
+
+    dates = [as_datetime(_) for _ in dates]
+    result = []
+
+    for n in _compress_dates(dates):
+        if isinstance(n, list):
+            result.extend([str(_) for _ in n])
+        else:
+            result.append(" ".join([str(n[0]), "to", str(n[1]), "by", str(n[2])]))
+
+    return result
+
+
+def print_dates(dates) -> None:
+    """Print a list of dates in a human-readable format.
+
+    Parameters
+    ----------
+    dates : list
+        A list of dates, as datetime objects or strings.
+    """
+    print(compress_dates(dates))

tests/test_frequency.py

@@ -0,0 +1,37 @@
+# (C) Copyright 2023 European Centre for Medium-Range Weather Forecasts.
+# This software is licensed under the terms of the Apache Licence Version 2.0
+# which can be obtained at http://www.apache.org/licenses/LICENSE-2.0.
+# In applying this licence, ECMWF does not waive the privileges and immunities
+# granted to it by virtue of its status as an intergovernmental organisation
+# nor does it submit to any jurisdiction.
+
+import datetime
+
+from anemoi.utils.dates import frequency_to_string
+from anemoi.utils.dates import frequency_to_timedelta
+
+
+def test_frequency_to_string():
+    assert frequency_to_string(datetime.timedelta(hours=1)) == "1h"
+    assert frequency_to_string(datetime.timedelta(hours=1, minutes=30)) == "1:30:00"
+    assert frequency_to_string(datetime.timedelta(days=10)) == "10d"
+    assert frequency_to_string(datetime.timedelta(minutes=10)) == "10m"
+    assert frequency_to_string(datetime.timedelta(minutes=90)) == "1:30:00"
+
+
+def test_frequency_to_timedelta():
+    assert frequency_to_timedelta("1s") == datetime.timedelta(seconds=1)
+    assert frequency_to_timedelta("3m") == datetime.timedelta(minutes=3)
+    assert frequency_to_timedelta("1h") == datetime.timedelta(hours=1)
+    assert frequency_to_timedelta("3d") == datetime.timedelta(days=3)
+    assert frequency_to_timedelta("90m") == datetime.timedelta(hours=1, minutes=30)
+    assert frequency_to_timedelta("0:30") == datetime.timedelta(minutes=30)
+    assert frequency_to_timedelta("0:30:10") == datetime.timedelta(minutes=30, seconds=10)
+    assert frequency_to_timedelta("1:30:10") == datetime.timedelta(hours=1, minutes=30, seconds=10)
+
+    assert frequency_to_timedelta("PT10M") == datetime.timedelta(minutes=10)
+
+
+if __name__ == "__main__":
+    test_frequency_to_string()
+    test_frequency_to_timedelta()