def _normalize_datetime(
val: Union[pd.Timestamp, str, np.datetime64, datetime.datetime,
datetime.date]
) -> Tuple[np.datetime64, np.timedelta64]:
'''
Break up a datetime into numpy date and numpy timedelta.
Args:
val: The datetime to normalize. Can be an array or a single datetime as a string, pandas timestamp, numpy datetime
or python date or datetime
>>> print(_normalize_datetime(pd.Timestamp('2016-05-01 3:55:00')))
(numpy.datetime64('2016-05-01'), numpy.timedelta64(14100000000000,'ns'))
>>> print(_normalize_datetime('2016-05-01'))
(numpy.datetime64('2016-05-01'), numpy.timedelta64(0,'D'))
>>> x = pd.DataFrame({'x' : [np.datetime64('2015-01-01 05:00:00'), np.datetime64('2015-02-01 06:00:00')]})
>>> print(_normalize_datetime(x.x))
(array(['2015-01-01', '2015-02-01'], dtype='datetime64[D]'), array([18000000000000, 21600000000000], dtype='timedelta64[ns]'))
>>> x = pd.DataFrame({'x' : [1, 2]}, index = [np.datetime64('2015-01-01 05:00:00'), np.datetime64('2015-02-01 06:00:00')])
>>> print(_normalize_datetime(x.index))
(array(['2015-01-01', '2015-02-01'], dtype='datetime64[D]'), array([18000000000000, 21600000000000], dtype='timedelta64[ns]'))
'''
if isinstance(val, pd.Timestamp):
datetime = val.to_datetime64()
elif isinstance(val, pd.Series) or isinstance(val, pd.DatetimeIndex):
datetime = val.values
elif isinstance(val, np.ndarray) and np.issubdtype(val.dtype,
np.datetime64):
datetime = val
else:
datetime = np.datetime64(val)
date = datetime.astype('M8[D]')
time_delta = datetime - date
return date, time_delta
def merra2_stare_time(ds, iTime=None, tType=2, centered=True):
if centered:
start_time_mn = ds['time'].begin_time / 100
start_time_sec = ds['time'].begin_time % 100
resolution = stare_temporal_resolutions[tType]['1/2hr']
else:
start_time_mn = 0
start_time_sec = 0
resolution = stare_temporal_resolutions[tType]['1hr']
yr, mo, dy = merra2_parse_begin_date(ds['time'].begin_date)
tm = []
if iTime is None:
i0 = 0
i1 = 24
else:
i0 = iTime
i1 = iTime + 1
for i in range(i0, i1):
# hr,mn = merra2_make_time(ds['time'].begin_date,ds['time'][i])
hr, mn = merra2_make_time(start_time_mn, ds['time'][i])
sc = start_time_sec
tm.append(format_time(yr, mo, dy, hr, mn, sc))
dt = np.array(tm, dtype='datetime64[ms]')
idx = ps.from_utc(dt.astype(np.int64), resolution)
return idx
def normalize_datetime(val):
'''
Break up a datetime into date and time.
Args:
val: The datetime to normalize. Can be an array or a single datetime as a string, pandas timestamp, numpy datetime
or python date or datetime
Return:
tuple of numpy datetime64('D') and np.timedelta64
File "__main__", line 55, in __main__._normalize_date_time
>>> print(normalize_datetime(pd.Timestamp('2016-05-01 3:55:00')))
(numpy.datetime64('2016-05-01'), numpy.timedelta64(14100000000000,'ns'))
>>> print(normalize_datetime('2016-05-01'))
(numpy.datetime64('2016-05-01'), numpy.timedelta64(0,'D'))
>>> x = pd.DataFrame({'x' : [np.datetime64('2015-01-01 05:00:00'), np.datetime64('2015-02-01 06:00:00')]})
>>> print(normalize_datetime(x.x))
(array(['2015-01-01', '2015-02-01'], dtype='datetime64[D]'), array([18000000000000, 21600000000000], dtype='timedelta64[ns]'))
>>> x = pd.DataFrame({'x' : [1, 2]}, index = [np.datetime64('2015-01-01 05:00:00'), np.datetime64('2015-02-01 06:00:00')])
>>> print(normalize_datetime(x.index))
(array(['2015-01-01', '2015-02-01'], dtype='datetime64[D]'), array([18000000000000, 21600000000000], dtype='timedelta64[ns]'))
'''
if isinstance(val, pd.Timestamp):
datetime = val.to_datetime64()
elif isinstance(val, pd.Series) or isinstance(val, pd.DatetimeIndex):
datetime = val.values
else:
datetime = np.datetime64(val)
date = datetime.astype('M8[D]')
time_delta = datetime - date
return date, time_delta
def read(self, dataset, filename, timeshift=0, **kargs):
fh = codecs.open(filename, encoding='utf-8-sig', errors='ignore')
for line in fh.readlines():
a = line.encode('utf-8')
if a == b'\x00'*len(a):
fh.close()
msg = "File {} contains line of binary 0's"
raise MiniDoasException(msg.format(filename))
fh.seek(0)
dt = np.dtype([('station', 'S2'), ('date', 'S10'), ('time', np.float),
('stept', np.int), ('angle', np.float),
('intt', np.int), ('nspec', np.int),
('specin', np.float), ('counts', np.int, (482,))])
def date_converter(x):
y = x.decode('ascii')
return '%s-%s-%s' % (y[0:4], y[4:6], y[6:8])
data = np.loadtxt(fh, converters={1: date_converter},
dtype=dt, delimiter=',', ndmin=1)
fh.close()
# Construct datetimes
date = data['date'].astype('datetime64')
hours = (data['time']/3600.).astype(int)
minutes = ((data['time'] - hours*3600.)/60.).astype(int)
fseconds = data['time'] - hours*3600. - minutes*60.
iseconds = (fseconds).astype(int)
mseconds = np.round((fseconds - iseconds), 3)*1e3
datetime = date + hours.astype('timedelta64[h]') \
+ minutes.astype('timedelta64[m]') \
+ iseconds.astype('timedelta64[s]') \
+ mseconds.astype('timedelta64[ms]')
datetime -= np.timedelta64(int(timeshift), 'h')
# Convert radians to decimal degrees
angles = data['angle']*360./(2.*np.pi)
rdtb = RawDataTypeBuffer(d_var_unit='ppm-m',
ind_var_unit='nm',
name='measurement',
acquisition='stationary')
wavelengths = np.arange(30, 512)
rb = RawDataBuffer(inc_angle=angles,
datetime=datetime.astype(str),
ind_var=wavelengths,
d_var=data['counts'],
integration_time=data['intt'])
return {str(rb): rb, str(rdtb): rdtb}
def goes10_img_stare_time(ds,tType=2,centered=True): resolution = stare_temporal_resolutions[2]['1/4hr'] dt = np.array(ds['time'][0]*1000,dtype='datetime64[ms]').reshape([1]) return ps.from_utc(dt.astype(np.int64),resolution)