def group_attrib(self):
'''
return a namedtuple containing all attributes attached
to groups of which the given series is a member
for each group of which the series is a member
'''
group_attributes = [g.attrib for g in self.dataset.groups if self in g]
if group_attributes:
return concat_namedtuples(*group_attributes)
def test_concat_namedtuples():
num = list(range(26))
chars = [chr(65 + i) for i in num]
limits = [0, 4, 5, 8, 14, 22, 25]
tuples = []
for i in range(len(limits) - 1):
newtype = namedtuple_factory('Test', chars[limits[i]:limits[i + 1]])
t = newtype(*num[limits[i]:limits[i + 1]])
tuples.append(t)
concat1 = concat_namedtuples(*tuples)
assert isinstance(concat1, tuple)
assert concat1.A == 0
def iter_pd_series(self, iter_series, dim_at_obs, dtype,
attributes, reverse_obs, fromfreq, parse_time):
with_obs_attr = 'o' in attributes
for series in iter_series:
# Generate the 3 main columns: index, values and attributes
obs_zip = list(zip(*series.obs(with_values=dtype,
with_attributes=with_obs_attr, reverse_obs=reverse_obs)))
# Are there observations at all?
if obs_zip:
obs_dim = obs_zip[0]
obs_values = NP.array(obs_zip[1], dtype=dtype)
obs_attrib = obs_zip[2]
l = len(obs_dim)
# Generate the index
# Get frequency if present
if 'FREQ' in series.key._fields:
f = series.key.FREQ
elif series.attrib and 'FREQUENCY' in series.attrib._fields:
f = series.attrib.FREQUENCY
elif 'FREQUENCY' in series.key._fields:
f = series.key.FREQUENCY
elif series.attrib and 'FREQ' in series.attrib._fields:
f = series.attrib.FREQ
else:
f = None
if parse_time and dim_at_obs == 'TIME_PERIOD':
# First, handle half-yearly and bimonthly freqs
# and format such as '2010-S1' format dim
# pandas cannot parse those. So convert them
if f == 'H':
f = '2Q'
# patch the dim values
obs_dim = ['Q'.join((od[:-2], '1' if od[-1] == '1' else '3'))
for od in obs_dim]
# Check if we can build the index based on start and freq
# Constructing the index from the first value and FREQ should only
# occur if 'fromfreq' and hence f is True
if fromfreq and f: # So there is a freq and we must use it
series_index = PD.period_range(start=PD.Period(obs_dim[0], freq=f), periods=l,
freq=f, name=dim_at_obs)
else:
# There is no ffreq or we must not use it.
# So generate the index from all the obs dim values
series_index = PD.PeriodIndex(
(PD.Period(d, freq=f) for d in obs_dim), name=dim_at_obs)
elif parse_time and dim_at_obs == 'TIME':
if fromfreq and f:
series_index = PD.date_range(
start=PD.datetime(obs_dim[0]), periods=l, freq=f, name=dim_at_obs)
else:
series_index = PD.DatetimeIndex(
(PD.datetime(d) for d in obs_dim),
name=dim_at_obs)
else:
# Not a datetime or period index or don't parse it
series_index = PD.Index(obs_dim, name=dim_at_obs)
if dtype:
value_series = PD.Series(
obs_values, index=series_index, name=series.key)
if attributes:
# Assemble attributes of dataset, group and series if
# needed
gen_attrib = [attr
for flag, attr in (('s', series.attrib),
('g', series.group_attrib), ('d', series.dataset.attrib))
if (flag in attributes) and attr]
if gen_attrib:
gen_attrib = concat_namedtuples(*gen_attrib)
else:
gen_attrib = None
if 'o' in attributes:
# concat with general attributes if any
if gen_attrib:
attrib_iter = (concat_namedtuples(a, gen_attrib,
name='Attrib') for a in obs_attrib)
else:
# Simply take the obs attributes
attrib_iter = obs_attrib
else:
# Make iterator yielding the constant general attribute set
# It may be None.
# for each obs
attrib_iter = (gen_attrib for d in obs_attrib)
attrib_series = PD.Series(attrib_iter,
index=series_index, dtype='object', name=series.key)
else:
# There are no observations. So generate empty DataFrames
if dtype:
value_series = PD.Series(name=series.key)
if attributes:
attrib_series = PD.Series(name=series.key)
# decide what to yield
if dtype and attributes:
yield value_series, attrib_series
elif dtype:
yield value_series
elif attributes:
yield attrib_series
else:
raise ValueError(
"At least one of 'dtype' or 'attributes' args must be True.")
def iter_pd_series(self, iter_series, dim_at_obs, dtype, attributes, reverse_obs, fromfreq, parse_time):
for series in iter_series:
# Generate the 3 main columns: index, values and attributes
obs_zip = iter(zip(*series.obs(dtype, attributes, reverse_obs)))
obs_dim = next(obs_zip)
l = len(obs_dim)
obs_values = NP.array(next(obs_zip), dtype=dtype)
if attributes:
obs_attrib = next(obs_zip)
# Generate the index
if parse_time and dim_at_obs == "TIME_PERIOD":
# Check if we can build the index based on start and freq
# Constructing the index from the first value and FREQ should only
# occur if 'fromfreq' is True
# and there is a FREQ dimension at all.
if fromfreq and "FREQ" in series.key._fields:
f = series.key.FREQ
od0 = obs_dim[0]
year, subdiv = map(int, (od0[:4], od0[-1]))
if f == "Q":
start_date = PD.datetime(year, (subdiv - 1) * 3 + 1, 1)
series_index = PD.period_range(start=start_date, periods=l, freq="Q", name=dim_at_obs)
elif "S" in od0:
# pandas cannot represent semesters as periods. So we
# use date_range.
start_date = PD.datetime(year, (subdiv - 1) * 6 + 1, 1)
series_index = PD.date_range(start=start_date, periods=l, freq="6M", name=dim_at_obs)
else:
series_index = PD.period_range(start=od0, periods=l, freq=f, name=dim_at_obs)
elif "FREQ" in series.key._fields:
# fromfreq is False. So generate the index from all the
# strings
f = series.key.FREQ
# Generate arrays for years and subdivisions (quarters or
# semesters
if f == "Q":
series_index = PD.Index(
(PD.Period(year=int(d[:4]), quarter=int(d[-1]), freq="Q") for d in obs_dim), name=dim_at_obs
)
elif f == "H":
series_index = PD.Index(
(PD.datetime(int(d[:4]), (int(d[-1]) - 1) * 6 + 1, 1) for d in obs_dim), name=dim_at_obs
)
else: # other freq such as 'A' or 'M'
series_index = PD.PeriodIndex(obs_dim, freq=f, name=dim_at_obs)
elif parse_time and dim_at_obs == "TIME":
if fromfreq and "FREQ" in series.key._fields:
f = series.key.FREQ
series_index = PD.date_range(start=obs_dim[0], periods=l, freq=f, name=dim_at_obs)
else:
series_index = PD.DatetimeIndex(obs_dim, name=dim_at_obs)
# Not a datetime or period index or don't parse it
else:
series_index = PD.Index(obs_dim, name=dim_at_obs)
if dtype:
value_series = PD.Series(obs_values, index=series_index, name=series.key)
if attributes:
# Assemble attributes of dataset, group and series if needed
gen_attrib = [
attr
for flag, attr in (("s", series.attrib), ("g", series.group_attrib), ("d", series.dataset.attrib))
if (flag in attributes) and attr
]
if gen_attrib:
gen_attrib = concat_namedtuples(*gen_attrib)
else:
gen_attrib = None
if "o" in attributes:
# concat with general attributes if any
if gen_attrib:
attrib_iter = (concat_namedtuples(a, gen_attrib, name="Attrib") for a in obs_attrib)
else:
# Simply take the obs attributes
attrib_iter = obs_attrib
else:
# Make iterator yielding the constant general attribute set
# It may be None.
# for each obs
attrib_iter = (gen_attrib for d in obs_attrib)
attrib_series = PD.Series(attrib_iter, index=series_index, dtype="object", name=series.key)
# decide what to yield
if dtype and attributes:
yield value_series, attrib_series
elif dtype:
yield value_series
elif attributes:
yield attrib_series
else:
raise ValueError("At least one of 'dtype' or 'attributes' args must be True.")
def iter_pd_series(self, iter_series, dim_at_obs, dtype, attributes,
reverse_obs, fromfreq, parse_time):
for series in iter_series:
# Generate the 3 main columns: index, values and attributes
obs_zip = iter(zip(*series.obs(dtype, attributes, reverse_obs)))
obs_dim = next(obs_zip)
l = len(obs_dim)
obs_values = NP.array(next(obs_zip), dtype=dtype)
if attributes:
obs_attrib = next(obs_zip)
# Generate the index
if parse_time and dim_at_obs == 'TIME_PERIOD':
# Check if we can build the index based on start and freq
# Constructing the index from the first value and FREQ should only
# occur if 'fromfreq' is True
# and there is a FREQ dimension at all.
if fromfreq and 'FREQ' in series.key._fields:
f = series.key.FREQ
od0 = obs_dim[0]
year, subdiv = map(int, (od0[:4], od0[-1]))
if f == 'Q':
start_date = PD.datetime(year, (subdiv - 1) * 3 + 1, 1)
series_index = PD.period_range(start=start_date,
periods=l,
freq='Q',
name=dim_at_obs)
elif 'S' in od0:
# pandas cannot represent semesters as periods. So we
# use date_range.
start_date = PD.datetime(year, (subdiv - 1) * 6 + 1, 1)
series_index = PD.date_range(start=start_date,
periods=l,
freq='6M',
name=dim_at_obs)
else:
series_index = PD.period_range(start=od0,
periods=l,
freq=f,
name=dim_at_obs)
elif 'FREQ' in series.key._fields:
# fromfreq is False. So generate the index from all the
# strings
f = series.key.FREQ
# Generate arrays for years and subdivisions (quarters or
# semesters
if f == 'Q':
series_index = PD.Index((PD.Period(
year=int(d[:4]), quarter=int(d[-1]), freq='Q')
for d in obs_dim),
name=dim_at_obs)
elif f == 'H':
series_index = PD.Index(
(PD.datetime(int(d[:4]),
(int(d[-1]) - 1) * 6 + 1, 1)
for d in obs_dim),
name=dim_at_obs)
else: # other freq such as 'A' or 'M'
series_index = PD.PeriodIndex(obs_dim,
freq=f,
name=dim_at_obs)
elif parse_time and dim_at_obs == 'TIME':
if fromfreq and 'FREQ' in series.key._fields:
f = series.key.FREQ
series_index = PD.date_range(start=obs_dim[0],
periods=l,
freq=f,
name=dim_at_obs)
else:
series_index = PD.DatetimeIndex(obs_dim, name=dim_at_obs)
# Not a datetime or period index or don't parse it
else:
series_index = PD.Index(obs_dim, name=dim_at_obs)
if dtype:
value_series = PD.Series(obs_values,
index=series_index,
name=series.key)
if attributes:
# Assemble attributes of dataset, group and series if needed
gen_attrib = [
attr for flag, attr in (('s', series.attrib),
('g', series.group_attrib),
('d', series.dataset.attrib))
if (flag in attributes) and attr
]
if gen_attrib:
gen_attrib = concat_namedtuples(*gen_attrib)
else:
gen_attrib = None
if 'o' in attributes:
# concat with general attributes if any
if gen_attrib:
attrib_iter = (concat_namedtuples(a,
gen_attrib,
name='Attrib')
for a in obs_attrib)
else:
# Simply take the obs attributes
attrib_iter = obs_attrib
else:
# Make iterator yielding the constant general attribute set
# It may be None.
# for each obs
attrib_iter = (gen_attrib for d in obs_attrib)
attrib_series = PD.Series(attrib_iter,
index=series_index,
dtype='object',
name=series.key)
# decide what to yield
if dtype and attributes:
yield value_series, attrib_series
elif dtype:
yield value_series
elif attributes:
yield attrib_series
else:
raise ValueError(
"At least one of 'dtype' or 'attributes' args must be True."
)
