def write(filename, d):
d2 = ds.copy(d)
for var in ds.get_vars(d2):
if isinstance(d2[var], np.ndarray):
d2[var] = d2[var].tolist()
with open(filename, 'w') as f:
json.dump(d2, f)
def read(filename, variables=None, sel=None, full=False, jd=False):
with open(filename) as f:
d = json.load(f)
for var in ds.get_vars(d):
if type(d[var]) is list:
d[var] = np.array(d[var])
return d
def read(filename, variables=None, sel=None, full=False, jd=False):
if type(filename) is bytes and str != bytes:
filename = os.fsdecode(filename)
with Dataset(filename, 'r') as f:
d = {}
d['.'] = {}
d['.']['.'] = read_attrs(f)
for name in f.variables.keys():
if variables is not None and name not in variables:
if full:
_, d['.'][name] = read_var(f, name, sel, False)
else:
d[name], d['.'][name] = read_var(f, name, sel)
if jd:
for name in ds.get_vars(d):
process_datetime_var(d, name)
return d
def output_sample(d, tres, output_sampling):
t = d['time_bnds'][0, 0]
r = (t + 0.5) % output_sampling
t1 = t - r
t2 = t1 + output_sampling
dims = ds.get_dims(d)
n = dims['time']
n2 = int(output_sampling / tres)
time = d['time']
time_bnds = d['time_bnds']
time_half2 = np.linspace(t1, t2, n2 + 1)
time2 = 0.5 * (time_half2[1:] + time_half2[:-1])
for var in ds.get_vars(d):
if 'time' not in d['.'][var]['.dims']:
continue
i = d['.'][var]['.dims'].index('time')
x = d[var]
size = x.shape
size2 = list(size)
size2[i] = n2
x2 = np.full(size2, np.nan, dtype=x.dtype)
for j in range(n2):
mask = np.maximum(
0,
np.minimum(time_bnds[:, 1], time_half2[j + 1]) -
np.maximum(time_bnds[:, 0], time_half2[j]))
s = np.sum(mask)
if s > 0.:
mask /= np.sum(mask)
sel2 = tuple(
[slice(None) if l != i else j for l in range(len(size))])
for k in np.argwhere(mask > 0):
sel = tuple(
[slice(None) if l != i else k for l in range(len(size))])
x2[sel2] = np.where(np.isnan(x2[sel2])*mask[k], 0, x2[sel2]) + \
x[sel]*mask[k]
d[var] = x2
d['time'] = time2
d['time_bnds'] = np.full((n2, 2), np.nan, np.float64)
d['time_bnds'][:, 0] = time_half2[:-1]
d['time_bnds'][:, 1] = time_half2[1:]
def output_sample(d, tres, output_sampling):
t1 = d['time'][0] - ((d['time'][0] + 0.5) % 1.0)
t2 = t1 + output_sampling
dims = ds.get_dims(d)
n = dims['time']
n2 = int(output_sampling/tres)
time = d['time']
time_half = np.zeros(n + 1, dtype=np.float64)
time_half[1:-1] = 0.5*(time[1:] + time[:-1])
time_half[0] = time[0] - 0.5*(time[1] - time[0])
time_half[-1] = time[-1] + 0.5*(time[-1] - time[-2])
time_half2 = np.linspace(t1, t2, n2 + 1)
time2 = 0.5*(time_half2[1:] + time_half2[:-1])
for var in ds.get_vars(d):
if 'time' not in d['.'][var]['.dims']:
continue
i = d['.'][var]['.dims'].index('time')
x = d[var]
size = x.shape
size2 = list(size)
size2[i] = n2
x2 = np.full(size2, np.nan, dtype=x.dtype)
for j in range(n2):
mask = np.maximum(0,
np.minimum(time_half[1:], time_half2[j + 1]) -
np.maximum(time_half[:-1], time_half2[j])
)
s = np.sum(mask)
if s > 0.:
mask /= np.sum(mask)
sel2 = tuple([
slice(None) if l != i else j
for l in range(len(size))
])
for k in range(len(mask)):
sel = tuple([
slice(None) if l != i else k
for l in range(len(size))
])
x2[sel2] = np.where(np.isnan(x2[sel2])*mask[k], 0, x2[sel2]) + \
x[sel]*mask[k]
d[var] = x2
d['time'] = time2
def write(filename, d):
ds.validate(d)
if type(filename) is bytes and hasattr(os, 'fsdecode'):
filename = os.fsdecode(filename)
with Dataset(filename, 'w') as f:
dims = ds.get_dims(d)
for k, v in dims.items():
f.createDimension(k, v)
for name in ds.get_vars(d):
data = ds.get_var(d, name)
if data.dtype == 'O' and \
len(data.flatten()) > 0 and \
type(data.flatten()[0]) is str:
dtype = str
else:
dtype = data.dtype
v = f.createVariable(name, dtype, ds.get_dims(d, name))
v.setncatts(ds.get_attrs(d, name))
v[::] = data
f.setncatts(ds.get_attrs(d))
def tsample(d, tres):
w = d['time_bnds'][:, 1] - d['time_bnds'][:, 0]
d['time_bnds'] = np.array(
[[np.amin(d['time_bnds'][:, 0]),
np.amax(d['time_bnds'][:, 1])]])
d['time'] = np.array([np.mean(d['time_bnds'], axis=1)])
if 'backscatter_sd' in d:
n, m = d['backscatter_sd'].shape
d['backscatter_sd'] = np.sqrt(1. / n * np.average(
d['backscatter_sd']**2,
axis=0,
weights=w,
))
d['backscatter_sd'] = d['backscatter_sd'].reshape([1, m])
for var in ds.get_vars(d):
if var in ('time', 'time_bnds', 'backscatter_sd'):
continue
if 'time' not in d['.'][var]['.dims']:
continue
i = d['.'][var]['.dims'].index('time')
shape = list(d[var].shape)
d[var] = np.average(d[var], axis=i, weights=w)
shape[i] = 1
d[var] = d[var].reshape(shape)
def read(dirname, track, warnings=[], step=6. / 24.):
d_ll = ds.read(os.path.join(dirname, 'LL125.nc'),
['latitude', 'longitude', 'z'])
lat_ll = d_ll['latitude']
lon_ll = d_ll['longitude']
orog_ll = d_ll['z'][0, :, :] / 9.80665
dd_idx = ds.readdir(
dirname,
variables=['time', 'latitude', 'longitude'],
jd=True,
full=True,
warnings=warnings,
)
start_time = track['time'][0]
end_time = track['time'][-1]
d_out = {}
for var in VARS:
dd = []
var2 = TRANS[var]
for d_idx in dd_idx:
if var not in d_idx['.']:
continue
time = d_idx['time']
lat = d_idx['latitude']
lon = d_idx['longitude']
filename = d_idx['filename']
ii = np.nonzero((time >= start_time - step * 0.5)
& (time < end_time + step * 0.5))[0]
for i in ii:
t = time[i]
i2 = np.argmin(np.abs(track['time'] - time[i]))
lat0 = track['lat'][i2]
lon0 = track['lon'][i2]
j = np.argmin(np.abs(lat - lat0))
k = np.argmin(np.abs(lon - lon0))
j_ll = np.argmin(np.abs(lat_ll - lat0))
k_ll = np.argmin(np.abs(lon_ll - lon0))
d = ds.read(
filename,
VARS_AUX + [var],
sel={
'time': [i],
'latitude': j,
'longitude': k,
},
jd=True,
)
for a, b in TRANS.items():
if a in d.keys():
ds.rename(d, a, b)
d['lat'] = np.array([d['lat']])
d['lon'] = np.array([d['lon']])
d['orog'] = np.array([orog_ll[j_ll, k_ll]])
d['.']['lat']['.dims'] = ['time']
d['.']['lon']['.dims'] = ['time']
d['.']['orog'] = {'.dims': ['time']}
if 'pfull' in ds.get_vars(d):
d['pfull'] = d['pfull'].reshape([1, len(d['pfull'])])
d['.']['pfull']['.dims'] = ['time', 'pfull']
d['pfull'] = d['pfull'][:, ::-1]
d[var2] = d[var2][:, ::-1]
ds.select(d, {'pfull': np.arange(27)})
dd.append(d)
d = ds.op.merge(dd, 'time')
for var_aux in VARS_AUX:
if TRANS[var_aux] in ds.get_vars(d_out) \
and TRANS[var_aux] in ds.get_vars(d) \
and not np.all(d_out[TRANS[var_aux]] == d[TRANS[var_aux]]):
raise ValueError('%s: Field differs between input files' %
TRANS[var_aux])
d_out.update(d)
d_out['pfull'] = d_out['pfull'] * 1e2
if 'time' in d_out:
d_out['time_bnds'] = misc.time_bnds(d_out['time'], step, start_time,
end_time)
d_out['time'] = np.mean(d_out['time_bnds'], axis=1)
d_out['.'] = META
return d_out
