def test_Stream(test_file):
dataset, request, key_count = TEST_FILES[test_file]
path = cdscommon.ensure_data(dataset, request, name='cds-' + test_file + '-{uuid}.grib')
stream = cfgrib.FileStream(path)
leader = stream.first()
assert len(leader) == key_count
assert sum(1 for _ in stream) == leader['count']
def test_reanalysis_Stream(test_file):
dataset, request, key_count = TEST_FILES[test_file]
path = cdscommon.ensure_data(dataset, request, name=test_file + "{ext}")
stream = cfgrib.FileStream(path)
leader = stream.first()
assert len(leader) == key_count
assert sum(1 for _ in stream) == leader["count"]
def _convert_unpacked_to_netcdf4(self, download_folder):
# A function that handles the many small GRIB files that were unpacked and formats those into NetCDF4
grib_files = glob.glob(
str(Path(download_folder).joinpath("HA40_N25_*_GB")))
lats, lons = self._build_grid_block(grib_files[0])
for grib_file in grib_files:
time_prediction_made, predicted_hour = self._get_time_from_file(
grib_file[len(download_folder) + 1:])
cfg_file = cfgrib.FileStream(grib_file)
file_dataset = None
for cfg_message in cfg_file:
# We skip any rotated grid data from the first file in each set of files. Only regular ll grids will do.
if cfg_message["gridType"] == "regular_ll":
field_name, line_dataset = self._process_cfg_message_to_line_dataset(
cfg_message=cfg_message,
lats=lats,
lons=lons,
time_prediction_made=time_prediction_made,
predicted_hour=predicted_hour)
if line_dataset is not None:
if file_dataset is None:
file_dataset = line_dataset
else:
file_dataset[field_name] = line_dataset[field_name]
save_filename = Path(download_folder).joinpath(
self.file_prefix + str(time_prediction_made.year) +
str(time_prediction_made.month).zfill(2) +
str(time_prediction_made.day).zfill(2) + "_" +
str(time_prediction_made.hour).zfill(2) +
"00_prediction_for_" + str(predicted_hour).zfill(2) + "00.nc")
file_dataset = file_dataset.unstack("coord")
file_dataset.time.encoding["units"] = "hours since 2018-01-01"
file_dataset.to_netcdf(path=save_filename, format="NETCDF4")
self.logger.info(f"Saved dataset as {save_filename}")
def _build_grid_block(file: str):
cf_streamed_file = cfgrib.FileStream(file)
line_to_use = None
for cf_line in cf_streamed_file:
# Make sure we don't pick up the rotated grid that exists in the first hourly file (0000) for each
# prediction..
if cf_line["gridType"] == "regular_ll":
line_to_use = cf_line
break
lat_first = float(line_to_use["latitudeOfFirstGridPointInDegrees"])
lat_last = float(line_to_use["latitudeOfLastGridPointInDegrees"])
lon_first = float(line_to_use["longitudeOfFirstGridPointInDegrees"])
lon_last = float(line_to_use["longitudeOfLastGridPointInDegrees"])
lat_step = float(line_to_use["jDirectionIncrement"])
lon_step = float(line_to_use["iDirectionIncrement"])
latitudes = list(
range(
int(lat_first * 1000),
int(lat_last * 1000) + int(lat_step),
int(lat_step),
))
longitudes = list(
range(
int(lon_first * 1000),
int(lon_last * 1000) + int(lon_step),
int(lon_step),
))
latitudes = [x / 1000 for x in latitudes]
longitudes = [x / 1000 for x in longitudes]
return latitudes, longitudes
