def copy_log_from_where(rundate, pipeline, args):
kwargs = dict()
for a in args.split():
if "=" in a:
a = a.split("=", maxsplit=1)
kwargs[a[0].lstrip("-")] = a[1]
file_vars = dict(
**config.program_vars(rundate, pipeline, use_options=False, **kwargs),
**config.date_vars(rundate))
log_level = config.where.runner.log_level.str
current_level = "none"
try:
with config.files.open("log", file_vars=file_vars) as fid:
for line in fid:
line_level, _, text = line.partition(" ")
line_level = line_level.strip().lower()
current_level = line_level if line_level else current_level
text = text.strip()
if getattr(LogLevel, current_level) >= getattr(
LogLevel, log_level) and text:
# strip the 20 date characters from the text
log.log(text[20:], current_level)
except FileNotFoundError as err:
log.warn(f"'{err}'")
def list_datasets(rundate, tech, session, stage, **kwargs):
"""List datasets in a given dataset file
Args:
rundate: Datetime, the model run date.
tech: String, the technique.
stage: String, the stage.
kwargs: Other arguments are passed to files.open.
Returns:
List of strings describing the datasets.
"""
file_vars = dict(
config.program_vars(rundate,
tech,
session=session,
stage=stage,
**kwargs), **config.date_vars(rundate))
try:
with files.open("dataset_json", file_vars=file_vars) as fid:
json_data = json.load(fid)
except FileNotFoundError:
return list()
log.fatal(
f"No data found for {tech.upper()} {stage} {rundate.strftime(config.FMT_date)}"
)
return sorted(k for k in json_data.keys()
if not k.startswith("_") and "/" in k)
def parse_dataset_id(rundate, tech, stage, dataset_name, dataset_id, **kwargs):
"""Allow for some advanced handling of dataset_id
In addition to using regular numbers as dataset_id, some text keywords can be used:
+ 'last': Use the last dataset_id written to file, default 0 if no file is previously written.
+ 'all': Return a list of all dataset_ids in the file.
"""
if isinstance(dataset_id, (float, int)):
return dataset_id
# Use the JSON-file to find information about the dataset ids
file_vars = dict(
config.program_vars(rundate,
tech,
session=dataset_name,
stage=stage,
**kwargs), **config.date_vars(rundate))
try:
with files.open("dataset_json", file_vars=file_vars) as fid:
json_data = json.load(fid)
except FileNotFoundError:
json_data = dict()
if dataset_id == "last":
# If _last_dataset_id is not given, use dataset_id=0 as default
return json_data.get(dataset_name, dict()).get("_last_dataset_id", 0)
if dataset_id == "all":
return [
int(k.split("/")[-1]) for k in json_data.keys()
if k.startswith("{}/".format(dataset_name))
]
def __init__(self,
rundate,
tech,
stage,
dataset_name,
dataset_id,
empty=False,
**kwargs):
"""Create a new Dataset or read an existing one
Note:
Be aware that the implementation is dependent on ``self._fields`` being the first attribute to be set. See
:func:`__setattr__` for more information.
Args:
rundate: Date, the model run date.
tech: String, the technique.
stage: String, the stage.
dataset_name: String, the name of the dataset.
dataset_id: Int, id of the dataset.
empty: Boolean, if False (default) will read dataset from disk if available.
"""
self._fields = dict()
self._data = dict()
self._num_obs = 0
self._default_field_suffix = None
self._kwargs = kwargs
self._kwargs.setdefault("session",
dataset_name) # TODO: Can this be removed?
self.vars = dict(
config.program_vars(**dict(
kwargs,
rundate=rundate,
tech_name=tech,
stage=stage,
dataset_name=dataset_name,
dataset_id=str(dataset_id),
)))
self.vars.update(**kwargs)
self.rundate = rundate
dataset_id = _data.parse_dataset_id(rundate, tech, stage, dataset_name,
dataset_id)
self.name = "{name}/{id:04d}".format(name=dataset_name, id=dataset_id)
self.meta = dict()
# Try to read dataset from disk unless explicitly told to create an empty dataset
if not empty:
try:
self.read()
except FileNotFoundError:
pass
def copy_log_from_where(rundate, pipeline, session):
file_vars = dict(**config.program_vars(rundate, pipeline, session),
**config.date_vars(rundate))
log_level = config.where.runner.log_level.str
current_level = "none"
try:
with files.open("log", file_vars=file_vars) as fid:
for line in fid:
line_level, _, text = line.partition(" ")
line_level = line_level.strip().lower()
current_level = line_level if line_level else current_level
text = text.strip()
if getattr(LogLevel, current_level) >= getattr(
LogLevel, log_level) and text:
log.log(text, current_level)
except FileNotFoundError as err:
log.warn(f"'{err}'")
def _concatenate_datasets(from_date: date, to_date: date, dset_vars: Dict[str,
str],
only_for_rundate: bool) -> np.ndarray:
"""Concatenate datasets
Args:
from_date: Start date for reading Dataset.
to_date: End date for reading Dataset.
dset_vars: Common Dataset variables.
only_for_rundate: Concatenate only data for given rundate.
"""
merged_vars = config.program_vars(rundate=from_date,
tech_name=dset_vars["tech"],
**dset_vars)
merged_vars["id"] += "_concatenated"
dset_merged = data.Dataset(
**dict(merged_vars, rundate=from_date, empty=True))
date_to_read = from_date
while date_to_read <= to_date:
dset = data.Dataset(rundate=date_to_read, **dset_vars)
current_date = date_to_read
date_to_read += timedelta(days=1)
if dset.num_obs == 0:
log.info(f"No data to read for {current_date}")
continue
if only_for_rundate:
_keep_data_only_for_rundate(dset)
if dset.num_obs == 0:
log.info(f"No data to read for {current_date}")
continue
log.info(f"Reading data for {current_date}")
if not dset_merged:
dset_merged.copy_from(dset)
else:
dset_merged.extend(dset)
return dset_merged
def concatenate_datasets(from_date, to_date, dset_vars):
merged_vars = config.program_vars(rundate=from_date, tech_name=dset_vars["tech"], **dset_vars)
merged_vars["id"] += "_concatenated"
dset_merged = data.Dataset(**dict(merged_vars, rundate=from_date, empty=True))
date_to_read = from_date
while date_to_read <= to_date:
dset = data.Dataset(rundate=date_to_read, **dset_vars)
current_date = date_to_read
date_to_read += timedelta(days=1)
if dset.num_obs == 0:
log.info(f"No data to read for {current_date}")
continue
log.info(f"Reading data for {current_date}")
if not dset_merged:
dset_merged.copy_from(dset)
else:
dset_merged.extend(dset)
return dset_merged
