from __future__ import annotations
import abc
import copy
from typing import Any, Iterable, Optional, Dict, Tuple
from cosapp.drivers.driver import Driver, System
from cosapp.utils.logging import logging
from cosapp.core.execution import (
Pool,
ExecutionPolicy,
ExecutionType,
Task,
Job,
Batch,
TaskAction,
TaskResponseStatus,
FunctionCallBehavior,
ops,
)
logger = logging.getLogger(__name__)
# TODO
# [ ] Quid for vector variables
[docs]
class AbstractSetOfCases(Driver):
"""
This driver builds a set of cases from a list
Parameters
----------
name : str
Name of the driver
owner : System, optional
:py:class:`~cosapp.systems.system.System` to which driver belongs; defaults to `None`
**kwargs : Any
Keyword arguments will be used to set driver options
"""
__slots__ = ("cases", "_transients_variables", "_execution_policy")
def __init__(
self,
name: str,
owner: Optional[System] = None,
execution_policy: ExecutionPolicy = ExecutionPolicy(workers_count=1, execution_type=ExecutionType.SINGLE_THREAD),
**options
) -> None:
"""Initialize driver
Parameters
----------
name: str, optional
Name of the `Module`
owner: System, optional
:py:class:`~cosapp.systems.system.System` to which driver belongs; defaults to `None`
**kwargs:
Additional keywords arguments forwarded to base class.
"""
super().__init__(name, owner, **options)
# TODO Fred - is this not too much oriented for MonteCarlo or DoE? What about a mission for which input
# variables may not be the same on all points.
self.cases = None # type: Optional[Iterable[Any]]
# desc="List of cases to be carried out."
self._transients_variables = {}
self._execution_policy: ExecutionPolicy = execution_policy # Execution policy to use for computation
def _precase(self, index: int, case: Any):
"""Hook to be called before running each case.
Parameters
----------
case_idx : int
Index of the case
case : Any
Parameters for this case
"""
self.status = ""
self.error_code = "0"
@abc.abstractmethod
def _build_cases(self) -> None:
"""Generator of cases."""
pass
def _postcase(self, index: int, case: Any):
"""Hook to be called after running each case.
Parameters
----------
case_idx : int
Index of the case
case : Any
Parameters for this case
"""
if (recorder := self._recorder):
recorder.record_state(index, self.status, self.error_code)
[docs]
def setup_run(self):
"""Actions performed prior to the `Module.compute` call."""
super().setup_run()
self._build_cases()
time_problem = self._owner.assembled_time_problem()
self._transients_variables = {
varname: copy.copy(transient.value)
for varname, transient in time_problem.transients.items()
}
[docs]
def run_children(self) -> None:
"""Runs all driver children.
"""
for child in self.children.values():
child.run_once()
if len(child.status) > 0:
self.status = child.status
if child.error_code != "0":
self.error_code = child.error_code
[docs]
def compute(self) -> None:
"""Contains the customized `Module` calculation, to execute after children.
"""
exec_policy = self._execution_policy
if exec_policy.is_sequential():
self._compute_sequential(self)
else:
self._compute_parallel()
@staticmethod
def _compute_sequential(driver: AbstractSetOfCases) -> None:
"""Contains the customized `Module` calculation, to execute after children.
"""
for case_idx, case in enumerate(driver.cases):
if len(case) > 0:
driver._precase(case_idx, case)
driver.run_children()
driver._postcase(case_idx, case)
@staticmethod
def _get_driver(system: System, name: str) -> AbstractSetOfCases:
return system.drivers[name]
@staticmethod
def _prepare_recorders(
driver: AbstractSetOfCases, exec_type: ExecutionType, chunk_id: int
) -> AbstractSetOfCases:
for d in driver.tree():
rec = d.recorder
if rec:
rec._enable_parallel_execution(exec_type, chunk_id)
rec.clear()
return driver
@staticmethod
def _modify_cases(driver: AbstractSetOfCases, rng) -> None:
lrng = list(rng)
start, stop = lrng[0], lrng[-1] + 1
driver.cases = driver.cases[start:stop]
return driver
@staticmethod
def _get_recorders_raw_data(driver: AbstractSetOfCases) -> Dict[str, Any]:
"""Gets recorders raw data.
Returns
-------
Dict[str, Any]
Recorders raw data dict, where keys are the drivers' names and values
are their recorders' raw data
"""
def make_items(driver: Driver) -> Tuple[str, Any]:
if (recorder := driver.recorder):
value = recorder._raw_data
else:
value = None
return (driver.name, value)
return dict(map(make_items, driver.tree(downwards=True)))
@staticmethod
def _compute_and_return_results(driver: AbstractSetOfCases) -> Dict[str, Any]:
"""Computes the driver and returns recorders raw data.
Returns
-------
Dict[str, Any]
Recorders raw data dict, where keys are the drivers' names and values
are their recorders' raw data
"""
AbstractSetOfCases._compute_sequential(driver)
return AbstractSetOfCases._get_recorders_raw_data(driver)
def _get_parallel_results(self, batch: Batch) -> Optional[Exception]:
"""Gets and dispatches results from parallel execution.
Parameters
----------
batch: Batch
Batch of jobs from which the results must be gathered
Returns
-------
Optional[Exception]
Exceptions raised workers' side if any
"""
mc_tree = list(self.tree(downwards=True))
for job in batch.jobs:
status, data = job.tasks[-1].result
if status != TaskResponseStatus.OK:
return data
for driver in mc_tree:
rec_data = data.get(driver.name, None)
if rec_data:
driver.recorder._batch_record(rec_data)
def _compute_parallel(self) -> None:
"""Computes the driver in parallel."""
exec_policy = self._execution_policy
pool = Pool.from_policy(exec_policy)
def create_job(id_and_range: Tuple[int, range]):
worker_id, index_range = id_and_range
store_system = Task(
TaskAction.FUNC_CALL,
FunctionCallBehavior.STORE_RETURNED_OBJECT,
(ops.return_arg, (self._owner,)),
)
get_mc = Task(
TaskAction.FUNC_CALL,
FunctionCallBehavior.CHAINED,
(self._get_driver, (self.name,)),
)
prepare_recorders = Task(
TaskAction.FUNC_CALL,
FunctionCallBehavior.CHAINED,
(self._prepare_recorders, (pool._type, worker_id)),
)
modify_cases = Task(
TaskAction.FUNC_CALL,
FunctionCallBehavior.CHAINED,
(self._modify_cases, (index_range,)),
)
compute_and_return_results = Task(
TaskAction.FUNC_CALL,
FunctionCallBehavior.CHAINED | FunctionCallBehavior.RETURN_OBJECT,
(self._compute_and_return_results, ()),
)
return Job([store_system, get_mc, prepare_recorders, modify_cases, compute_and_return_results])
blocks = Batch.compute_blocks(len(self.cases), exec_policy.workers_count)
batch = Batch(map(create_job, enumerate(blocks)))
with pool.activate():
pool.run_batch(batch)
batch.join()
error = self._get_parallel_results(batch)
if error:
raise error
[docs]
def run_once(self) -> None:
"""Run the driver once.
Do not call children `run_once` method, this is handled by the derived drivers.
"""
with self.log_context(" - run_once"):
if self.is_active():
self._precompute()
logger.debug(f"Call {self.name}.compute_before()")
self.compute_before()
logger.debug(f"Call {self.name}.compute()")
self._compute_calls += 1
self.compute()
self._postcompute()
self.computed.emit()
else:
logger.debug(f"Skip {self.name} execution - Inactive")
