def replan_with_optimizers(evaluations, external_plan, domain, optimizers):
# TODO: return multiple plans?
# TODO: can instead have multiple goal binding combinations
# TODO: can replan using samplers as well
if not is_plan(external_plan):
return None
optimizers = list(
filter(lambda s: isinstance(s, ComponentStream), optimizers))
if not optimizers:
return None
stream_plan, function_plan = partition_external_plan(external_plan)
free_parameters = {o for r in stream_plan for o in r.output_objects}
#free_parameters = {o for r in stream_plan for o in r.output_objects if isinstance(o, OptimisticObject)}
initial_evaluations = {
e: n
for e, n in evaluations.items() if n.result == INIT_EVALUATION
}
#initial_evaluations = evaluations
goal_facts = set()
for result in stream_plan:
goal_facts.update(
filter(
lambda f: evaluation_from_fact(f) not in initial_evaluations,
result.get_certified()))
visited_facts = set()
new_results = []
for fact in goal_facts:
retrace_instantiation(fact, optimizers, initial_evaluations,
free_parameters, visited_facts, new_results)
# TODO: ensure correct ordering
new_results = list(
filter(lambda r: isinstance(r, ComponentStream), new_results))
#from pddlstream.algorithms.scheduling.recover_streams import get_achieving_streams, extract_stream_plan
#node_from_atom = get_achieving_streams(evaluations, stream_results) # TODO: make these lower effort
#extract_stream_plan(node_from_atom, target_facts, stream_plan)
optimizer_results = []
for optimizer in {get_optimizer(r)
for r in new_results
}: # None is like a unique optimizer:
relevant_results = [
r for r in new_results if get_optimizer(r) == optimizer
]
optimizer_results.extend(
combine_optimizer_plan(relevant_results, function_plan))
#print(str_from_object(set(map(fact_from_evaluation, evaluations))))
#print(str_from_object(set(goal_facts)))
# TODO: can do the flexibly sized optimizers search
from pddlstream.algorithms.scheduling.postprocess import reschedule_stream_plan
optimizer_plan = reschedule_stream_plan(initial_evaluations,
goal_facts,
copy.copy(domain),
(stream_plan + optimizer_results),
unique_binding=True)
if not is_plan(optimizer_plan):
return None
return optimizer_plan + function_plan
def iterative_plan_streams(all_evaluations, externals, optimistic_solve_fn,
complexity_limit, **effort_args):
# Previously didn't have unique optimistic objects that could be constructed at arbitrary depths
complexity_evals = {
e: n
for e, n in all_evaluations.items() if n.complexity <= complexity_limit
}
num_iterations = 0
while True:
num_iterations += 1
results, exhausted = optimistic_process_streams(
complexity_evals, externals, complexity_limit, **effort_args)
stream_plan, action_plan, cost, final_depth = hierarchical_plan_streams(
complexity_evals,
externals,
results,
optimistic_solve_fn,
complexity_limit,
depth=0,
constraints=None,
**effort_args)
print('Attempt: {} | Results: {} | Depth: {} | Success: {}'.format(
num_iterations, len(results), final_depth, is_plan(action_plan)))
if is_plan(action_plan):
return stream_plan, action_plan, cost
if final_depth == 0:
status = INFEASIBLE if exhausted else FAILED
return status, status, cost
def replan_with_optimizers(evaluations, external_plan, domain, externals):
# TODO: return multiple plans?
# TODO: can instead have multiple goal binding combinations
# TODO: can replan using samplers as well
if not is_plan(external_plan):
return external_plan
optimizer_streams = list(
filter(lambda s: type(s) in [VariableStream, ConstraintStream],
externals))
if not optimizer_streams:
return external_plan
stream_plan, function_plan = partition_external_plan(external_plan)
goal_facts = set()
for result in stream_plan:
goal_facts.update(
filter(lambda f: evaluation_from_fact(f) not in evaluations,
result.get_certified()))
visited_facts = set()
new_results = []
for fact in goal_facts:
retrace_instantiation(fact, optimizer_streams, evaluations,
visited_facts, new_results)
variable_results = filter(lambda r: isinstance(r.external, VariableStream),
new_results)
constraint_results = filter(
lambda r: isinstance(r.external, ConstraintStream), new_results)
new_results = variable_results + constraint_results # TODO: ensure correct ordering
#from pddlstream.algorithms.scheduling.recover_streams import get_achieving_streams, extract_stream_plan
#node_from_atom = get_achieving_streams(evaluations, stream_results) # TODO: make these lower effort
#extract_stream_plan(node_from_atom, target_facts, stream_plan)
optimizer_results = []
for optimizer in {get_optimizer(r)
for r in new_results
}: # None is like a unique optimizer:
relevant_results = [
r for r in new_results if get_optimizer(r) == optimizer
]
optimizer_results.extend(
combine_optimizer_plan(relevant_results, function_plan))
#print(str_from_object(set(map(fact_from_evaluation, evaluations))))
#print(str_from_object(set(goal_facts)))
# TODO: can do the flexibly sized optimizers search
from pddlstream.algorithms.scheduling.postprocess import reschedule_stream_plan
combined_plan = reschedule_stream_plan(evaluations,
goal_facts,
copy.copy(domain),
(stream_plan + optimizer_results),
unique_binding=True,
unit_efforts=True)
if not is_plan(combined_plan):
return external_plan
return combined_plan + function_plan
def decompose_stream_plan(stream_plan):
if not is_plan(stream_plan):
return stream_plan
new_stream_plan = []
for result in stream_plan:
new_stream_plan.extend(decompose_result(result))
return new_stream_plan
def is_refined(stream_plan):
# TODO: lazily expand the shared objects in some cases to prevent increase in size
if not is_plan(stream_plan):
return True
# TODO: some of these opt_index equal None
return all((result.opt_index is None) or (result.opt_index == 0)
for result in stream_plan)
def value_from_obj_plan(obj_plan):
if not is_plan(obj_plan):
return obj_plan
#return [(action,) + tuple(values_from_objects(args)) for action, args in obj_plan]
#return [(action, tuple(values_from_objects(args))) for action, args in obj_plan]
value_plan = []
for action in obj_plan:
if len(action) == 3:
name, inputs, outputs = action
new_inputs = params_from_objects(inputs) # values_from_objects
new_outputs = outputs
if isinstance(new_outputs, collections.Sequence):
new_outputs = params_from_objects(
new_outputs) # values_from_objects
new_action = (name, new_inputs, new_outputs)
elif isinstance(action, DurativeAction):
name, args, start, duration = action
name, index = name[:-2], int(name[-1])
if index != 0:
continue
new_action = DurativeAction(name,
tuple(map(param_from_object, args)),
start, duration)
else:
new_action = transform_action_args(
action, param_from_object) # values_from_objects
value_plan.append(new_action)
return value_plan
def hierarchical_plan_streams(evaluations, externals, results, optimistic_solve_fn, complexity_limit,
depth, constraints, **effort_args):
if MAX_DEPTH <= depth:
return OptSolution(None, None, INF), depth
stream_plan, opt_plan, cost = optimistic_solve_fn(evaluations, results, constraints)
if not is_plan(opt_plan) or is_refined(stream_plan):
return OptSolution(stream_plan, opt_plan, cost), depth
#action_plan, preimage_facts = opt_plan
#dump_plans(stream_plan, action_plan, cost)
#create_visualizations(evaluations, stream_plan, depth)
#print(depth, get_length(stream_plan))
#print('Stream plan ({}, {:.3f}): {}\nAction plan ({}, {:.3f}): {}'.format(
# get_length(stream_plan), compute_plan_effort(stream_plan), stream_plan,
# get_length(action_plan), cost, str_from_plan(action_plan)))
#if is_plan(stream_plan):
# for result in stream_plan:
# effort = compute_result_effort(result, unit_efforts=True)
# if effort != 0:
# print(result, effort)
#print()
# TODO: identify control parameters that can be separated across actions
new_depth = depth + 1
new_results, bindings = optimistic_stream_evaluation(evaluations, stream_plan)
if not (CONSTRAIN_STREAMS or CONSTRAIN_PLANS):
return OptSolution(FAILED, FAILED, INF), new_depth
#if CONSTRAIN_STREAMS:
# next_results = compute_stream_results(evaluations, new_results, externals, complexity_limit, **effort_args)
#else:
next_results, _ = optimistic_process_streams(evaluations, externals, complexity_limit, **effort_args)
next_constraints = None
if CONSTRAIN_PLANS:
next_constraints = compute_skeleton_constraints(opt_plan, bindings)
return hierarchical_plan_streams(evaluations, externals, next_results, optimistic_solve_fn, complexity_limit,
new_depth, next_constraints, **effort_args)
def solve_restart(problem, max_time=INF, max_restarts=0, iteration_time=INF, abort=True, **kwargs):
# TODO: iteratively lower the cost bound
# TODO: a sequence of different planner configurations
# TODO: reset objects and/or streams
if (max_restarts >= 1) and (iteration_time == INF):
iteration_time = min(2 * 60, iteration_time)
assert (max_restarts == 0) or (iteration_time != INF)
assert max_restarts >= 0
start_time = time.time()
for attempt in irange(1+max_restarts):
iteration_start_time = time.time()
if elapsed_time(start_time) > max_time:
break
if attempt >= 1:
print(SEPARATOR)
#solution = planner_fn(problem) # Or include the problem in the lambda
remaining_time = min(iteration_time, max_time-elapsed_time(start_time))
solution = solve(problem, max_time=remaining_time, **kwargs)
plan, cost, certificate = solution
if is_plan(plan): # TODO: INFEASIBLE
return solution
if abort and (elapsed_time(iteration_start_time) < remaining_time):
break # TODO: return the cause of failure
certificate = Certificate(all_facts=[], preimage_facts=[]) # TODO: aggregate
return Solution(None, INF, certificate)
def stream_plan_complexity(evaluations, stream_plan, stream_calls, complexity_op=COMPLEXITY_OP):
if not is_plan(stream_plan):
return INF
# TODO: difference between a result having a particular complexity and the next result having something
#optimistic_facts = {}
optimistic_facts = {fact: evaluations[evaluation_from_fact(fact)].complexity
for fact in stream_plan_preimage(stream_plan)}
result_complexities = []
#complexity = 0
for i, result in enumerate(stream_plan):
# if result.external.get_complexity(num_calls=INF) == 0: # TODO: skip if true
result_complexity = complexity_op([0] + [optimistic_facts[fact]
#optimistic_complexity(evaluations, optimistic_facts, fact)
for fact in result.get_domain()])
# if stream_calls is None:
# num_calls = result.instance.num_calls
# else:
num_calls = stream_calls[i]
result_complexity += result.external.get_complexity(num_calls)
result_complexities.append(result_complexity)
#complexity = complexity_op(complexity, result_complexity)
for fact in result.get_certified():
if fact not in optimistic_facts:
optimistic_facts[fact] = result_complexity
complexity = complexity_op([0] + result_complexities)
return complexity
def hierarchical_plan_streams(evaluations, externals, results,
optimistic_solve_fn, complexity_limit, depth,
constraints, **effort_args):
if MAX_DEPTH <= depth:
return None, INF, depth
combined_plan, cost = optimistic_solve_fn(evaluations, results,
constraints)
if not is_plan(combined_plan):
return combined_plan, cost, depth
stream_plan, action_plan = separate_plan(combined_plan, stream_only=False)
#dump_plans(stream_plan, action_plan, cost)
#create_visualizations(evaluations, stream_plan, depth)
#print(depth, get_length(stream_plan))
if is_refined(stream_plan):
return combined_plan, cost, depth
new_results, bindings = optimistic_stream_evaluation(
evaluations, stream_plan)
if not CONSTRAIN_STREAMS and not CONSTRAIN_PLANS:
return None, INF, depth + 1
if CONSTRAIN_STREAMS:
next_results = compute_stream_results(evaluations, new_results,
externals, **effort_args)
else:
next_results, _ = optimistic_process_streams(evaluations, externals,
complexity_limit,
**effort_args)
next_constraints = None
if CONSTRAIN_PLANS:
next_constraints = compute_skeleton_constraints(action_plan, bindings)
return hierarchical_plan_streams(evaluations, externals, next_results,
optimistic_solve_fn, complexity_limit,
depth + 1, next_constraints,
**effort_args)
def stream_plan_complexity(evaluations, stream_plan, stream_calls=None):
if not is_plan(stream_plan):
return INF
# TODO: difference between a result having a particular complexity and the next result having something
optimistic_facts = {}
total_complexity = 0
for i, result in enumerate(stream_plan):
result_complexity = 0
for fact in result.get_domain():
evaluation = evaluation_from_fact(fact)
if evaluation in evaluations:
fact_complexity = evaluations[evaluation].complexity
else:
fact_complexity = optimistic_facts[fact]
result_complexity = COMPLEXITY_OP(result_complexity,
fact_complexity)
if stream_calls is None:
result_complexity += result.instance.num_calls + 1
elif i < len(stream_calls):
result_complexity += stream_calls[i] + 1
else:
result_complexity += 1
for fact in result.get_certified():
if fact not in optimistic_facts:
optimistic_facts[fact] = result_complexity
total_complexity = COMPLEXITY_OP(total_complexity, result_complexity)
return total_complexity
def hierarchical_plan_streams(evaluations, externals, results, optimistic_solve_fn, complexity_limit,
depth, constraints, **effort_args):
if MAX_DEPTH <= depth:
return None, INF, depth
combined_plan, cost = optimistic_solve_fn(evaluations, results, constraints)
if not is_plan(combined_plan):
return combined_plan, cost, depth
stream_plan, action_plan = separate_plan(combined_plan, stream_only=False)
#dump_plans(stream_plan, action_plan, cost)
#create_visualizations(evaluations, stream_plan, depth)
#print(depth, get_length(stream_plan))
#print('Stream plan ({}, {:.3f}): {}\nAction plan ({}, {:.3f}): {}'.format(
# get_length(stream_plan), compute_plan_effort(stream_plan), stream_plan,
# get_length(action_plan), cost, str_from_plan(action_plan)))
#if is_plan(stream_plan):
# for result in stream_plan:
# effort = compute_result_effort(result, unit_efforts=True)
# if effort != 0:
# print(result, effort)
#print()
if is_refined(stream_plan):
return combined_plan, cost, depth
new_results, bindings = optimistic_stream_evaluation(evaluations, stream_plan)
if not CONSTRAIN_STREAMS and not CONSTRAIN_PLANS:
return None, INF, depth + 1
if CONSTRAIN_STREAMS:
next_results = compute_stream_results(evaluations, new_results, externals, **effort_args)
else:
next_results, _ = optimistic_process_streams(evaluations, externals, complexity_limit, **effort_args)
next_constraints = None
if CONSTRAIN_PLANS:
next_constraints = compute_skeleton_constraints(action_plan, bindings)
return hierarchical_plan_streams(evaluations, externals, next_results, optimistic_solve_fn, complexity_limit,
depth + 1, next_constraints, **effort_args)
def combine_optimizers_greedy(evaluations, external_plan):
if not is_plan(external_plan):
return external_plan
# The key thing is that a variable must be grounded before it can used in a non-stream thing
# TODO: construct variables in order
# TODO: graph cut algorithm to minimize the number of constraints that are excluded
# TODO: reorder to ensure that constraints are done first since they are likely to fail as tests
incoming_edges, outgoing_edges = neighbors_from_orders(
get_partial_orders(external_plan))
queue = []
functions = []
for v in external_plan:
if not incoming_edges[v]:
(functions if isinstance(v, FunctionResult) else queue).append(v)
current = []
ordering = []
while queue:
optimizer = get_optimizer(current[-1]) if current else None
for v in queue:
if optimizer == get_optimizer(v):
current.append(v)
break
else:
ordering.extend(combine_optimizer_plan(current, functions))
current = [queue[0]]
v1 = current[-1]
queue.remove(v1)
for v2 in outgoing_edges[v1]:
incoming_edges[v2].remove(v1)
if not incoming_edges[v2]:
(functions
if isinstance(v2, FunctionResult) else queue).append(v2)
ordering.extend(combine_optimizer_plan(current, functions))
return ordering + functions
def value_from_obj_plan(obj_plan):
if not is_plan(obj_plan):
return obj_plan
value_plan = []
for action in obj_plan:
# TODO: I shouldn't need this decomposition any more, right?
if isinstance(action, StreamAction):
name, inputs, outputs = action
new_inputs = params_from_objects(inputs)
new_outputs = outputs
#if isinstance(new_outputs, collections.Sequence): # TODO: what was this for?
new_outputs = params_from_objects(new_outputs)
new_action = StreamAction(name, new_inputs, new_outputs)
elif isinstance(action, DurativeAction):
name, args, start, duration = action
name, index = name[:-2], int(name[-1])
if index != 0:
continue
new_action = DurativeAction(name,
tuple(map(param_from_object, args)),
start, duration)
elif isinstance(action, Action):
new_action = transform_action_args(
action, param_from_object) # values_from_objects
elif isinstance(action, Assignment):
new_action = transform_action_args(action, param_from_object)
else:
raise ValueError(action)
value_plan.append(new_action)
return value_plan
def compute_plan_effort(stream_plan, **kwargs):
# TODO: compute effort in the delete relaxation way
if not is_plan(stream_plan):
return INF
if not stream_plan:
return 0
return sum(result.get_effort(**kwargs) for result in stream_plan)
def add_plan(self, plan, cost):
# TODO: double-check that plan is a solution
if not is_plan(plan) or (self.best_cost <= cost):
return
solution = Solution(plan, cost)
self.best_plan, self.best_cost = solution
self.solutions.append(solution)
def process_stream_plan(store, domain, disabled, stream_plan, action_plan, cost, bind=True, max_failures=0):
# Bad old implementation of this method
# The only advantage of this vs skeleton is that this can avoid the combinatorial growth in bindings
if not is_plan(stream_plan):
return
stream_plan = [result for result in stream_plan if result.optimistic]
free_objects = get_free_objects(stream_plan)
bindings = {}
bound_plan = []
for i, opt_result in enumerate(stream_plan):
if (store.best_cost <= cost) or (max_failures < (i - len(bound_plan))):
# TODO: this terminates early when bind=False
break
opt_inputs = [inp for inp in opt_result.instance.input_objects if inp in free_objects]
if (not bind and opt_inputs) or not all(inp in bindings for inp in opt_inputs):
continue
bound_result = opt_result.remap_inputs(bindings)
bound_instance = bound_result.instance
if bound_instance.enumerated or not is_instance_ready(store.evaluations, bound_instance):
continue
# TODO: could remove disabled and just use complexity_limit
new_results = process_instance(store, domain, bound_instance)
if not bound_instance.enumerated:
disabled.add(bound_instance)
if not new_results:
continue
bound_plan.append(new_results[0])
bindings = update_bindings(bindings, bound_result, bound_plan[-1])
cost = update_cost(cost, opt_result, bound_plan[-1])
if len(stream_plan) == len(bound_plan):
store.add_plan(bind_action_plan(action_plan, bindings), cost)
def compute_plan_effort(stream_plan, **kwargs):
if not is_plan(stream_plan):
return INF
if not stream_plan:
return 0
return sum(
compute_result_effort(result, **kwargs) for result in stream_plan)
def reorder_stream_plan(stream_plan, **kwargs):
if not is_plan(stream_plan):
return stream_plan
stream_orders = get_partial_orders(stream_plan)
in_stream_orders, out_stream_orders = neighbors_from_orders(stream_orders)
valid_combine = lambda v, subset: out_stream_orders[v] <= subset
#valid_combine = lambda v, subset: in_stream_orders[v] & subset
return dynamic_programming(stream_plan, valid_combine, get_stream_stats, **kwargs)
def compute_expected_cost(stream_plan, stats_fn=Performance.get_statistics):
if not is_plan(stream_plan):
return INF
expected_cost = 0.
for result in reversed(stream_plan):
p_success, overhead = stats_fn(result)
expected_cost = overhead + p_success * expected_cost
return expected_cost
def test_init_goal(problem, **kwargs):
problem = create_simplified_problem(problem,
use_actions=False,
use_streams=False,
new_goal=None)
plan, cost, certificate = solve(problem, **kwargs)
assert not plan
is_goal = is_plan(plan)
return is_goal, certificate
def compute_expected_cost(stream_plan, stats_fn=get_stream_stats):
# TODO: prioritize cost functions as they can prune when we have a better plan
if not is_plan(stream_plan):
return INF
expected_cost = 0
for result in reversed(stream_plan):
p_success, overhead = stats_fn(result)
expected_cost = overhead + p_success * expected_cost
return expected_cost
def solve_incremental(problem, constraints=PlanConstraints(),
unit_costs=False, success_cost=INF,
max_iterations=INF, max_time=INF,
start_complexity=0, complexity_step=1, max_complexity=INF,
verbose=False, **search_args):
"""
Solves a PDDLStream problem by alternating between applying all possible streams and searching
:param problem: a PDDLStream problem
:param constraints: PlanConstraints on the set of legal solutions
:param max_time: the maximum amount of time to apply streams
:param max_iterations: the maximum amount of search iterations
:param unit_costs: use unit action costs rather than numeric costs
:param success_cost: an exclusive (strict) upper bound on plan cost to terminate
:param start_complexity: the stream complexity on the first iteration
:param complexity_step: the increase in the complexity limit after each iteration
:param max_complexity: the maximum stream complexity
:param verbose: if True, this prints the result of each stream application
:param search_args: keyword args for the search subroutine
:return: a tuple (plan, cost, evaluations) where plan is a sequence of actions
(or None), cost is the cost of the plan, and evaluations is init but expanded
using stream applications
"""
# max_complexity = 0 => current
# complexity_step = INF => exhaustive
# success_cost = terminate_cost = decision_cost
evaluations, goal_expression, domain, externals = parse_problem(
problem, constraints=constraints, unit_costs=unit_costs)
store = SolutionStore(evaluations, max_time, success_cost, verbose) # TODO: include other info here?
ensure_no_fluent_streams(externals)
if UPDATE_STATISTICS:
load_stream_statistics(externals)
num_iterations = num_calls = 0
complexity_limit = start_complexity
instantiator = Instantiator(externals, evaluations)
num_calls += process_stream_queue(instantiator, store, complexity_limit, verbose=verbose)
while not store.is_terminated() and (num_iterations <= max_iterations) and (complexity_limit <= max_complexity):
num_iterations += 1
print('Iteration: {} | Complexity: {} | Calls: {} | Evaluations: {} | Solved: {} | Cost: {} | Time: {:.3f}'.format(
num_iterations, complexity_limit, num_calls, len(evaluations),
store.has_solution(), store.best_cost, store.elapsed_time()))
plan, cost = solve_finite(evaluations, goal_expression, domain,
max_cost=min(store.best_cost, constraints.max_cost), **search_args)
if is_plan(plan):
store.add_plan(plan, cost)
if not instantiator:
break
if complexity_step is None:
# TODO: option to select the next k-smallest complexities
complexity_limit = instantiator.min_complexity()
else:
complexity_limit += complexity_step
num_calls += process_stream_queue(instantiator, store, complexity_limit, verbose=verbose)
#retrace_stream_plan(store, domain, goal_expression)
#print('Final queue size: {}'.format(len(instantiator)))
if UPDATE_STATISTICS:
write_stream_statistics(externals, verbose)
return store.extract_solution()
def check_dominated(skeleton_queue, stream_plan):
if not is_plan(stream_plan):
return True
for skeleton in skeleton_queue.skeletons:
# TODO: has stream_plans and account for different output object values
if frozenset(stream_plan) <= frozenset(skeleton.stream_plan):
print(stream_plan)
print(skeleton.stream_plan)
raise NotImplementedError()
def solve_skeleton(evaluations={}, opt_solutions=[], max_time=INF, success_cost=0,
max_complexity=INF, reorder=False, visualize=True):
store = SolutionStore(evaluations=evaluations, max_time=max_time, success_cost=success_cost, verbose=True)
skeleton_queue = SkeletonQueue(store, domain=None, disable=True)
for opt_solution in opt_solutions:
# TODO: add and then process later
stream_plan, opt_plan, cost = opt_solution
# stream_plan = replan_with_optimizers(evaluations, stream_plan, domain, externals) or stream_plan
stream_plan = combine_optimizers(evaluations, stream_plan)
# stream_plan = get_synthetic_stream_plan(stream_plan, # evaluations
# [s for s in synthesizers if not s.post_only])
# stream_plan = recover_optimistic_outputs(stream_plan)
if reorder:
# TODO: this blows up memory wise for long stream plans
stream_plan = reorder_stream_plan(store, stream_plan)
num_optimistic = sum(r.optimistic for r in stream_plan) if stream_plan else 0
action_plan = opt_plan.action_plan if is_plan(opt_plan) else opt_plan
print('Stream plan ({}, {}, {:.3f}): {}\nAction plan ({}, {:.3f}): {}'.format(
get_length(stream_plan), num_optimistic, compute_plan_effort(stream_plan), stream_plan,
get_length(action_plan), cost, str_from_plan(action_plan)))
if is_plan(stream_plan) and visualize:
num_iterations = 0
log_plans(stream_plan, action_plan, num_iterations)
#create_visualizations(evaluations, stream_plan, num_iterations)
visualize_stream_plan(stream_plan)
################
# # optimizer_plan = replan_with_optimizers(evaluations, stream_plan, domain, optimizers)
# optimizer_plan = None
# if optimizer_plan is not None:
# # TODO: post process a bound plan
# print('Optimizer plan ({}, {:.3f}): {}'.format(
# get_length(optimizer_plan), compute_plan_effort(optimizer_plan), optimizer_plan))
# skeleton_queue.new_skeleton(optimizer_plan, opt_plan, cost)
# TODO: these are quite different
#skeleton_queue.process(stream_plan, opt_plan, cost, complexity_limit=INF, max_time=INF)
skeleton_queue.process(stream_plan, opt_plan, cost, complexity_limit=max_complexity, max_time=0)
return store.extract_solution()
def decompose_stream_plan(stream_plan):
if not is_plan(stream_plan):
return stream_plan
new_stream_plan = []
for result in stream_plan:
if isinstance(result, SynthStreamResult):
new_stream_plan.extend(result.decompose())
else:
new_stream_plan.append(result)
return new_stream_plan
def process(self, stream_plan, action_plan, cost, complexity_limit, max_time=0):
# TODO: manually add stream_plans for synthesizers/optimizers
start_time = time.time()
if is_plan(stream_plan):
#print([result for result in stream_plan if result.optimistic])
#raw_input('New skeleton')
self.new_skeleton(stream_plan, action_plan, cost)
self.greedily_process()
elif stream_plan is INFEASIBLE:
# TODO: use complexity_limit
self.process_until_new()
self.timed_process(max_time - elapsed_time(start_time))
self.accelerate_best_bindings()
def recover_optimistic_outputs(stream_plan):
if not is_plan(stream_plan):
return stream_plan
new_mapping = {}
new_stream_plan = []
for result in stream_plan:
new_result = result.remap_inputs(new_mapping)
new_stream_plan.append(new_result)
if isinstance(new_result, StreamResult):
opt_result = new_result.instance.opt_results[
0] # TODO: empty if disabled
new_mapping.update(
safe_zip(new_result.output_objects, opt_result.output_objects))
return new_stream_plan
def reorder_combined_plan(evaluations, combined_plan, action_info, domain, **kwargs):
if not is_plan(combined_plan):
return combined_plan
stream_plan, action_plan = separate_plan(combined_plan)
orders = get_combined_orders(evaluations, stream_plan, action_plan, domain)
_, out_orders = neighbors_from_orders(orders)
valid_combine = lambda v, subset: out_orders[v] <= subset
def stats_fn(operator):
if isinstance(operator, Result):
return get_stream_stats(operator)
name, _ = operator
info = action_info[name]
return info.p_success, info.overhead
return dynamic_programming(combined_plan, valid_combine, stats_fn, **kwargs)
def process(self, stream_plan, action_plan, cost, complexity_limit, max_time=0, accelerate=False):
start_time = time.time()
if is_plan(stream_plan):
self.new_skeleton(stream_plan, action_plan, cost)
self.greedily_process()
elif (stream_plan is INFEASIBLE) and not self.process_until_new():
# Move this after process_complexity
return INFEASIBLE
if not self.queue:
return FAILED
# TODO: add and process
self.timed_process(max_time=(max_time - elapsed_time(start_time)))
self.process_complexity(complexity_limit)
if accelerate:
self.accelerate_best_bindings()
return FAILED
