# Copyright Swiss Data Science Center (SDSC). A partnership between
# École Polytechnique Fédérale de Lausanne (EPFL) and
# Eidgenössische Technische Hochschule Zürich (ETHZ).
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Plan execution."""
import itertools
import re
from functools import reduce
from pathlib import Path
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, cast
from pydantic import ConfigDict, validate_call
from renku.command.command_builder import inject
from renku.core import errors
from renku.core.interface.activity_gateway import IActivityGateway
from renku.core.interface.plan_gateway import IPlanGateway
from renku.core.lfs import check_external_storage, pull_paths_from_storage
from renku.core.plugin.provider import execute
from renku.core.util import communication
from renku.core.util.datetime8601 import local_now
from renku.core.util.os import is_subpath, safe_read_yaml
from renku.core.workflow.model.concrete_execution_graph import ExecutionGraph
from renku.core.workflow.plan import is_plan_removed
from renku.core.workflow.plan_factory import delete_indirect_files_list
from renku.core.workflow.value_resolution import ValueResolver
from renku.domain_model.project_context import project_context
from renku.domain_model.provenance.activity import Activity, ActivityCollection, WorkflowFileActivityCollection
from renku.domain_model.workflow.plan import AbstractPlan
from renku.domain_model.workflow.workflow_file import WorkflowFileCompositePlan
if TYPE_CHECKING:
from networkx import DiGraph
[docs]@inject.params(activity_gateway=IActivityGateway, plan_gateway=IPlanGateway)
def execute_workflow_graph(
dag: "DiGraph",
activity_gateway: IActivityGateway,
plan_gateway: IPlanGateway,
provider="toil",
config=None,
workflow_file_plan: Optional[WorkflowFileCompositePlan] = None,
):
"""Execute a Run with/without subprocesses.
Args:
dag(DiGraph): The workflow graph to execute.
activity_gateway(IActivityGateway): The injected activity gateway.
plan_gateway(IPlanGateway): The injected plan gateway.
provider: Provider to run the workflow with (Default value = "toil").
config: Path to config for the workflow provider (Default value = None).
workflow_file_plan (Optional[WorkflowFileCompositePlan): If passed, a workflow file is executed, so, store
related metadata.
"""
inputs = {i.actual_value for p in dag.nodes for i in p.inputs}
# NOTE: Pull inputs from Git LFS or other storage backends
if check_external_storage():
pull_paths_from_storage(project_context.repository, *inputs)
# check whether the none generated inputs of workflows are available
outputs = {o.actual_value for p in dag.nodes for o in p.outputs}
inputs = {i for i in inputs if i not in outputs and not any(is_subpath(path=i, base=o) for o in outputs)}
for i in inputs:
if not Path(i).exists():
raise errors.ParameterError(f"Invalid input value: Input '{i}' does not exist!", show_prefix=False)
delete_indirect_files_list(project_context.path)
if config:
config = safe_read_yaml(config)
started_at_time = local_now()
execute(dag=dag, basedir=project_context.path, provider=provider, config=config)
ended_at_time = local_now()
activities = []
for plan in dag.nodes:
# NOTE: Update plans are copies of Plan objects. We need to use the original Plan objects to avoid duplicates.
original_plan = plan_gateway.get_by_id(plan.id)
# NOTE: Workflow files don't have an original plan
if not original_plan:
original_plan = plan
activity = Activity.from_plan(
plan=plan,
repository=project_context.repository,
started_at_time=started_at_time,
ended_at_time=ended_at_time,
)
activity.association.plan = original_plan
activity_gateway.add(activity)
activities.append(activity)
if workflow_file_plan:
activity_collection = WorkflowFileActivityCollection.from_activities(
activities=activities, plan=workflow_file_plan
)
activity_gateway.add_activity_collection(activity_collection)
elif len(activities) > 1:
activity_collection = ActivityCollection(activities=activities)
activity_gateway.add_activity_collection(activity_collection)
[docs]def check_for_cycles(graph: ExecutionGraph):
"""Check for cycles in the graph and raises an error if there are any."""
if not graph.cycles:
return
cycles_str = []
for cycle in graph.cycles:
nodes = []
for node in cycle:
if isinstance(node, AbstractPlan):
nodes.append(f"[{node.name}]")
else:
cls = node.__class__.__name__.replace("Command", "")
nodes.append(f"{cls}: {node.actual_value}")
cycles_str.append(" -> ".join(nodes))
message = "Circular workflows are not supported in Renku. Please remove these cycles:\n\t"
message += "\n\t".join(cycles_str)
raise errors.GraphCycleError(message=message, cycles=[])
[docs]@inject.autoparams()
@validate_call(config=ConfigDict(arbitrary_types_allowed=True))
def execute_workflow(
name_or_id: str,
set_params: List[str],
provider: str,
config: Optional[str],
values: Optional[str],
plan_gateway: IPlanGateway,
):
"""Execute a plan with specified values.
Args:
name_or_id(str): Name or id of the Plan to iterate.
set_params(List[str]): List of values specified for workflow parameters.
provider(str): Name of the workflow provider backend to use for execution.
config(Optional[str]): Path to config for the workflow provider.
values(Optional[str]): Path to YAMl file containing values specified for workflow parameters.
plan_gateway(IPlanGateway): The plan gateway.
"""
workflow = plan_gateway.get_by_name_or_id(name_or_id)
if is_plan_removed(workflow):
raise errors.ParameterError(f"The specified workflow '{name_or_id}' cannot be found.")
# apply the provided parameter settings provided by user
override_params = dict()
if values:
override_params.update(safe_read_yaml(values))
if set_params:
from deepmerge import always_merger
for param in set_params:
name, value = param.split("=", maxsplit=1)
keys = name.split(".")
set_param = reduce(lambda x, y: {y: x}, reversed(keys), value) # type: ignore
override_params = always_merger.merge(override_params, set_param)
rv = ValueResolver.get(workflow, override_params)
workflow = rv.apply()
if rv.missing_parameters:
communication.warn(
f'Could not resolve the following parameters in "{workflow.name}" workflow: '
f'{",".join(rv.missing_parameters)}'
)
graph = ExecutionGraph([workflow], virtual_links=True)
execute_workflow_graph(dag=graph.workflow_graph, provider=provider, config=config)
def _extract_iterate_parameters(values: Dict[str, Any], index_pattern: re.Pattern, tag_separator: str = "@"):
"""Recursively extracts the iteration parameters from the workflow values given by the user.
Args:
values(Dict[str, Any]): Plan values to iterate over.
index_pattern(re.Pattern): Pattern for parameter indexes.
tag_separator(str, optional): Separator for tagged values (Default value = "@").
Returns:
Tuple of ``(iter_params, params)`` where ``params`` are regular parameters
and ``iter_params`` are parameters with iteration values.
"""
iter_params: Dict[str, Any] = {"indexed": {}, "params": {}, "tagged": {}}
params: Dict[str, Any] = {}
for param_name, param_value in values.items():
if isinstance(param_value, str) and index_pattern.search(param_value):
iter_params["indexed"][param_name] = param_value
params[param_name] = param_value
elif isinstance(param_value, list):
if len(param_value) == 1:
communication.warn(
f"The parameter '{param_name}' has only one element '{param_value}', "
"changing it to be a fixed parameter!"
)
params[param_name] = param_value[0]
continue
if tag_separator in param_name:
name, tag = param_name.split(tag_separator, maxsplit=1)
if tag in iter_params["tagged"]:
iter_params["tagged"][tag][name] = param_value
else:
iter_params["tagged"][tag] = {name: param_value}
params[name] = param_value
else:
iter_params["params"][param_name] = param_value
params[param_name] = param_value
elif isinstance(param_value, dict):
inner_iter_params, inner_params = _extract_iterate_parameters(param_value, index_pattern, tag_separator)
iter_params["params"].update([(f"{param_name}.{ik}", iv) for ik, iv in inner_iter_params["params"].items()])
iter_params["indexed"].update(
[(f"{param_name}.{ik}", iv) for ik, iv in inner_iter_params["indexed"].items()]
)
for tag, param in inner_iter_params["tagged"].items():
if tag in iter_params["tagged"]:
iter_params["tagged"][tag].update([(f"{param_name}.{ik}", iv) for ik, iv in param.items()])
else:
iter_params["tagged"][tag] = {f"{param_name}.{ik}": iv for ik, iv in param.items()}
params[param_name] = inner_params
else:
params[param_name] = param_value
return iter_params, params
def _validate_iterate_parameters(
workflow: AbstractPlan, workflow_params: Dict[str, Any], iter_params: Dict[str, Any]
) -> Optional[Dict[str, Any]]:
"""Validates the user provided iteration parameters.
Args:
workflow(AbstractPlan): The Plan to validate parameters against.
workflow_params(Dict[str, Any]): The plain parameters to check.
iter_params(Dict[str, Any]): The iterative parameters to check.
Returns:
Dictionary of validated iteration parameters.
"""
import copy
rv = ValueResolver.get(copy.deepcopy(workflow), workflow_params)
rv.apply()
mp_paths = [mp.split(".") for mp in rv.missing_parameters]
for collection in [iter_params["indexed"], iter_params["params"], *iter_params["tagged"].values()]:
remove_keys = []
for p in collection.keys():
parameter_path = p.split(".")
if any(parameter_path[: len(mp)] == mp for mp in mp_paths):
remove_keys.append(p)
for rk in remove_keys:
collection.pop(rk)
# validate tagged
empty_tags = []
for k, tagged_params in iter_params["tagged"].items():
if len(tagged_params) == 0:
empty_tags.append(k)
else:
tagged_params_values = list(tagged_params.values())
tag_size = len(tagged_params_values[0])
for p in tagged_params_values[1:]:
num_params = len(p)
if tag_size != num_params:
communication.error(
f"'{k}' tagged parameters '{tagged_params}' has different number of possible values!"
)
return None
for et in empty_tags:
iter_params["tagged"].pop(et)
if (len(iter_params["indexed"]) == 0) and (len(iter_params["params"]) == 0) and (len(iter_params["tagged"]) == 0):
raise errors.UsageError(
"Please check the provided mappings as none of the "
f"parameters are present in the '{workflow.name}' workflow"
)
if rv.missing_parameters:
communication.confirm(
f'Could not resolve the following parameters in "{workflow.name}" workflow: '
f'{", ".join(rv.missing_parameters)}. Resume the execution?',
abort=True,
)
return iter_params
def _build_iterations(
workflow: AbstractPlan, workflow_params: Dict[str, Any], iter_params: Dict[str, Any], index_pattern: re.Pattern
) -> Tuple[List[AbstractPlan], List[Dict]]:
"""Instantiate the workflows for each iteration.
Args:
workflow(AbstractPlan): The base workflow to use as a template.
workflow_params(Dict[str, Any]): The plain parameters to use.
iter_params(Dict[str, Any]): The iterative parameters to use.
index_pattern(re.Pattern): The pattern for the index placeholder.
Returns:
Tuple of ``(plans, itervalues)`` with ``plans`` being a list of all
plans for each iteration and ``itervalues`` being a list of all values
for each iteration.
"""
import copy
from deepmerge import always_merger
plans = []
execute_plan = []
columns = list(iter_params["params"].keys())
tagged_values = []
for tag in iter_params["tagged"].values():
columns.extend(tag.keys())
tagged_values.append(zip(*tag.values()))
def _flatten(values):
for i in values:
if isinstance(i, (list, tuple)):
yield from i
else:
yield i
for i, values in enumerate(itertools.product(*iter_params["params"].values(), *tagged_values)):
plan_params = copy.deepcopy(workflow_params)
iteration_values = {}
for k, v in iter_params["indexed"].items():
value = index_pattern.sub(str(i), v)
set_param = reduce(lambda x, y: {y: x}, reversed(k.split(".")), value) # type: ignore
plan_params = always_merger.merge(plan_params, set_param)
iteration_values[k] = value
for param_key, param_value in zip(columns, _flatten(values)):
set_param = reduce(lambda x, y: {y: x}, reversed(param_key.split(".")), param_value) # type: ignore
plan_params = always_merger.merge(plan_params, set_param)
iteration_values[param_key] = param_value
execute_plan.append(iteration_values)
rv = ValueResolver.get(copy.deepcopy(workflow), plan_params)
plans.append(rv.apply())
return plans, execute_plan
[docs]@inject.autoparams()
@validate_call(config=ConfigDict(arbitrary_types_allowed=True))
def iterate_workflow(
name_or_id: str,
mapping_path: Optional[str],
mappings: List[str],
dry_run: bool,
provider: str,
config: Optional[str],
plan_gateway: IPlanGateway,
):
"""Iterate a workflow repeatedly with differing values.
Args:
name_or_id(str): Name or id of the Plan to iterate.
mapping_path(str): Path to file defining workflow mappings.
mappings(List[str]): List of workflow mappings.
dry_run(bool): Whether to preview execution or actually run it.
provider(str): Name of the workflow provider backend to use for execution.
config(Optional[str]): Path to config for the workflow provider.
plan_gateway(IPlanGateway): The plan gateway.
"""
import ast
from deepmerge import always_merger
from renku.core.util.tabulate import tabulate
if mapping_path is None and len(mappings) == 0:
raise errors.UsageError("No mapping has been given for the iteration!")
workflow = plan_gateway.get_by_name_or_id(name_or_id)
if is_plan_removed(workflow):
raise errors.ParameterError(f"The specified workflow '{name_or_id}' cannot be found.")
tag_separator = "@"
index_pattern = re.compile(r"{iter_index}")
iter_params: Dict[str, Any] = {"indexed": {}, "params": {}, "tagged": {}}
workflow_params = {}
if mapping_path:
mapping = safe_read_yaml(mapping_path)
iter_params, workflow_params = _extract_iterate_parameters(mapping, index_pattern, tag_separator=tag_separator)
for m in mappings:
param_name, param_value = m.split("=", maxsplit=1)
if index_pattern.search(param_value):
iter_params["indexed"][param_name] = param_value # type: ignore
else:
try:
evaluated_param_value = ast.literal_eval(param_value)
except Exception:
raise errors.ParameterError(
f"The value of '{param_name}' parameter is neither a list nor templated variable!"
)
if isinstance(evaluated_param_value, list) and len(evaluated_param_value) == 1:
communication.warn(
f"The parameter '{param_name}' has only one element '{param_value}', "
"changing it to be a fixed parameter!"
)
workflow_params[param_name] = param_value[0]
continue
elif not isinstance(evaluated_param_value, list):
workflow_params[param_name] = param_value
continue
if tag_separator in param_name:
name, tag = param_name.split(tag_separator, maxsplit=1)
if tag in iter_params["tagged"]:
iter_params["tagged"][tag][name] = evaluated_param_value
else:
iter_params["tagged"][tag] = {name: evaluated_param_value}
param_name = name
else:
iter_params["params"][param_name] = evaluated_param_value
set_param = reduce(lambda x, y: {y: x}, reversed(param_name.split(".")), param_value) # type: ignore
workflow_params = always_merger.merge(workflow_params, set_param)
validated_iter_params = _validate_iterate_parameters(workflow, workflow_params, cast(Dict[str, Any], iter_params))
if validated_iter_params is None:
return
plans, execute_plan = _build_iterations(workflow, workflow_params, validated_iter_params, index_pattern)
communication.echo(f"\n\n{tabulate(execute_plan, execute_plan[0].keys())}")
if not dry_run:
graph = ExecutionGraph(workflows=plans, virtual_links=True)
execute_workflow_graph(dag=graph.workflow_graph, provider=provider, config=config)