taskflow/taskflow/utils/flow_utils.py

221 lines
8.9 KiB
Python

# -*- coding: utf-8 -*-
# Copyright (C) 2013 Yahoo! Inc. All Rights Reserved.
#
# 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.
import logging
import threading
import networkx as nx
from taskflow import exceptions
from taskflow import flow
from taskflow.patterns import graph_flow as gf
from taskflow.patterns import linear_flow as lf
from taskflow.patterns import unordered_flow as uf
from taskflow import retry
from taskflow import task
from taskflow.utils import graph_utils as gu
from taskflow.utils import lock_utils as lu
from taskflow.utils import misc
LOG = logging.getLogger(__name__)
# Use the 'flatten' attribute as the need to add an edge here, which is useful
# for doing later analysis of the edges (to determine why the edges were
# created).
FLATTEN_EDGE_DATA = {
'flatten': True,
}
class Flattener(object):
def __init__(self, root, freeze=True):
self._root = root
self._graph = None
self._history = set()
self._freeze = bool(freeze)
self._lock = threading.Lock()
self._edge_data = FLATTEN_EDGE_DATA.copy()
def _add_new_edges(self, graph, nodes_from, nodes_to, edge_attrs=None):
"""Adds new edges from nodes to other nodes in the specified graph,
with the following edge attributes (defaulting to the class provided
edge_data if None), if the edge does not already exist.
"""
if edge_attrs is None:
edge_attrs = self._edge_data
else:
edge_attrs = edge_attrs.copy()
edge_attrs.update(self._edge_data)
for u in nodes_from:
for v in nodes_to:
if not graph.has_edge(u, v):
# NOTE(harlowja): give each edge its own attr copy so that
# if it's later modified that the same copy isn't modified.
graph.add_edge(u, v, attr_dict=edge_attrs.copy())
def _flatten(self, item):
functor = self._find_flattener(item)
if not functor:
raise TypeError("Unknown type requested to flatten: %s (%s)"
% (item, type(item)))
self._pre_item_flatten(item)
graph = functor(item)
self._post_item_flatten(item, graph)
return graph
def _find_flattener(self, item):
"""Locates the flattening function to use to flatten the given item."""
if isinstance(item, lf.Flow):
return self._flatten_linear
elif isinstance(item, uf.Flow):
return self._flatten_unordered
elif isinstance(item, gf.Flow):
return self._flatten_graph
elif isinstance(item, task.BaseTask):
return self._flatten_task
elif isinstance(item, retry.Retry):
raise TypeError("Retry controller %s (%s) is used not as a flow "
"parameter" % (item, type(item)))
else:
return None
def _connect_retry(self, retry, graph):
graph.add_node(retry)
# All graph nodes that has not predecessors should be depended on its
# retry
for n in gu.get_no_predecessors(graph):
if n != retry:
# modified that the same copy isn't modified.
graph.add_edge(retry, n, FLATTEN_EDGE_DATA.copy())
# Add link to retry for each node of subgraph that hasn't
# a parent retry
for n in graph.nodes_iter():
if n != retry and 'retry' not in graph.node[n]:
graph.add_node(n, {'retry': retry})
def _flatten_linear(self, flow):
"""Flattens a linear flow."""
graph = nx.DiGraph(name=flow.name)
previous_nodes = []
for item in flow:
subgraph = self._flatten(item)
graph = gu.merge_graphs([graph, subgraph])
# Find nodes that have no predecessor, make them have a predecessor
# of the previous nodes so that the linearity ordering is
# maintained. Find the ones with no successors and use this list
# to connect the next subgraph (if any).
self._add_new_edges(graph,
previous_nodes,
list(gu.get_no_predecessors(subgraph)))
# There should always be someone without successors, otherwise we
# have a cycle A -> B -> A situation, which should not be possible.
previous_nodes = list(gu.get_no_successors(subgraph))
return graph
def _flatten_unordered(self, flow):
"""Flattens a unordered flow."""
graph = nx.DiGraph(name=flow.name)
for item in flow:
# NOTE(harlowja): we do *not* connect the graphs together, this
# retains that each item (translated to subgraph) is disconnected
# from each other which will result in unordered execution while
# running.
graph = gu.merge_graphs([graph, self._flatten(item)])
return graph
def _flatten_task(self, task):
"""Flattens a individual task."""
graph = nx.DiGraph(name=task.name)
graph.add_node(task)
return graph
def _flatten_graph(self, flow):
"""Flattens a graph flow."""
graph = nx.DiGraph(name=flow.name)
# Flatten all nodes into a single subgraph per node.
subgraph_map = {}
for item in flow:
subgraph = self._flatten(item)
subgraph_map[item] = subgraph
graph = gu.merge_graphs([graph, subgraph])
# Reconnect all node edges to there corresponding subgraphs.
for (u, v) in flow.graph.edges_iter():
# Retain and update the original edge attributes.
u_v_attrs = gu.get_edge_attrs(flow.graph, u, v)
# Connect the ones with no predecessors in v to the ones with no
# successors in u (thus maintaining the edge dependency).
self._add_new_edges(graph,
list(gu.get_no_successors(subgraph_map[u])),
list(gu.get_no_predecessors(subgraph_map[v])),
edge_attrs=u_v_attrs)
return graph
def _pre_item_flatten(self, item):
"""Called before a item is flattened; any pre-flattening actions."""
if id(item) in self._history:
raise ValueError("Already flattened item: %s (%s), recursive"
" flattening not supported" % (item, id(item)))
LOG.debug("Starting to flatten '%s'", item)
self._history.add(id(item))
def _post_item_flatten(self, item, graph):
"""Called before a item is flattened; any post-flattening actions."""
if isinstance(item, flow.Flow) and item.retry:
self._connect_retry(item.retry, graph)
LOG.debug("Finished flattening '%s'", item)
# NOTE(harlowja): this one can be expensive to calculate (especially
# the cycle detection), so only do it if we know debugging is enabled
# and not under all cases.
if LOG.isEnabledFor(logging.DEBUG):
LOG.debug("Translated '%s' into a graph:", item)
for line in gu.pformat(graph).splitlines():
# Indent it so that it's slightly offset from the above line.
LOG.debug(" %s", line)
def _pre_flatten(self):
"""Called before the flattening of the item starts."""
self._history.clear()
def _post_flatten(self, graph):
"""Called after the flattening of the item finishes successfully."""
dup_names = misc.get_duplicate_keys(graph.nodes_iter(),
key=lambda node: node.name)
if dup_names:
dup_names = ', '.join(sorted(dup_names))
raise exceptions.InvariantViolation("Tasks with duplicate names "
"found: %s" % (dup_names))
self._history.clear()
@lu.locked
def flatten(self):
"""Flattens a item (a task or flow) into a single execution graph."""
if self._graph is not None:
return self._graph
self._pre_flatten()
graph = self._flatten(self._root)
self._post_flatten(graph)
if self._freeze:
self._graph = nx.freeze(graph)
else:
self._graph = graph
return self._graph
def flatten(item, freeze=True):
"""Flattens a item (a task or flow) into a single execution graph."""
return Flattener(item, freeze=freeze).flatten()