def _create_digraph(self):
digraph = DiGraph()
digraph.add_nodes(self.jobs)
for edge in self._graph:
if edge[1] is not None:
digraph.add_edge(edge)
self._digraph = digraph
def _generate_directed_graph(int_list):
g = DirectedGraph()
max_index = len(int_list) - 1
g.add_nodes(list(range(len(int_list))))
for i, x in enumerate(int_list):
if i + x <= max_index:
g.add_edge((i, i + x))
if i - x >= 0:
g.add_edge((i, i - x))
return g
def __init__(self):
self.digraph = Digraph()
# A map of human readable labels to help users
self.nodemap = {} # {'label' : <node object> }
# A dictionary of nodes evaluated data. Values will be None if node needs evaluating
self.cache = {} # {'label' : {data} }
# A dictionary of errors associated with each node
self.errors = {} # {'label' : [ "errors"]}
# a set of updated nodes since last DAG evaluation
self.updated = set() # ['labels']
class DAG(object):
def __init__(self):
self.digraph = Digraph()
# A map of human readable labels to help users
self.nodemap = {} # {'label' : <node object> }
# A dictionary of nodes evaluated data. Values will be None if node needs evaluating
self.cache = {} # {'label' : {data} }
# A dictionary of errors associated with each node
self.errors = {} # {'label' : [ "errors"]}
# a set of updated nodes since last DAG evaluation
self.updated = set() # ['labels']
def write(self, location):
import pickle
fp = open(location, 'wb')
pickle.dump(self, fp, protocol=2)
fp.close()
def get_dot(self):
'Return string of dotgraph representation'
d = pygraph.readwrite.dot.write(self.digraph)
return d
def write_dot(self, filename):
dot = self.get_dot()
fh = open(filename, 'w')
fh.write(dot)
fh.close()
def get_dic(self):
'Return a dictionary representation of digraph'
st = breadth_first_search(self.digraph)[0]
return st
def get_edges(self):
edges = self.digraph.edges()
return edges
def add_node(self, node, label):
'Add a node to the DAG'
# Check for node instance
if not isinstance(node, Node):
raise Exception("Provided node must be of type 'Node'")
# Check if node object already in graph
if self.nodemap.has_key(label):
if node == self.nodemap[label]:
# No need to re-enter the same object
return
else:
raise Exception("Label '%s' already in use" % label)
elif node in self.nodemap.values():
raise Exception("Node already exists in DAG with label '%s'" % self.get_label_from_node(node))
# Register DAG for node notifications
node.register(self)
# Add node to graph
self.nodemap[label] = node
self.cache[label] = None
self.errors[label] = []
self.digraph.add_node(label)
def del_node(self, label):
'Remove a node from the DAG by label'
if not isinstance(label, str):
raise Exception("Provided label not of type 'str'")
if self.nodemap.has_key(label):
node = self.nodemap[label]
# Unregister DAG from node
node.unregister(self)
# Remove node
self.digraph.del_node(label)
del self.nodemap[label]
if self.cache.has_key(label):
del self.cache[label]
if self.errors.has_key(label):
del self.errors[label]
if label in self.updated:
self.updated.remove(label)
else:
raise Exception("Provided label not found in DAG")
def add_edge(self, parent_label, child_label):
'Add an edge to DAG between parent_label and child_label'
if not self.nodemap.has_key(parent_label):
raise Exception("Provided parent label not found in DAG")
elif not self.nodemap.has_key(child_label):
raise Exception("Provided child label not found in DAG")
edge = (parent_label, child_label)
self.digraph.add_edge(edge)
def del_edge(self, parent_label, child_label):
'Remove an edge from DAG between parent_label and child_label'
if not self.nodemap.has_key(parent_label):
raise Exception("Provided parent label not found in DAG")
elif not self.nodemap.has_key(child_label):
raise Exception("Provided child label not found in DAG")
edge = (parent_label, child_label)
self.digraph.del_edge(edge)
def get_nodes_from_labels(self, labels):
if isinstance(labels, list):
newlist = []
for label in labels:
node = self.get_node_from_label(label)
newlist.append(node)
elif isinstance(labels, str):
node = self.get_node_from_label(label)
return node
def evaluate_dag(self):
'Force every node to be evaluated'
# First invalidate all changed caches
self._invalidate_caches()
# Then call update on every node
sg = topological_sorting(self.digraph) # sorted graph
for label in sg:
self._update_node(label)
def evaluate_node(self, label):
'Return the correct data for a node after evaluation'
if isinstance(label, Node):
label = self.get_label_from_node(label)
elif not isinstance(label, str):
raise Exception("Provided label not of type 'str' or 'Node'")
if not self.nodemap.has_key(label):
raise Exception("Could not find node with label '%s' in DAG" % label)
# first we invalidate the caches for nodes that have changed
self._invalidate_caches()
# then update the node
self._update_node(label)
#
if self.cache[label] == None:
errors = self.errors[label]
plural = ""
if len(errors) > 1:
plural = "s"
msg = "Node '%s' could not be evaluated due to the following error%s:\n" % (label, plural)
for error in errors:
msg += "\t - %s\n" % error
raise Exception(msg)
else:
return self.cache[label]
def mark_updated(self, node):
'Method to add a node to the invalidated list - so it can be checked on next execution'
label = self.get_label_from_node(node)
self.updated.add(label)
def _invalidate_caches(self):
'invalidate the downstream caches of updated nodes'
if len(self.updated) == 0:
return
# Sort the nodes in worklist and remove duplicates
sg = topological_sorting(self.digraph) # sorted graph
worklist = []
# insert nodes into worklist in sorted order
for node in sg:
if node in self.updated:
worklist.append(node)
self.updated.clear()
# iterate through worklist
while worklist:
node = worklist.pop() # one item at a time
downstream = breadth_first_search(self.digraph, root=node)[1] # get all downstream labels
for n in downstream:
if n in worklist:
# remove labels that will already be done
worklist.remove(n)
# remove cache entries
self.cache[n] = None
def _update_node(self, label, rgraph=None):
'Recursively updates nodes up the graph'
# If cache already exists, this node is ok
if self.cache[label] != None:
return
# reverse digraph then get parents
if rgraph is None:
rgraph = self.digraph.reverse()
parents = rgraph.neighbors(label)
# recursively evaluate all parent nodes to check if updates are required
for i in parents:
self._update_node(i, rgraph)
# If we are here, cache needs to be generated
# default state of cache entry is None
self.cache[label] = None
self.errors[label] = []
# temp variables for final node attributes
errors = []
cache = {}
node = self.get_node_from_label(label)
while parents:
node1 = parents.pop() # Compare one parent node against all others at a time
if parents == []:
# merge evaluated node1 if nothing to compare with
c = self.cache[node1]
if c == None:
errors.append("Node '%s' could not be evaluated due to errors in upstream node '%s'" % (label, node1) )
# break because the error is upstream
break
cache.update(c)
else:# clear the update
for node2 in parents:
# get merge and conflict info
lhs = self.cache[node1]
if lhs == None:
errors.append("Node '%s' could not be evaluated due to errors in upstream node '%s'" % (label, node1) )
# break because the error is upstream
break
rhs = self.cache[node2]
if rhs == None:
errors.append("Node '%s' could not be evaluated due to errors in upstream node '%s'" % (label, node2) )
# break because the error is upstream
break
merge, conflicts = self._dict_merge(lhs, rhs)
for key in conflicts:
# Ignore conflict if it would be overwritten anyway
if key in node._data.keys():
pass
else:
errors.append("Key conflict between '%s' and '%s' for key '%s'" % (node1, node2, key))
# Dont break, because we can list all that is wront with this node by continuing
# update the cache with the merge
cache.update(merge)
# finally update the cache dic with the node data
cache.update(node._data)
# If there were any erros, append them to the object
self.errors[label] = errors
# write the cache entry
if self.errors[label] == []:
self.cache[label] = deepcopy(cache)
def _dict_merge(self, lhs, rhs):
'Merge two nodes data, returning the merged data and any conflicting keys'
merge = {} # merged dics
conflicts = [] # list of conflicting keys
for key in lhs.keys():
# auto-merge for missing key on right-hand-side.
if (not rhs.has_key(key)):
merge[key] = lhs[key]
# on collision, raise exception
elif (lhs[key] != rhs[key]):
conflicts.append(key)
for key in rhs.keys():
# auto-merge for missing key on left-hand-side.
if (not lhs.has_key(key)):
merge[key] = rhs[key]
return merge, conflicts
def get_nodes(self):
'Return all node objects in DAG'
return self.nodemap.values()
def get_labels(self):
'Return labels for all nodes in DAG'
return self.nodemap.keys()
def get_dict(self):
'Return a dictionary of {\'label\' : <node object>}'
return self.nodemap
def get_node_from_label(self, label):
'Return the node for given label'
if self.nodemap.has_key(label):
return self.nodemap[label]
else:
raise Exception("Could not find node in DAG with label '%s'" % label)
def get_label_from_node(self, node):
'Return the label for given node'
for k,v in self.nodemap.iteritems():
if v == node:
return k
raise Exception("Could not find node in DAG")
def show(self):
'Simple method to generate a dotgraph and render it with graphviz'
import gv
import os
import time
dot = self.get_dot()
# write the dotfile out for testing
self.write_dot('/tmp/dag.dot')
# Apply the dot layout to the graph
gvo = gv.readstring(dot) # graphviz object
gv.layout(gvo, 'dot')
# render the layout into the node attributes
gv.render(gvo)
# write to a temp file and display in default viewer
fileout = '/tmp/out.png'
if os.path.exists(fileout):
os.remove(fileout)
gv.render(gvo, 'png', fileout)
time.sleep(1)
os.system('xdg-open %s 2> /dev/null' % fileout)
