35
30
graph -- sequence of pairs of node->parents_list.
37
The result is a list of node names, such that all parents come before their
32
The result is a list of node names, such that all parents come before
40
35
node identifiers can be any hashable object, and are typically strings.
42
This function has the same purpose as the TopoSorter class, but uses a
43
different algorithm to sort the graph. That means that while both return a
44
list with parents before their child nodes, the exact ordering can be
47
topo_sort is faster when the whole list is needed, while when iterating
48
over a part of the list, TopoSorter.iter_topo_order should be used.
50
kg = _mod_graph.KnownGraph(dict(graph))
37
return TopoSorter(graph).sorted()
54
40
class TopoSorter(object):
56
42
def __init__(self, graph):
57
43
"""Topological sorting of a graph.
59
45
:param graph: sequence of pairs of node_name->parent_names_list.
60
46
i.e. [('C', ['B']), ('B', ['A']), ('A', [])]
61
47
For this input the output from the sort or
62
48
iter_topo_order routines will be:
65
51
node identifiers can be any hashable object, and are typically strings.
67
53
If you have a graph like [('a', ['b']), ('a', ['c'])] this will only use
70
56
The graph is sorted lazily: until you iterate or sort the input is
71
57
not processed other than to create an internal representation.
73
iteration or sorting may raise GraphCycleError if a cycle is present
59
iteration or sorting may raise GraphCycleError if a cycle is present
76
# store a dict of the graph.
62
# a dict of the graph.
77
63
self._graph = dict(graph)
64
self._visitable = set(self._graph)
66
# self._original_graph = dict(graph)
68
# this is a stack storing the depth first search into the graph.
69
self._node_name_stack = []
70
# at each level of 'recursion' we have to check each parent. This
71
# stack stores the parents we have not yet checked for the node at the
72
# matching depth in _node_name_stack
73
self._pending_parents_stack = []
74
# this is a set of the completed nodes for fast checking whether a
75
# parent in a node we are processing on the stack has already been
76
# emitted and thus can be skipped.
77
self._completed_node_names = set()
80
80
"""Sort the graph and return as a list.
82
82
After calling this the sorter is empty and you must create a new one.
84
84
return list(self.iter_topo_order())
96
96
def iter_topo_order(self):
97
97
"""Yield the nodes of the graph in a topological order.
99
99
After finishing iteration the sorter is empty and you cannot continue
103
visitable = set(graph)
105
# this is a stack storing the depth first search into the graph.
106
pending_node_stack = []
107
# at each level of 'recursion' we have to check each parent. This
108
# stack stores the parents we have not yet checked for the node at the
109
# matching depth in pending_node_stack
110
pending_parents_stack = []
112
# this is a set of the completed nodes for fast checking whether a
113
# parent in a node we are processing on the stack has already been
114
# emitted and thus can be skipped.
115
completed_node_names = set()
118
103
# now pick a random node in the source graph, and transfer it to the
119
# top of the depth first search stack of pending nodes.
120
node_name, parents = graph.popitem()
121
pending_node_stack.append(node_name)
122
pending_parents_stack.append(list(parents))
124
# loop until pending_node_stack is empty
125
while pending_node_stack:
126
parents_to_visit = pending_parents_stack[-1]
127
# if there are no parents left, the revision is done
104
# top of the depth first search stack.
105
node_name, parents = self._graph.popitem()
106
self._push_node(node_name, parents)
107
while self._node_name_stack:
108
# loop until this call completes.
109
parents_to_visit = self._pending_parents_stack[-1]
110
# if all parents are done, the revision is done
128
111
if not parents_to_visit:
129
112
# append the revision to the topo sorted list
130
# all the nodes parents have been added to the output,
131
# now we can add it to the output.
132
popped_node = pending_node_stack.pop()
133
pending_parents_stack.pop()
134
completed_node_names.add(popped_node)
113
# all the nodes parents have been added to the output, now
114
# we can add it to the output.
115
yield self._pop_node()
137
# recurse depth first into a single parent
138
next_node_name = parents_to_visit.pop()
140
if next_node_name in completed_node_names:
141
# parent was already completed by a child, skip it.
143
if next_node_name not in visitable:
144
# parent is not a node in the original graph, skip it.
147
# transfer it along with its parents from the source graph
148
# into the top of the current depth first search stack.
150
parents = graph.pop(next_node_name)
152
# if the next node is not in the source graph it has
153
# already been popped from it and placed into the
154
# current search stack (but not completed or we would
155
# have hit the continue 6 lines up). this indicates a
157
raise errors.GraphCycleError(pending_node_stack)
158
pending_node_stack.append(next_node_name)
159
pending_parents_stack.append(list(parents))
117
while self._pending_parents_stack[-1]:
118
# recurse depth first into a single parent
119
next_node_name = self._pending_parents_stack[-1].pop()
120
if next_node_name in self._completed_node_names:
121
# this parent was completed by a child on the
122
# call stack. skip it.
124
if next_node_name not in self._visitable:
126
# otherwise transfer it from the source graph into the
127
# top of the current depth first search stack.
129
parents = self._graph.pop(next_node_name)
131
# if the next node is not in the source graph it has
132
# already been popped from it and placed into the
133
# current search stack (but not completed or we would
134
# have hit the continue 4 lines up.
135
# this indicates a cycle.
136
raise errors.GraphCycleError(self._node_name_stack)
137
self._push_node(next_node_name, parents)
138
# and do not continue processing parents until this 'call'
142
def _push_node(self, node_name, parents):
143
"""Add node_name to the pending node stack.
145
Names in this stack will get emitted into the output as they are popped
148
self._node_name_stack.append(node_name)
149
self._pending_parents_stack.append(list(parents))
152
"""Pop the top node off the stack
154
The node is appended to the sorted output.
156
# we are returning from the flattened call frame:
157
# pop off the local variables
158
node_name = self._node_name_stack.pop()
159
self._pending_parents_stack.pop()
161
self._completed_node_names.add(node_name)
162
165
def merge_sort(graph, branch_tip, mainline_revisions=None, generate_revno=False):
163
166
"""Topological sort a graph which groups merges.
165
168
:param graph: sequence of pairs of node->parents_list.
166
:param branch_tip: the tip of the branch to graph. Revisions not
169
:param branch_tip: the tip of the branch to graph. Revisions not
167
170
reachable from branch_tip are not included in the
169
172
:param mainline_revisions: If not None this forces a mainline to be
283
285
C is the end of a cluster due to rule 1.
284
D is not the end of a cluster from rule 1, but is from rule 2: E
286
D is not the end of a cluster from rule 1, but is from rule 2: E
285
287
is not its left most ancestor
286
288
E is the end of a cluster due to rule 1
287
289
F might be but we need more data.
289
291
we show connecting lines to a parent when:
290
292
- The parent is the start of a merge within this cluster.
291
That is, the merge was not done to the mainline before this cluster
293
That is, the merge was not done to the mainline before this cluster
292
294
was merged to the mainline.
293
295
This can be detected thus:
294
* The parent has a higher merge depth and is the next revision in
296
* The parent has a higher merge depth and is the next revision in
297
299
The next revision in the list constraint is needed for this case:
299
B 1 [C, F] # we do not want to show a line to F which is depth 2
301
B 1 [C, F] # we do not want to show a line to F which is depth 2
301
C 1 [H] # note that this is a long line to show back to the
303
C 1 [H] # note that this is a long line to show back to the
302
304
ancestor - see the end of merge rules.
402
# the scheduling order is: F, E, D, C, B, A
398
# the scheduling order is: F, E, D, C, B, A
403
399
# that is - 'left subtree, right subtree, node'
404
400
# which would mean that when we schedule A we can emit the entire tree.
405
401
self._scheduled_nodes = []
406
# This records for each node when we have processed its left most
402
# This records for each node when we have processed its left most
407
403
# unmerged subtree. After this subtree is scheduled, all other subtrees
408
404
# have their merge depth increased by one from this nodes merge depth.
409
405
# it contains tuples - name, merge_depth
410
406
self._left_subtree_pushed_stack = []
412
408
# seed the search with the tip of the branch
413
if (branch_tip is not None and
414
branch_tip != _mod_revision.NULL_REVISION and
415
branch_tip != (_mod_revision.NULL_REVISION,)):
409
if branch_tip is not None:
416
410
parents = self._graph.pop(branch_tip)
417
411
self._push_node(branch_tip, 0, parents)
419
413
def sorted(self):
420
414
"""Sort the graph and return as a list.
422
416
After calling this the sorter is empty and you must create a new one.
424
418
return list(self.iter_topo_order())
426
420
def iter_topo_order(self):
427
421
"""Yield the nodes of the graph in a topological order.
429
423
After finishing iteration the sorter is empty and you cannot continue
555
537
if not left_subtree_pushed_stack[-1]:
556
538
# recurse depth first into the primary parent
557
539
next_node_name = pending_parents_stack[-1].pop(0)
558
is_left_subtree = True
559
left_subtree_pushed_stack[-1] = True
561
541
# place any merges in right-to-left order for scheduling
562
542
# which gives us left-to-right order after we reverse
563
# the scheduled queue. XXX: This has the effect of
543
# the scheduled queue. XXX: This has the effect of
564
544
# allocating common-new revisions to the right-most
565
# subtree rather than the left most, which will
545
# subtree rather than the left most, which will
566
546
# display nicely (you get smaller trees at the top
567
547
# of the combined merge).
568
548
next_node_name = pending_parents_stack[-1].pop()
569
is_left_subtree = False
570
549
if next_node_name in completed_node_names:
571
550
# this parent was completed by a child on the
572
551
# call stack. skip it.
added
removed
bzrlib/tsort.py
