31
34
graph -- sequence of pairs of node->parents_list.
33
The result is a list of node names, such that all parents come before
36
The result is a list of node names, such that all parents come before their
36
39
node identifiers can be any hashable object, and are typically strings.
41
This function has the same purpose as the TopoSorter class, but uses a
42
different algorithm to sort the graph. That means that while both return a
43
list with parents before their child nodes, the exact ordering can be
46
topo_sort is faster when the whole list is needed, while when iterating
47
over a part of the list, TopoSorter.iter_topo_order should be used.
38
return TopoSorter(graph).sorted()
49
kg = _mod_graph.KnownGraph(dict(graph))
41
53
class TopoSorter(object):
43
55
def __init__(self, graph):
44
56
"""Topological sorting of a graph.
46
58
:param graph: sequence of pairs of node_name->parent_names_list.
47
59
i.e. [('C', ['B']), ('B', ['A']), ('A', [])]
48
60
For this input the output from the sort or
49
61
iter_topo_order routines will be:
52
64
node identifiers can be any hashable object, and are typically strings.
54
66
If you have a graph like [('a', ['b']), ('a', ['c'])] this will only use
57
69
The graph is sorted lazily: until you iterate or sort the input is
58
70
not processed other than to create an internal representation.
60
iteration or sorting may raise GraphCycleError if a cycle is present
72
iteration or sorting may raise GraphCycleError if a cycle is present
63
# a dict of the graph.
75
# store a dict of the graph.
64
76
self._graph = dict(graph)
65
self._visitable = set(self._graph)
67
# self._original_graph = dict(graph)
69
# this is a stack storing the depth first search into the graph.
70
self._node_name_stack = []
71
# at each level of 'recursion' we have to check each parent. This
72
# stack stores the parents we have not yet checked for the node at the
73
# matching depth in _node_name_stack
74
self._pending_parents_stack = []
75
# this is a set of the completed nodes for fast checking whether a
76
# parent in a node we are processing on the stack has already been
77
# emitted and thus can be skipped.
78
self._completed_node_names = set()
81
79
"""Sort the graph and return as a list.
83
81
After calling this the sorter is empty and you must create a new one.
85
83
return list(self.iter_topo_order())
97
95
def iter_topo_order(self):
98
96
"""Yield the nodes of the graph in a topological order.
100
98
After finishing iteration the sorter is empty and you cannot continue
102
visitable = set(graph)
104
# this is a stack storing the depth first search into the graph.
105
pending_node_stack = []
106
# at each level of 'recursion' we have to check each parent. This
107
# stack stores the parents we have not yet checked for the node at the
108
# matching depth in pending_node_stack
109
pending_parents_stack = []
111
# this is a set of the completed nodes for fast checking whether a
112
# parent in a node we are processing on the stack has already been
113
# emitted and thus can be skipped.
114
completed_node_names = set()
104
117
# now pick a random node in the source graph, and transfer it to the
105
# top of the depth first search stack.
106
node_name, parents = self._graph.popitem()
107
self._push_node(node_name, parents)
108
while self._node_name_stack:
109
# loop until this call completes.
110
parents_to_visit = self._pending_parents_stack[-1]
111
# if all parents are done, the revision is done
118
# top of the depth first search stack of pending nodes.
119
node_name, parents = graph.popitem()
120
pending_node_stack.append(node_name)
121
pending_parents_stack.append(list(parents))
123
# loop until pending_node_stack is empty
124
while pending_node_stack:
125
parents_to_visit = pending_parents_stack[-1]
126
# if there are no parents left, the revision is done
112
127
if not parents_to_visit:
113
128
# append the revision to the topo sorted list
114
# all the nodes parents have been added to the output, now
115
# we can add it to the output.
116
yield self._pop_node()
129
# all the nodes parents have been added to the output,
130
# now we can add it to the output.
131
popped_node = pending_node_stack.pop()
132
pending_parents_stack.pop()
133
completed_node_names.add(popped_node)
118
while self._pending_parents_stack[-1]:
119
# recurse depth first into a single parent
120
next_node_name = self._pending_parents_stack[-1].pop()
121
if next_node_name in self._completed_node_names:
122
# this parent was completed by a child on the
123
# call stack. skip it.
125
if next_node_name not in self._visitable:
127
# otherwise transfer it from the source graph into the
128
# top of the current depth first search stack.
130
parents = self._graph.pop(next_node_name)
132
# if the next node is not in the source graph it has
133
# already been popped from it and placed into the
134
# current search stack (but not completed or we would
135
# have hit the continue 4 lines up.
136
# this indicates a cycle.
137
raise errors.GraphCycleError(self._node_name_stack)
138
self._push_node(next_node_name, parents)
139
# and do not continue processing parents until this 'call'
143
def _push_node(self, node_name, parents):
144
"""Add node_name to the pending node stack.
146
Names in this stack will get emitted into the output as they are popped
149
self._node_name_stack.append(node_name)
150
self._pending_parents_stack.append(list(parents))
153
"""Pop the top node off the stack
155
The node is appended to the sorted output.
157
# we are returning from the flattened call frame:
158
# pop off the local variables
159
node_name = self._node_name_stack.pop()
160
self._pending_parents_stack.pop()
162
self._completed_node_names.add(node_name)
136
# recurse depth first into a single parent
137
next_node_name = parents_to_visit.pop()
139
if next_node_name in completed_node_names:
140
# parent was already completed by a child, skip it.
142
if next_node_name not in visitable:
143
# parent is not a node in the original graph, skip it.
146
# transfer it along with its parents from the source graph
147
# into the top of the current depth first search stack.
149
parents = graph.pop(next_node_name)
151
# if the next node is not in the source graph it has
152
# already been popped from it and placed into the
153
# current search stack (but not completed or we would
154
# have hit the continue 6 lines up). this indicates a
156
raise errors.GraphCycleError(pending_node_stack)
157
pending_node_stack.append(next_node_name)
158
pending_parents_stack.append(list(parents))
166
161
def merge_sort(graph, branch_tip, mainline_revisions=None, generate_revno=False):
167
162
"""Topological sort a graph which groups merges.
169
164
:param graph: sequence of pairs of node->parents_list.
170
:param branch_tip: the tip of the branch to graph. Revisions not
165
:param branch_tip: the tip of the branch to graph. Revisions not
171
166
reachable from branch_tip are not included in the
173
168
:param mainline_revisions: If not None this forces a mainline to be
286
281
C is the end of a cluster due to rule 1.
287
D is not the end of a cluster from rule 1, but is from rule 2: E
282
D is not the end of a cluster from rule 1, but is from rule 2: E
288
283
is not its left most ancestor
289
284
E is the end of a cluster due to rule 1
290
285
F might be but we need more data.
292
287
we show connecting lines to a parent when:
293
288
- The parent is the start of a merge within this cluster.
294
That is, the merge was not done to the mainline before this cluster
289
That is, the merge was not done to the mainline before this cluster
295
290
was merged to the mainline.
296
291
This can be detected thus:
297
* The parent has a higher merge depth and is the next revision in
292
* The parent has a higher merge depth and is the next revision in
300
295
The next revision in the list constraint is needed for this case:
302
B 1 [C, F] # we do not want to show a line to F which is depth 2
297
B 1 [C, F] # we do not want to show a line to F which is depth 2
304
C 1 [H] # note that this is a long line to show back to the
299
C 1 [H] # note that this is a long line to show back to the
305
300
ancestor - see the end of merge rules.
405
# the scheduling order is: F, E, D, C, B, A
400
# the scheduling order is: F, E, D, C, B, A
406
401
# that is - 'left subtree, right subtree, node'
407
402
# which would mean that when we schedule A we can emit the entire tree.
408
403
self._scheduled_nodes = []
409
# This records for each node when we have processed its left most
404
# This records for each node when we have processed its left most
410
405
# unmerged subtree. After this subtree is scheduled, all other subtrees
411
406
# have their merge depth increased by one from this nodes merge depth.
412
407
# it contains tuples - name, merge_depth
545
553
if not left_subtree_pushed_stack[-1]:
546
554
# recurse depth first into the primary parent
547
555
next_node_name = pending_parents_stack[-1].pop(0)
556
is_left_subtree = True
557
left_subtree_pushed_stack[-1] = True
549
559
# place any merges in right-to-left order for scheduling
550
560
# which gives us left-to-right order after we reverse
551
# the scheduled queue. XXX: This has the effect of
561
# the scheduled queue. XXX: This has the effect of
552
562
# allocating common-new revisions to the right-most
553
# subtree rather than the left most, which will
563
# subtree rather than the left most, which will
554
564
# display nicely (you get smaller trees at the top
555
565
# of the combined merge).
556
566
next_node_name = pending_parents_stack[-1].pop()
567
is_left_subtree = False
557
568
if next_node_name in completed_node_names:
558
569
# this parent was completed by a child on the
559
570
# call stack. skip it.
added
removed
bzrlib/tsort.py
