Class: Puppet::Graph::SimpleGraph

Inherits:
Object
  • Object
show all
Defined in:
lib/puppet/graph/simple_graph.rb

Overview

A hopefully-faster graph class to replace the use of GRATR.

API:

  • public

Direct Known Subclasses

RelationshipGraph, Resource::Catalog

Defined Under Namespace

Classes: VertexWrapper

Class Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize ⇒ SimpleGraph

All public methods of this class must maintain (assume ^ ensure) the following invariants, where “=~=” means equiv. up to order:

@in_to.keys =~= @out_to.keys =~= all vertices
@in_to.values.collect { |x| x.values }.flatten =~= @out_from.values.collect { |x| x.values }.flatten =~= all edges
@in_to[v1][v2] =~= @out_from[v2][v1] =~= all edges from v1 to v2
@in_to   [v].keys =~= vertices with edges leading to   v
@out_from[v].keys =~= vertices with edges leading from v
no operation may shed reference loops (for gc)
recursive operation must scale with the depth of the spanning trees, or better (e.g. no recursion over the set
    of all vertices, etc.)

This class is intended to be used with DAGs. However, if the graph has a cycle, it will not cause non-termination of any of the algorithms.

API:

  • public



24
25
26
27
28
29
 
# File 'lib/puppet/graph/simple_graph.rb', line 24

def initialize
  @in_to = {}
  @out_from = {}
  @upstream_from = {}
  @downstream_from = {}
end

Class Attribute Details

.use_new_yaml_format ⇒ Object

API:

  • public



484
485
486
 
# File 'lib/puppet/graph/simple_graph.rb', line 484

def use_new_yaml_format
  @use_new_yaml_format
end

Instance Method Details

#add_edge(e, *a) ⇒ Object

Add a new edge. The graph user has to create the edge instance, since they have to specify what kind of edge it is.

API:

  • public



295
296
297
298
299
300
301
302
303
 
# File 'lib/puppet/graph/simple_graph.rb', line 295

def add_edge(e,*a)
  return add_relationship(e,*a) unless a.empty?
  @upstream_from.clear
  @downstream_from.clear
  add_vertex(e.source)
  add_vertex(e.target)
  @in_to[   e.target][e.source] ||= []; @in_to[   e.target][e.source] |= [e]
  @out_from[e.source][e.target] ||= []; @out_from[e.source][e.target] |= [e]
end

#add_relationship(source, target, label = nil) ⇒ Object

API:

  • public



305
306
307
 
# File 'lib/puppet/graph/simple_graph.rb', line 305

def add_relationship(source, target, label = nil)
  add_edge Puppet::Relationship.new(source, target, label)
end

#add_vertex(vertex) ⇒ Object

Add a new vertex to the graph.

API:

  • public



268
269
270
271
 
# File 'lib/puppet/graph/simple_graph.rb', line 268

def add_vertex(vertex)
  @in_to[vertex]    ||= {}
  @out_from[vertex] ||= {}
end

#adjacent(v, options = {}) ⇒ Object

Find adjacent edges.

API:

  • public



338
339
340
341
 
# File 'lib/puppet/graph/simple_graph.rb', line 338

def adjacent(v, options = {})
  return [] unless ns = (options[:direction] == :in) ? @in_to[v] : @out_from[v]
  (options[:type] == :edges) ? ns.values.flatten : ns.keys
end

#clear ⇒ Object

Clear our graph.

API:

  • public



32
33
34
35
36
37
 
# File 'lib/puppet/graph/simple_graph.rb', line 32

def clear
  @in_to.clear
  @out_from.clear
  @upstream_from.clear
  @downstream_from.clear
end

#dependencies(resource) ⇒ Object

Which resources depend upon the given resource.

API:

  • public



40
41
42
 
# File 'lib/puppet/graph/simple_graph.rb', line 40

def dependencies(resource)
  vertex?(resource) ? upstream_from_vertex(resource).keys : []
end

#dependents(resource) ⇒ Object

API:

  • public



44
45
46
 
# File 'lib/puppet/graph/simple_graph.rb', line 44

def dependents(resource)
  vertex?(resource) ? downstream_from_vertex(resource).keys : []
end

#direct_dependencies_of(v) ⇒ Object

API:

  • public



396
397
398
 
# File 'lib/puppet/graph/simple_graph.rb', line 396

def direct_dependencies_of(v)
  (@in_to[v] || {}).keys
end

#direct_dependents_of(v) ⇒ Object

API:

  • public



382
383
384
 
# File 'lib/puppet/graph/simple_graph.rb', line 382

def direct_dependents_of(v)
  (@out_from[v] || {}).keys
end

#directed? ⇒ Boolean

Whether our graph is directed. Always true. Used to produce dot files.

Returns:

API:

  • public



49
50
51
 
# File 'lib/puppet/graph/simple_graph.rb', line 49

def directed?
  true
end

#dotty(params = {}, dotfile = 'graph.dot') ⇒ Object

Call dotty for the graph which is written to the file ‘graph.dot’ in the # current directory.

API:

  • public



462
463
464
465
 
# File 'lib/puppet/graph/simple_graph.rb', line 462

def dotty (params = {}, dotfile = 'graph.dot')
  File.open(dotfile, 'w') {|f| f << to_dot(params) }
  system('dotty', dotfile)
end

#downstream_from_vertex(v) ⇒ Object

API:

  • public



372
373
374
375
376
377
378
379
380
 
# File 'lib/puppet/graph/simple_graph.rb', line 372

def downstream_from_vertex(v)
  return @downstream_from[v] if @downstream_from[v]
  result = @downstream_from[v] = {}
  @out_from[v].keys.each do |node|
    result[node] = 1
    result.update(downstream_from_vertex(node))
  end
  result
end

#each_edge ⇒ Object

API:

  • public



323
324
325
 
# File 'lib/puppet/graph/simple_graph.rb', line 323

def each_edge
  @in_to.each { |t,ns| ns.each { |s,es| es.each { |e| yield e }}}
end

#edge?(source, target) ⇒ Boolean

Is there an edge between the two vertices?

Returns:

API:

  • public



315
316
317
 
# File 'lib/puppet/graph/simple_graph.rb', line 315

def edge?(source, target)
  vertex?(source) and vertex?(target) and @out_from[source][target]
end

#edges ⇒ Object

API:

  • public



319
320
321
 
# File 'lib/puppet/graph/simple_graph.rb', line 319

def edges
  @in_to.values.collect { |x| x.values }.flatten
end

#edges_between(source, target) ⇒ Object

Find all matching edges.

API:

  • public



310
311
312
 
# File 'lib/puppet/graph/simple_graph.rb', line 310

def edges_between(source, target)
  (@out_from[source] || {})[target] || []
end

#find_cycles_in_graph ⇒ Object

Find all cycles in the graph by detecting all the strongly connected components, then eliminating everything with a size of one as uninteresting - which it is, because it can’t be a cycle. :)

This has an unhealthy relationship with the ‘tarjan’ method above, which it uses to implement the detection of strongly connected components.

API:

  • public



160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
 
# File 'lib/puppet/graph/simple_graph.rb', line 160

def find_cycles_in_graph
  state = {
    :number => 0, :index => {}, :lowlink => {}, :scc => [],
    :stack => [], :seen => {}
  }

  # we usually have a disconnected graph, must walk all possible roots
  vertices.each do |vertex|
    if ! state[:index][vertex] then
      tarjan vertex, state
    end
  end

  # To provide consistent results to the user, given that a hash is never
  # assured to return the same order, and given our graph processing is
  # based on hash tables, we need to sort the cycles internally, as well as
  # the set of cycles.
  #
  # Given we are in a failure state here, any extra cost is more or less
  # irrelevant compared to the cost of a fix - which is on a human
  # time-scale.
  state[:scc].select do |component|
    multi_vertex_component?(component) || single_vertex_referring_to_self?(component)
  end.map do |component|
    component.sort
  end.sort
end

#instance_variable_get(v) ⇒ Object

instance_variable_get is used by Object.to_zaml to get instance variables. Override it so that we can simulate the presence of instance variables @edges and @vertices for serialization.

API:

  • public



505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
 
# File 'lib/puppet/graph/simple_graph.rb', line 505

def instance_variable_get(v)
  case v.to_s
  when '@edges' then
    edges
  when '@vertices' then
    if self.class.use_new_yaml_format
      vertices
    else
      result = {}
      vertices.each do |vertex|
        adjacencies = {}
        [:in, :out].each do |direction|
          adjacencies[direction] = {}
          adjacent(vertex, :direction => direction, :type => :edges).each do |edge|
            other_vertex = direction == :in ? edge.source : edge.target
            (adjacencies[direction][other_vertex] ||= Set.new).add(edge)
          end
        end
        result[vertex] = Puppet::Graph::SimpleGraph::VertexWrapper.new(vertex, adjacencies)
      end
      result
    end
  else
    super(v)
  end
end

#leaves(vertex, direction = :out) ⇒ Object

Determine all of the leaf nodes below a given vertex.

API:

  • public



54
55
56
 
# File 'lib/puppet/graph/simple_graph.rb', line 54

def leaves(vertex, direction = :out)
  tree_from_vertex(vertex, direction).keys.find_all { |c| adjacent(c, :direction => direction).empty? }
end

#matching_edges(event, base = nil) ⇒ Object

Collect all of the edges that the passed events match. Returns an array of edges.

API:

  • public



60
61
62
63
64
65
66
67
68
69
70
71
 
# File 'lib/puppet/graph/simple_graph.rb', line 60

def matching_edges(event, base = nil)
  source = base || event.resource

  unless vertex?(source)
    Puppet.warning "Got an event from invalid vertex #{source.ref}"
    return []
  end
  # Get all of the edges that this vertex should forward events
  # to, which is the same thing as saying all edges directly below
  # This vertex in the graph.
  @out_from[source].values.flatten.find_all { |edge| edge.match?(event.name) }
end

#path_between(f, t) ⇒ Object

Return an array of the edge-sets between a series of n+1 vertices (f=v0,v1,v2…t=vn)

connecting the two given verticies.  The ith edge set is an array containing all the
edges between v(i) and v(i+1); these are (by definition) never empty.

  * if f == t, the list is empty
  * if they are adjacent the result is an array consisting of
    a single array (the edges from f to t)
  * and so on by induction on a vertex m between them
  * if there is no path from f to t, the result is nil

This implementation is not particularly efficient; it’s used in testing where clarity

is more important than last-mile efficiency.

API:

  • public



413
414
415
416
417
418
419
420
421
422
423
424
 
# File 'lib/puppet/graph/simple_graph.rb', line 413

def path_between(f,t)
  if f==t
    []
  elsif direct_dependents_of(f).include?(t)
    [edges_between(f,t)]
  elsif dependents(f).include?(t)
    m = (dependents(f) & direct_dependencies_of(t)).first
    path_between(f,m) + path_between(m,t)
  else
    nil
  end
end

#paths_in_cycle(cycle, max_paths = 1) ⇒ Object

Perform a BFS on the sub graph representing the cycle, with a view to generating a sufficient set of paths to report the cycle meaningfully, and ideally usefully, for the end user.

BFS is preferred because it will generally report the shortest paths through the graph first, which are more likely to be interesting to the user. I think; it would be interesting to verify that. –daniel 2011-01-23

Raises:

API:

  • public



195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
 
# File 'lib/puppet/graph/simple_graph.rb', line 195

def paths_in_cycle(cycle, max_paths = 1)
  raise ArgumentError, "negative or zero max_paths" if max_paths < 1

  # Calculate our filtered outbound vertex lists...
  adj = {}
  cycle.each do |vertex|
    adj[vertex] = adjacent(vertex).select{|s| cycle.member? s}
  end

  found = []

  # frame struct is vertex, [path]
  stack = [[cycle.first, []]]
  while frame = stack.shift do
    if frame[1].member?(frame[0]) then
      found << frame[1] + [frame[0]]
      break if found.length >= max_paths
    else
      adj[frame[0]].each do |to|
        stack.push [to, frame[1] + [frame[0]]]
      end
    end
  end

  return found.sort
end

#remove_edge!(e) ⇒ Object

Remove an edge from our graph.

API:

  • public



328
329
330
331
332
333
334
335
 
# File 'lib/puppet/graph/simple_graph.rb', line 328

def remove_edge!(e)
  if edge?(e.source,e.target)
    @upstream_from.clear
    @downstream_from.clear
    @in_to   [e.target].delete e.source if (@in_to   [e.target][e.source] -= [e]).empty?
    @out_from[e.source].delete e.target if (@out_from[e.source][e.target] -= [e]).empty?
  end
end

#remove_vertex!(v) ⇒ Object

Remove a vertex from the graph.

API:

  • public



274
275
276
277
278
279
280
281
 
# File 'lib/puppet/graph/simple_graph.rb', line 274

def remove_vertex!(v)
  return unless vertex?(v)
  @upstream_from.clear
  @downstream_from.clear
  (@in_to[v].values+@out_from[v].values).flatten.each { |e| remove_edge!(e) }
  @in_to.delete(v)
  @out_from.delete(v)
end

#report_cycles_in_graph ⇒ Object

Raises:

API:

  • public



222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
 
# File 'lib/puppet/graph/simple_graph.rb', line 222

def report_cycles_in_graph
  cycles = find_cycles_in_graph
  n = cycles.length           # where is "pluralize"? --daniel 2011-01-22
  return if n == 0
  s = n == 1 ? '' : 's'

  message = "Found #{n} dependency cycle#{s}:\n"
  cycles.each do |cycle|
    paths = paths_in_cycle(cycle)
    message += paths.map{ |path| '(' + path.join(" => ") + ')'}.join("\n") + "\n"
  end

  if Puppet[:graph] then
    filename = write_cycles_to_graph(cycles)
    message += "Cycle graph written to #{filename}."
  else
    message += "Try the '--graph' option and opening the "
    message += "resulting '.dot' file in OmniGraffle or GraphViz"
  end

  raise Puppet::Error, message
end

#reversal ⇒ Object

Return a reversed version of this graph.

API:

  • public



74
75
76
77
78
79
80
81
 
# File 'lib/puppet/graph/simple_graph.rb', line 74

def reversal
  result = self.class.new
  vertices.each { |vertex| result.add_vertex(vertex) }
  edges.each do |edge|
    result.add_edge edge.class.new(edge.target, edge.source, edge.label)
  end
  result
end

#size ⇒ Object

Return the size of the graph.

API:

  • public



84
85
86
 
# File 'lib/puppet/graph/simple_graph.rb', line 84

def size
  vertices.size
end

#tarjan(root, s) ⇒ Object

This is a simple implementation of Tarjan’s algorithm to find strongly connected components in the graph; this is a fairly ugly implementation, because I can’t just decorate the vertices themselves.

This method has an unhealthy relationship with the find_cycles_in_graph method below, which contains the knowledge of how the state object is maintained.

API:

  • public



99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
 
# File 'lib/puppet/graph/simple_graph.rb', line 99

def tarjan(root, s)
  # initialize the recursion stack we use to work around the nasty lack of a
  # decent Ruby stack.
  recur = [{ :node => root }]

  while not recur.empty? do
    frame = recur.last
    vertex = frame[:node]

    case frame[:step]
    when nil then
      s[:index][vertex]   = s[:number]
      s[:lowlink][vertex] = s[:number]
      s[:number]          = s[:number] + 1

      s[:stack].push(vertex)
      s[:seen][vertex] = true

      frame[:children] = adjacent(vertex)
      frame[:step]     = :children

    when :children then
      if frame[:children].length > 0 then
        child = frame[:children].shift
        if ! s[:index][child] then
          # Never seen, need to recurse.
          frame[:step] = :after_recursion
          frame[:child] = child
          recur.push({ :node => child })
        elsif s[:seen][child] then
          s[:lowlink][vertex] = [s[:lowlink][vertex], s[:index][child]].min
        end
      else
        if s[:lowlink][vertex] == s[:index][vertex] then
          this_scc = []
          begin
            top = s[:stack].pop
            s[:seen][top] = false
            this_scc << top
          end until top == vertex
          s[:scc] << this_scc
        end
        recur.pop               # done with this node, finally.
      end

    when :after_recursion then
      s[:lowlink][vertex] = [s[:lowlink][vertex], s[:lowlink][frame[:child]]].min
      frame[:step] = :children

    else
      fail "#{frame[:step]} is an unknown step"
    end
  end
end

#to_a ⇒ Object

API:

  • public



88
89
90
 
# File 'lib/puppet/graph/simple_graph.rb', line 88

def to_a
  vertices
end

#to_dot(params = {}) ⇒ Object

Output the dot format as a string

API:

  • public



458
 
# File 'lib/puppet/graph/simple_graph.rb', line 458

def to_dot (params={}) to_dot_graph(params).to_s; end

#to_dot_graph(params = {}) ⇒ Object

Return a DOT::DOTDigraph for directed graphs or a DOT::DOTSubgraph for an undirected Graph. params can contain any graph property specified in rdot.rb. If an edge or vertex label is a kind of Hash then the keys which match dot properties will be used as well.

API:

  • public



432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
 
# File 'lib/puppet/graph/simple_graph.rb', line 432

def to_dot_graph (params = {})
  params['name'] ||= self.class.name.gsub(/:/,'_')
  fontsize   = params['fontsize'] ? params['fontsize'] : '8'
  graph      = (directed? ? DOT::DOTDigraph : DOT::DOTSubgraph).new(params)
  edge_klass = directed? ? DOT::DOTDirectedEdge : DOT::DOTEdge
  vertices.each do |v|
    name = v.to_s
    params = {'name'     => '"'+name+'"',
      'fontsize' => fontsize,
      'label'    => name}
    v_label = v.to_s
    params.merge!(v_label) if v_label and v_label.kind_of? Hash
    graph << DOT::DOTNode.new(params)
  end
  edges.each do |e|
    params = {'from'     => '"'+ e.source.to_s + '"',
      'to'       => '"'+ e.target.to_s + '"',
      'fontsize' => fontsize }
    e_label = e.to_s
    params.merge!(e_label) if e_label and e_label.kind_of? Hash
    graph << edge_klass.new(params)
  end
  graph
end

#to_yaml_properties ⇒ Object

API:

  • public



532
533
534
535
 
# File 'lib/puppet/graph/simple_graph.rb', line 532

def to_yaml_properties
  (super + [:@vertices, :@edges] -
   [:@in_to, :@out_from, :@upstream_from, :@downstream_from]).uniq
end

#tree_from_vertex(start, direction = :out) ⇒ Object

A different way of walking a tree, and a much faster way than the one that comes with GRATR.

API:

  • public



364
365
366
367
368
369
370
 
# File 'lib/puppet/graph/simple_graph.rb', line 364

def tree_from_vertex(start, direction = :out)
  predecessor={}
  walk(start, direction) do |parent, child|
    predecessor[child] = parent
  end
  predecessor
end

#upstream_from_vertex(v) ⇒ Object

API:

  • public



386
387
388
389
390
391
392
393
394
 
# File 'lib/puppet/graph/simple_graph.rb', line 386

def upstream_from_vertex(v)
  return @upstream_from[v] if @upstream_from[v]
  result = @upstream_from[v] = {}
  @in_to[v].keys.each do |node|
    result[node] = 1
    result.update(upstream_from_vertex(node))
  end
  result
end

#vertex?(v) ⇒ Boolean

Test whether a given vertex is in the graph.

Returns:

API:

  • public



284
285
286
 
# File 'lib/puppet/graph/simple_graph.rb', line 284

def vertex?(v)
  @in_to.include?(v)
end

#vertices ⇒ Object

Return a list of all vertices.

API:

  • public



289
290
291
 
# File 'lib/puppet/graph/simple_graph.rb', line 289

def vertices
  @in_to.keys
end

#walk(source, direction) ⇒ Object

Just walk the tree and pass each edge.

API:

  • public



344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
 
# File 'lib/puppet/graph/simple_graph.rb', line 344

def walk(source, direction)
  # Use an iterative, breadth-first traversal of the graph. One could do
  # this recursively, but Ruby's slow function calls and even slower
  # recursion make the shorter, recursive algorithm cost-prohibitive.
  stack = [source]
  seen = Set.new
  until stack.empty?
    node = stack.shift
    next if seen.member? node
    connected = adjacent(node, :direction => direction)
    connected.each do |target|
      yield node, target
    end
    stack.concat(connected)
    seen << node
  end
end

#write_cycles_to_graph(cycles) ⇒ Object

API:

  • public



245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
 
# File 'lib/puppet/graph/simple_graph.rb', line 245

def write_cycles_to_graph(cycles)
  # This does not use the DOT graph library, just writes the content
  # directly.  Given the complexity of this, there didn't seem much point
  # using a heavy library to generate exactly the same content. --daniel 2011-01-27
  Puppet.settings.use(:graphing)

  graph = ["digraph Resource_Cycles {"]
  graph << '  label = "Resource Cycles"'

  cycles.each do |cycle|
    paths_in_cycle(cycle, 10).each do |path|
      graph << path.map { |v| '"' + v.to_s.gsub(/"/, '\\"') + '"' }.join(" -> ")
    end
  end

  graph << '}'

  filename = File.join(Puppet[:graphdir], "cycles.dot")
  File.open(filename, "w") { |f| f.puts graph }
  return filename
end

#write_graph(name) ⇒ Object

Produce the graph files if requested.

API:

  • public



468
469
470
471
472
473
474
475
476
477
 
# File 'lib/puppet/graph/simple_graph.rb', line 468

def write_graph(name)
  return unless Puppet[:graph]

  Puppet.settings.use(:graphing)

  file = File.join(Puppet[:graphdir], "#{name}.dot")
  File.open(file, "w") { |f|
    f.puts to_dot("name" => name.to_s.capitalize)
  }
end

#yaml_initialize(tag, var) ⇒ Object

API:

  • public



537
538
539
540
541
542
543
544
545
546
547
548
549
550
 
# File 'lib/puppet/graph/simple_graph.rb', line 537

def yaml_initialize(tag, var)
  initialize()
  vertices = var.delete('vertices')
  edges = var.delete('edges')
  if vertices.is_a?(Hash)
    # Support old (2.6) format
    vertices = vertices.keys
  end
  vertices.each { |v| add_vertex(v) }
  edges.each { |e| add_edge(e) }
  var.each do |varname, value|
    instance_variable_set("@#{varname}", value)
  end
end