Class: Gecode::Model

Inherits:
Object
  • Object
show all
Defined in:
lib/gecoder/interface/model.rb,
lib/gecoder/interface/branch.rb,
lib/gecoder/interface/search.rb,
lib/gecoder/interface/enum_wrapper.rb

Overview

Model is the base class that all models must inherit from.

Formulating problems

Models are used to formulate the problems that Gecode should solve. Below is an example of a model that formulates the problem of finding a solution to the following equation system.

Equation system:

x + y = z
x = y - 3
0 <= x,y,z <= 9

Model:

class EquationProblem < Gecode::Model
  attr :vars

  def initialize
    x, y, z = @vars = int_var_array(3, 0..9)

    (x + y).must == z
    x.must == y - 3

    branch_on @vars
  end
end

A model typically consists of three main parts:

Variables

Variables specify how to view the problem. A solution is an assignment of the variables. In the example above we created an array of three integer variables with domains 0..9 and gave it the name variables.

There are three types of variables: integer variables (Gecode::IntVar, can be assigned one of many possible integer values), boolean variables (Gecode::BoolVar, can be assigned either true or false) and set variables (Gecode::SetVar, can be assigned a set of integers). Variables of the different types are constructed using #int_var, #int_var_array, #int_var_matrix, #bool_var, #bool_var_array, #bool_var_matrix, #set_var, #set_var_array and #set_var_matrix .

The various variables all have the functionality of Operand and have many properties depending on their type. For instance integer variables have the properties defined in Gecode::Int::IntOperand and enumerations of integer variables (such as the array variables we used) have the properties defined in

Gecode::IntEnum::IntEnumOperand .

Constraints

Constraints are placed on the variables to ensure that a valid assignment of the variables must also be a solution. In the example above we constrained the variables so that all equations were satisfied (which is exactly when we have found a solution).

The various constraints that can be placed on the various kinds of operands are found in the respective constraint receivers. For instance, the constraints that can be placed on integer operands are found in Gecode::Int::IntConstraintReceiver and the constraints that can be placed on enumerations of integer operands are found in Gecode::IntEnum::IntEnumConstraintReceiver .

Branching

“branch_on variables” in the example tells Gecode that it should explore the search space until it has assigned variables (or exhausted the search space). It also tells Gecode in what order the search space should be explore, which can have a big effect on the search performance. See #branch_on for details.

Finding solutions

Solutions to a formulated problem are found are found by using methods such as #solve!, #solution, #each_solution . If one wants to find a solution that optimizes a certain quantity (i.e. maximizes a certain variable) then one should have a look at #maximize!, #minimize! and #optimize! .

The first solution to the example above could for instance be found using

puts EquationProblem.new.solve!.vars.values.join(', ')

which would find the first solution to the problem, access the assigned values of variables and print them (in order x, y, z).

Shorter ways of formulating problems

Problems can also be formulated without defining a new class by using Gecode#solve et al.

Additionally one can use “foo_is_an …” to create an accessor of name foo, without having to use instance variables. The above problem becomes

class EquationProblem < Gecode::Model
  def initialize
    x, y, z = vars_is_an int_var_array(3, 0..9)

    (x + y).must == z
    x.must == y - 3

    branch_on vars
  end
end

Defined Under Namespace

Modules: Constants

Constant Summary collapse

MAX_INT =

The largest integer allowed in the domain of an integer variable.

Gecode::Raw::IntLimits::MAX
MIN_INT =

The smallest integer allowed in the domain of an integer variable.

Gecode::Raw::IntLimits::MIN
SET_MAX_INT =

The largest integer allowed in the domain of a set variable.

Gecode::Raw::SetLimits::MAX
SET_MIN_INT =

The smallest integer allowed in the domain of a set variable.

Gecode::Raw::SetLimits::MIN
LARGEST_INT_DOMAIN =

The largest possible domain for an integer variable.

MIN_INT..MAX_INT
NON_NEGATIVE_INT_DOMAIN =

The largest possible domain, without negative integers, for an integer variable.

0..MAX_INT
LARGEST_SET_BOUND =

The largest possible bound for a set variable.

SET_MIN_INT..SET_MAX_INT

Class Method Summary collapse

Instance Method Summary collapse

Dynamic Method Handling

This class handles dynamic methods through the method_missing method

#method_missing(name_symbol, *args) ⇒ Object

Wraps method missing to handle #foo_is_a and #foo_is_an .

“<variable_name>_is_a <variable>” or “<variable_name>_is_an <variable>”, replacing “<variable_name>” with the variable’s name and “<variable>” with the variable, adds an instance variable and accessor with the specified name.

The method also returns the variable given.

Example

# Add an instance variable and accessor named "foo" that return
# the integer variable.
foo_is_an int_var(0..9)

# Add an instance variable and accessor named "bar" that return
# the boolean variable array.
bar_is_a bool_var_array(2)



287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
 
# File 'lib/gecoder/interface/model.rb', line 287

def method_missing(name_symbol, *args)
  name = name_symbol.to_s
  if name =~ /._is_an?$/
    name.sub!(/_is_an?$/, '')
    unless args.size == 1
      raise ArgumentError, "Wrong number of argmuments (#{args.size} for 1)."
    end 
    if respond_to? name
      raise ArgumentError, "Method with name #{name} already exists."
    end
    if instance_variable_defined? "@#{name}"
      raise ArgumentError, 
        "Instance variable with name @#{name} already exists."
    end

    # We use the meta class to avoid defining the variable in all
    # other instances of the class.
    eval <<-"end_eval"
      @#{name} = args.first
      class <<self
        attr :#{name}
      end
    end_eval
    return args.first
  else
    super
  end
end

Class Method Details

.constrain(home, best) ⇒ Object

Called by spaces when they want to constrain as part of BAB-search.



173
174
175
176
177
178
179
 
# File 'lib/gecoder/interface/search.rb', line 173

def constrain(home, best) #:nodoc:
  if @constrain_proc.nil?
    raise NotImplementedError, 'Constrain method not implemented.' 
  else
    @constrain_proc.call(home, best)
  end
end

.constrain_proc=(proc) ⇒ Object

Sets the proc that should be used to handle constrain requests.



168
169
170
 
# File 'lib/gecoder/interface/search.rb', line 168

def constrain_proc=(proc) #:nodoc:
  @constrain_proc = proc
end

Instance Method Details

#active_space ⇒ Object

Retrieves the currently used space. Calling this method is only allowed when sanctioned by the model beforehand, e.g. when the model asks a constraint to post itself. Otherwise an RuntimeError is raised.

The space returned by this method should never be stored, it should be rerequested from the model every time that it’s needed.



223
224
225
226
227
228
229
 
# File 'lib/gecoder/interface/model.rb', line 223

def active_space #:nodoc:
  unless @allow_space_access
    raise 'Space access is restricted and the permission to access the ' + 
      'space has not been given.'
  end
  selected_space
end

#add_constraint(constraint) ⇒ Object

Adds the specified constraint to the model. Returns the newly added constraint.



233
234
235
236
237
238
 
# File 'lib/gecoder/interface/model.rb', line 233

def add_constraint(constraint) #:nodoc:
  add_interaction do
    constraint.post
  end
  return constraint
end

#add_interaction(&block) ⇒ Object

Adds a block containing something that interacts with Gecode to a queue where it is potentially executed.



242
243
244
 
# File 'lib/gecoder/interface/model.rb', line 242

def add_interaction(&block) #:nodoc:
  gecode_interaction_queue << block
end

#allow_space_access(&block) ⇒ Object

Allows the model’s active space to be accessed while the block is executed. Don’t use this unless you know what you’re doing. Anything that the space is used for (such as bound variables) must be released before the block ends.

Returns the result of the block.



252
253
254
255
256
257
258
259
260
261
 
# File 'lib/gecoder/interface/model.rb', line 252

def allow_space_access(&block) #:nodoc:
  # We store the old value so that nested calls don't become a problem, i.e.
  # access is allowed as long as one call to this method is still on the 
  # stack.
  old = @allow_space_access
  @allow_space_access = true
  res = yield
  @allow_space_access = old
  return res
end

#bool_var ⇒ Object

Creates a new boolean variable.



162
163
164
 
# File 'lib/gecoder/interface/model.rb', line 162

def bool_var
  BoolVar.new(self, variable_creation_space.new_bool_var)
end

#bool_var_array(count) ⇒ Object

Creates an array containing the specified number of boolean variables.



167
168
169
170
171
 
# File 'lib/gecoder/interface/model.rb', line 167

def bool_var_array(count)
  build_var_array(count) do
    BoolVar.new(self, variable_creation_space.new_bool_var)
  end
end

#bool_var_matrix(row_count, col_count) ⇒ Object

Creates a matrix containing the specified number rows and columns of boolean variables.



175
176
177
178
179
 
# File 'lib/gecoder/interface/model.rb', line 175

def bool_var_matrix(row_count, col_count)
  build_var_matrix(row_count, col_count) do
    BoolVar.new(self, variable_creation_space.new_bool_var)
  end
end

#branch_on(variables, options = {}) ⇒ Object

Specifies which variables that should be branched on (given as an enum of operands or as a single operand). One can optionally also select which of the variables that should be used first with the :variable option and which value in that variable’s domain that should be used with the :value option. If nothing is specified then :variable uses :none and value uses :min.

The following values can be used with :variable for integer and boolean enums:

:none

The first unassigned variable.

:smallest_min

The one with the smallest minimum.

:largest_min

The one with the largest minimum.

:smallest_max

The one with the smallest maximum.

:largest_max

The one with the largest maximum.

:smallest_size

The one with the smallest size.

:largest_size

The one with the larges size.

:smallest_degree

The one with the smallest degree. The degree of a variable is defined as the number of dependant propagators. In case of ties, choose the variable with smallest domain.

:largest_degree

The one with the largest degree. The degree of a variable is defined as the number of dependant propagators. In case of ties, choose the variable with smallest domain.

:smallest_min_regret

The one with the smallest min-regret. The min-regret of a variable is the difference between the smallest and second-smallest value still in the domain.

:largest_min_regret

The one with the largest min-regret. The min-regret of a variable is the difference between the smallest and second-smallest value still in the domain.

:smallest_max_regret

The one with the smallest max-regret The max-regret of a variable is the difference between the largest and second-largest value still in the domain.

:largest_max_regret

The one with the largest max-regret. The max-regret of a variable is the difference between the largest and second-largest value still in the domain.

The following values can be used with :value for integer and boolean enums:

:min

Selects the smallest value.

:med

Select the median value.

:max

Selects the largest vale

:split_min

Selects the lower half of the domain.

:split_max

Selects the upper half of the domain.

The following values can be used with :variable for set enums:

:none

The first unassigned set.

:smallest_cardinality

The one with the smallest cardinality.

:largest_cardinality

The one with the largest cardinality.

:smallest_unknown

The one with the smallest number of unknown elements

:largest_unknown

The one with the largest number of unknown elements

The following values can be used with :value set enums:

:min

Selects the smallest value in the unknown part of the set.

:max

Selects the largest value in the unknown part of the set.



64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
 
# File 'lib/gecoder/interface/branch.rb', line 64

def branch_on(variables, options = {})
  if variables.respond_to?(:to_int_var) or 
      variables.respond_to?(:to_bool_var) or 
      variables.respond_to?(:to_set_var)
    variables = wrap_enum [variables]
  end

  if variables.respond_to? :to_int_enum 
    add_branch(variables.to_int_enum, options,
      Constants::BRANCH_INT_VAR_CONSTANTS, 
      Constants::BRANCH_INT_VALUE_CONSTANTS)
  elsif variables.respond_to? :to_bool_enum
    add_branch(variables.to_bool_enum, options, 
      Constants::BRANCH_INT_VAR_CONSTANTS, 
      Constants::BRANCH_INT_VALUE_CONSTANTS)
  elsif variables.respond_to? :to_set_enum
    add_branch(variables.to_set_enum, options, 
      Constants::BRANCH_SET_VAR_CONSTANTS, 
      Constants::BRANCH_SET_VALUE_CONSTANTS)
  else
    raise TypeError, "Unknown type of variable enum #{variables.class}."
  end
end

#each_solution(&block) ⇒ Object

Yields each solution that the model has.



46
47
48
49
50
51
52
53
54
55
 
# File 'lib/gecoder/interface/search.rb', line 46

def each_solution(&block)
  dfs = dfs_engine
  next_solution = nil
  while not (next_solution = dfs.next).nil?
    self.active_space = next_solution
    @statistics = dfs.statistics
    yield self
  end
  self.reset!
end

#int_var(domain = LARGEST_INT_DOMAIN) ⇒ Object

Creates a new integer variable with the specified domain. The domain can either be a range, a single element, or an enumeration of elements. If no domain is specified then the largest possible domain is used.



134
135
136
137
 
# File 'lib/gecoder/interface/model.rb', line 134

def int_var(domain = LARGEST_INT_DOMAIN)
  args = domain_arguments(domain)
  IntVar.new(self, variable_creation_space.new_int_var(*args))
end

#int_var_array(count, domain = LARGEST_INT_DOMAIN) ⇒ Object

Creates an array containing the specified number of integer variables with the specified domain. The domain can either be a range, a single element, or an enumeration of elements. If no domain is specified then the largest possible domain is used.



143
144
145
146
147
148
 
# File 'lib/gecoder/interface/model.rb', line 143

def int_var_array(count, domain = LARGEST_INT_DOMAIN)
  args = domain_arguments(domain)
  build_var_array(count) do
    IntVar.new(self, variable_creation_space.new_int_var(*args))
  end
end

#int_var_matrix(row_count, col_count, domain = LARGEST_INT_DOMAIN) ⇒ Object

Creates a matrix containing the specified number rows and columns of integer variables with the specified domain. The domain can either be a range, a single element, or an enumeration of elements. If no domain is specified then the largest possible domain is used.



154
155
156
157
158
159
 
# File 'lib/gecoder/interface/model.rb', line 154

def int_var_matrix(row_count, col_count, domain = LARGEST_INT_DOMAIN)
  args = domain_arguments(domain)
  build_var_matrix(row_count, col_count) do
    IntVar.new(self, variable_creation_space.new_int_var(*args))
  end
end

#maximize!(var) ⇒ Object

Finds the solution that maximizes a given integer variable. The name of the method that accesses the variable from the model should be given. To for instance maximize a variable named “profit”, that’s accessible through the model, one would use the following.

model.maximize! :profit

Raises Gecode::NoSolutionError if no solution can be found.



136
137
138
139
140
141
142
143
144
145
 
# File 'lib/gecoder/interface/search.rb', line 136

def maximize!(var)
  variable = self.method(var).call
  unless variable.kind_of? Gecode::IntVar
    raise ArgumentError.new("Expected integer variable, got #{variable.class}.")
  end
  
  optimize! do |model, best_so_far|
    model.method(var).call.must > best_so_far.method(var).call.value
  end
end

#minimize!(var) ⇒ Object

Finds the solution that minimizes a given integer variable. The name of the method that accesses the variable from the model should be given. To for instance minimize a variable named “cost”, that’s accessible through the model, one would use the following.

model.minimize! :cost

Raises Gecode::NoSolutionError if no solution can be found.



155
156
157
158
159
160
161
162
163
164
 
# File 'lib/gecoder/interface/search.rb', line 155

def minimize!(var)
  variable = self.method(var).call
  unless variable.kind_of? Gecode::IntVar
    raise ArgumentError.new("Expected integer variable, got #{variable.class}.")
  end
  
  optimize! do |model, best_so_far|
    model.method(var).call.must < best_so_far.method(var).call.value
  end
end

#optimize!(&block) ⇒ Object

Finds the optimal solution. Optimality is defined by the provided block which is given one parameter, a solution to the problem. The block should constrain the solution so that that only “better” solutions can be new solutions. For instance if one wants to optimize a variable named price (accessible from the model) to be as low as possible then one should write the following.

model.optimize! do |model, best_so_far|
  model.price.must < best_so_far.price.val
end

Raises Gecode::NoSolutionError if no solution can be found.

Raises:



90
91
92
93
94
95
96
97
98
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
 
# File 'lib/gecoder/interface/search.rb', line 90

def optimize!(&block)
  # Execute constraints.
  perform_queued_gecode_interactions

  # Set the method used for constrain calls by the BAB-search.
  Model.constrain_proc = lambda do |home_space, best_space|
    self.active_space = best_space
    @variable_creation_space = home_space
    yield(self, self)
    self.active_space = home_space
    @variable_creation_space = nil
    
    perform_queued_gecode_interactions
  end

  # Perform the search.
  options = Gecode::Raw::Search::Options.new
  options.c_d = Gecode::Raw::Search::Config::MINIMAL_DISTANCE
  options.a_d = Gecode::Raw::Search::Config::ADAPTIVE_DISTANCE
  options.stop = nil
  bab = Gecode::Raw::BAB.new(selected_space, options)
  
  result = nil
  previous_solution = nil
  until (previous_solution = bab.next).nil?
    result = previous_solution
  end
  @statistics = bab.statistics
  
  # Reset the method used constrain calls and return the result.
  Model.constrain_proc = nil
  raise Gecode::NoSolutionError if result.nil?
  
  # Switch to the result.
  self.active_space = result
  return self
end

#reset! ⇒ Object

Returns to the original state, before any search was made (but propagation might have been performed). Returns the reset model.



25
26
27
28
29
 
# File 'lib/gecoder/interface/search.rb', line 25

def reset!
  self.active_space = base_space
  @statistics = nil
  return self
end

#search_stats ⇒ Object

Returns search statistics providing various information from Gecode about the search that resulted in the model’s current variable state. If the model’s variables have not undergone any search then nil is returned. The statistics is a hash with the following keys:

:propagations

The number of propagation steps performed.

:failures

The number of failed nodes in the search tree.

:clones

The number of clones created.

:commits

The number of commit operations performed.

:memory

The peak memory allocated to Gecode.



66
67
68
69
70
71
72
73
74
75
76
 
# File 'lib/gecoder/interface/search.rb', line 66

def search_stats
  return nil if @statistics.nil?
  
  return {
    :propagations => @statistics.propagate,
    :failures     => @statistics.fail,
    :clones       => @statistics.clone,
    :commits      => @statistics.commit,
    :memory       => @statistics.memory
  }
end

#set_var(glb_domain = [], lub_domain = LARGEST_SET_BOUND, cardinality_range = nil) ⇒ Object

Creates a set variable with the specified domain for greatest lower bound and least upper bound (specified as either a fixnum, range or enum). If no bounds are specified then the empty set is used as greatest lower bound and the largest possible set as least upper bound.

A range for the allowed cardinality of the set can also be specified, if none is specified, or nil is given, then the default range (anything) will be used. If only a single Fixnum is specified as cardinality_range then it’s used as lower bound.



190
191
192
193
194
 
# File 'lib/gecoder/interface/model.rb', line 190

def set_var(glb_domain = [], lub_domain = LARGEST_SET_BOUND,
    cardinality_range = nil)
  args = set_bounds_to_parameters(glb_domain, lub_domain, cardinality_range)
  SetVar.new(self, variable_creation_space.new_set_var(*args))
end

#set_var_array(count, glb_domain = [], lub_domain = LARGEST_SET_BOUND, cardinality_range = nil) ⇒ Object

Creates an array containing the specified number of set variables. The parameters beyond count are the same as for #set_var .



198
199
200
201
202
203
204
 
# File 'lib/gecoder/interface/model.rb', line 198

def set_var_array(count, glb_domain = [], lub_domain = LARGEST_SET_BOUND, 
    cardinality_range = nil)
  args = set_bounds_to_parameters(glb_domain, lub_domain, cardinality_range)
  build_var_array(count) do
    SetVar.new(self, variable_creation_space.new_set_var(*args))
  end
end

#set_var_matrix(row_count, col_count, glb_domain = [], lub_domain = LARGEST_SET_BOUND, cardinality_range = nil) ⇒ Object

Creates a matrix containing the specified number of rows and columns filled with set variables. The parameters beyond row and column counts are the same as for #set_var .



209
210
211
212
213
214
215
 
# File 'lib/gecoder/interface/model.rb', line 209

def set_var_matrix(row_count, col_count, glb_domain = [], 
    lub_domain = LARGEST_SET_BOUND, cardinality_range = nil)
  args = set_bounds_to_parameters(glb_domain, lub_domain, cardinality_range)
  build_var_matrix(row_count, col_count) do
    SetVar.new(self, variable_creation_space.new_set_var(*args))
  end
end

#solution(&block) ⇒ Object

Yields the first solution (if any) to the block. If no solution is found then the block is not used. Returns the result of the block (nil in case the block wasn’t run).



34
35
36
37
38
39
40
41
42
43
 
# File 'lib/gecoder/interface/search.rb', line 34

def solution(&block)
  begin
    solution = self.solve!
    res = yield solution
    self.reset!
    return res
  rescue Gecode::NoSolutionError
    return nil
  end
end

#solve! ⇒ Object

Finds the first solution to the modelled problem and updates the variables to that solution. The found solution is also returned. Raises Gecode::NoSolutionError if no solution can be found.

Raises:



14
15
16
17
18
19
20
21
 
# File 'lib/gecoder/interface/search.rb', line 14

def solve!
  dfs = dfs_engine
  space = dfs.next
  @statistics = dfs.statistics
  raise Gecode::NoSolutionError if space.nil?
  self.active_space = space
  return self
end

#track_variable(variable) ⇒ Object

Starts tracking a variable that depends on the space. All variables created should call this method for their respective models.



265
266
267
 
# File 'lib/gecoder/interface/model.rb', line 265

def track_variable(variable) #:nodoc:
  (@variables ||= []) << variable
end

#wrap_enum(enum) ⇒ Object

Wraps a custom enumerable so that constraints can be specified using it. The argument is altered and returned.



5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
 
# File 'lib/gecoder/interface/enum_wrapper.rb', line 5

def wrap_enum(enum)
  unless enum.kind_of? Enumerable
    raise TypeError, 'Only enumerables can be wrapped.'
  end
  if enum.kind_of? Gecode::EnumMethods
    raise ArgumentError, 'The enumration is already wrapped.'
  end
  elements = enum.to_a
  if elements.empty?
    raise ArgumentError, 'Enumerable must not be empty.'
  end
  
  if elements.all?{ |var| var.respond_to? :to_int_var }
    elements.map!{ |var| var.to_int_var }
    class <<enum
      include Gecode::IntEnumMethods
    end
  elsif elements.all?{ |var| var.respond_to? :to_bool_var }
    elements.map!{ |var| var.to_bool_var }
    class <<enum
      include Gecode::BoolEnumMethods
    end
  elsif elements.all?{ |var| var.respond_to? :to_set_var }
    elements.map!{ |var| var.to_set_var }
    class <<enum
      include Gecode::SetEnumMethods
    end
  elsif elements.all?{ |var| var.kind_of? Fixnum }
    class <<enum
      include Gecode::FixnumEnumMethods
    end
  else
    raise TypeError, "Enumerable doesn't contain operands #{enum.inspect}."
  end
  
  enum.model = self
  return enum
end