Major improvements in NetworkSimplex.
Main changes:
- Use -potenital[] instead of potential[] to conform to the usual
terminology.
- Use function parameter instead of #define commands to select pivot rule.
- Use much faster implementation for the candidate list pivot rule.
It is about 5-20 times faster now.
- Add a new pivot rule called "Limited Search" that is a modified
version of "Block Search". It is about 25 percent faster on rather
sparse graphs.
- By default "Limited Search" is used for sparse graphs and
"Block Search" is used otherwise. This combined method is the most
efficient on every input class.
- Change the name of private members to start with "_".
- Change the name of function parameters not to start with "_".
- Remove unnecessary documentation for private members.
- Many doc improvements.
3 * This file is a part of LEMON, a generic C++ optimization library
5 * Copyright (C) 2003-2008
6 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 * (Egervary Research Group on Combinatorial Optimization, EGRES).
9 * Permission to use, modify and distribute this software is granted
10 * provided that this copyright notice appears in all copies. For
11 * precise terms see the accompanying LICENSE file.
13 * This software is provided "AS IS" with no warranty of any kind,
14 * express or implied, and with no claim as to its suitability for any
21 \page named-param Named Parameters
23 \section named-func-param Named Function Parameters
25 C++ makes it possible to use default parameter values when calling a
26 function. In such a case we do not have to give value for parameters,
27 the program will use the default ones. Unfortunately sometimes this
28 is not enough. If we do not want to give values for all the
29 parameters, only for some of them we come across problems, because an
30 arbitrary set of parameters cannot be omitted. On the other hand
31 parameters have a fixed order in the head of the function. C++ can
32 apply the default values only in the back of the order, if we do not
33 give other value for them. So we can not give the function for
34 example the value of the first, and the third parameter, expecting
35 that the program will aplly the default value for the second
36 parameter. However sometimes we would like to use some functinos
37 exactly in this way. With a crafty trick and with some little
38 inconvenience this is possible. We have implemented this little trick
49 namedFn() : _id(0), _val(1), _dim(2) {}
50 namedFn& id(int p) { _id = p ; return *this; }
51 namedFn& val(double p) { _val = p ; return *this; }
52 namedFn& dim(int p) { _dim = p ; return *this; }
55 printf("Here is the function itself.");
61 The usage is the following.
63 We have to define a class, let's call it \c namedFn. Let us assume that
64 we would like to use a parameter, called \c X. In the \c namedFn class we
65 have to define an \c _X attribute, and a function \c X. The function
66 expects a parameter with the type of \c _X, and sets the value of
67 \c _X. After setting the value the function returns the class itself. The
68 class also have to have a function, called for example <tt>run()</tt>, we have
69 to implement here the original function itself. The constructor of the
70 class have to give all the attributes like \c _X the default values of
73 If we instantiate this class, the default values will be set for the
74 attributes (originally the parameters), initially. If we call function
75 \c X, we get a class with the modified parameter value of
76 \c X. Therefore we can modify any parameter-value, independently from the
77 order. To run the algorithm we have to call the <tt>run()</tt> function at the
82 namedFn().id(3).val(2).run();
85 \note Although it is a class, namedFn is used pretty much like as it were
86 a function. That it why it is called namedFn and not \c NamedFn.
88 \note In fact, the final <tt>.run()</tt> could be made unnecessary if the
89 actual function code were put in the destructor instead. This however would make
90 hard to implement functions with return values, and would also make the
91 implementation of \ref named-templ-func-param "named template parameters"
92 very problematic. <b>Therefore, by convention, <tt>.run()</tt> must be used
93 to explicitly execute function having named parameters in Lemon.</b>
96 \section traits-classes Traits Classes
98 The procedure above can also be applied when defining classes. In this
99 case the type of the attributes can be changed. Initially we have to
100 define a class with the default attribute types. This is the so called
101 Traits Class. Later on the types of these attributes can be changed,
102 as described below. In our software \ref lemon::DijkstraDefaultTraits is an
103 example of how a traits class looks like.
105 \section named-templ-param Named Class Template Parameters
107 If we would like to change the type of an attribute in a class that
108 was instantiated by using a traits class as a template parameter, and
109 the class contains named parameters, we do not have to reinstantiate
110 the class with new traits class. Instead of that, adaptor classes can
111 be used like in the following cases.
114 Dijkstra<>::SetPredNodeMap<NullMap<Node,Node> >::Create
117 It can also be used in conjunction with other named template
118 parameters in arbitrary order.
121 Dijkstra<>::SetDistMap<MyMap>::SetPredMap<NullMap<Node,Edge> >::Create
124 The result will be an instantiated Dijkstra class, in which the
125 DistMap and the PredMap is modified.
127 \section named-templ-func-param Named Function Template Parameters
129 If the class has so called wizard functions, the new class with the
130 modified tpye of attributes can be returned by the appropriate wizard
131 function. The usage of these wizard functions is the following: