/* -*- C++ -*-

 *

 * This file is a part of LEMON, a generic C++ optimization library

 *

 * Copyright (C) 2003-2007

 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport

 * (Egervary Research Group on Combinatorial Optimization, EGRES).

 *

 * Permission to use, modify and distribute this software is granted

 * provided that this copyright notice appears in all copies. For

 * precise terms see the accompanying LICENSE file.

 *

 * This software is provided "AS IS" with no warranty of any kind,

 * express or implied, and with no claim as to its suitability for any

 * purpose.

 *

 */

namespace lemon {

/*!

\page graph-io-page Graph Input-Output

The standard graph IO enables one to store graphs and additional maps

(i.e. functions on the nodes or edges) in a flexible and efficient way. 

Before you read this page you should be familiar with LEMON 

\ref graphs "graphs" and \ref maps-page "maps".

\section format The general file format

The file contains sections in the following order:

\li nodeset

\li edgeset

\li nodes

\li edges

\li attributes

Some of these sections can be omitted, but you will basicly need the nodeset

section (unless your graph has no nodes at all) and the edgeset section

(unless your graph has no edges at all). 

The nodeset section describes the nodes of your graph: it identifies the nodes

and gives the maps defined on them, if any. It starts with the

following line:

<tt>\@nodeset</tt>

The next line contains the names of the nodemaps, separated by whitespaces.  Each

following line describes a node in the graph: it contains the values of the

maps in the right order. The map named "label" should contain unique values

because it is regarded as a label map. These labels need not be numbers but they

must identify the nodes uniquely for later reference. For example:

\code

@nodeset

label  x-coord  y-coord  color

3   1.0      4.0      blue

5   2.3      5.7      red

12  7.8      2.3      green

\endcode

The edgeset section is very similar to the nodeset section, it has

the same coloumn oriented structure. It starts with the line 

<tt>\@edgeset</tt>

The next line contains the whitespace separated list of names of the edge

maps.  Each of the next lines describes one edge. The first two elements in

the line are the labels of the source and target (or tail and head) nodes of the

edge as they occur in the label node map of the nodeset section. You can also

have an optional label map on the edges for later reference (which has to be

unique in this case).

\code

@edgeset

             label      weight   note

3   5        a          4.3      a-edge

5   12       c          2.6      c-edge

3   12       g          3.4      g-edge

\endcode

The \e nodes section contains <em>labeled (distinguished) nodes</em> 

(i.e. nodes having a special

label on them). The section starts with

<tt> \@nodes </tt>

Each of the next lines contains a label for a node in the graph 

and then the label as described in the \e nodeset section.

\code

@nodes 

source 3

target 12

\endcode

The last section describes the <em>labeled (distinguished) edges</em>

(i.e. edges having a special label on them). It starts with \c \@edges

and then each line contains the name of the edge and the label.

\code

@edges 

observed c

\endcode

The file may contain empty lines and comment lines. The comment lines

start with an \c # character.

The attributes section can handle some information about the graph. It

contains key-value pairs in each line (a key and the mapped value to key). The

key should be a string without whitespaces, the value can be of various types.

\code

@attributes

title "Four colored planar graph"

author "Balazs DEZSO"

copyright "Lemon Library"

version 12

\endcode

Finally, the file should be closed with \c \@end line.

\section use Using graph input-output

The graph input and output is based on <em> reading and writing

commands</em>. The user gives reading and writing commands to the reader or

writer class, then he calls the \c run() method that executes all the given

commands.

\subsection write Writing a graph

The \ref lemon::GraphWriter "GraphWriter" template class

provides the graph output. To write a graph

you should first give writing commands to the writer. You can declare

writing command as \c NodeMap or \c EdgeMap writing and labeled Node and

Edge writing.

\code

GraphWriter<ListGraph> writer(std::cout, graph);

\endcode

The \ref lemon::GraphWriter::writeNodeMap() "writeNodeMap()"

function declares a \c NodeMap writing command in the

\ref lemon::GraphWriter "GraphWriter".

You should give a name to the map and the map

object as parameters. The NodeMap writing command with name "label" should write a 

unique map because it will be regarded as a label map.

\see IdMap, DescriptorMap  

\code

IdMap<ListGraph, Node> nodeLabelMap;

writer.writeNodeMap("label", nodeLabelMap);

writer.writeNodeMap("x-coord", xCoordMap);

writer.writeNodeMap("y-coord", yCoordMap);

writer.writeNodeMap("color", colorMap);

\endcode

With the \ref lemon::GraphWriter::writeEdgeMap() "writeEdgeMap()"

member function you can give an edge map

writing command similar to the NodeMaps.

\see IdMap, DescriptorMap  

\code

DescriptorMap<ListGraph, Edge, ListGraph::EdgeMap<int> > edgeDescMap(graph);

writer.writeEdgeMap("descriptor", edgeDescMap);

writer.writeEdgeMap("weight", weightMap);

writer.writeEdgeMap("note", noteMap);

\endcode

With \ref lemon::GraphWriter::writeNode() "writeNode()"

and \ref lemon::GraphWriter::writeEdge() "writeEdge()"

functions you can designate Nodes and

Edges in the graph. For example, you can write out the source and target node

of a maximum flow instance.

\code

writer.writeNode("source", sourceNode);

writer.writeNode("target", targetNode);

writer.writeEdge("observed", edge);

\endcode

With \ref lemon::GraphWriter::writeAttribute() "writeAttribute()"

function you can write an attribute to the file.

\code

writer.writeAttribute("author", "Balazs DEZSO");

writer.writeAttribute("version", 12);

\endcode

After you give all write commands you must call the

\ref lemon::GraphWriter::run() "run()" member

function, which executes all the writing commands.

\code

writer.run();

\endcode

\subsection reading Reading a graph

The file to be read may contain several maps and labeled nodes or edges.

If you read a graph you need not read all the maps and items just those

that you need. The interface of the \ref lemon::GraphReader "GraphReader"

is very similar to

the \ref lemon::GraphWriter "GraphWriter"

but the reading method does not depend on the order of the

given commands.

The reader object assumes that each not read value does not contain 

whitespaces, therefore it has some extra possibilities to control how

it should skip the values when the string representation contains spaces.

\code

GraphReader<ListGraph> reader(std::cin, graph);

\endcode

The \ref lemon::GraphReader::readNodeMap() "readNodeMap()"

function reads a map from the \c nodeset section.

If there is a map that you do not want to read from the file and there are

whitespaces in the string represenation of the values then you should

call the \ref lemon::GraphReader::skipNodeMap() "skipNodeMap()"

template member function with proper parameters.

\see QuotedStringReader

\code

reader.readNodeMap("x-coord", xCoordMap);

reader.readNodeMap("y-coord", yCoordMap);

reader.readNodeMap<QuotedStringReader>("label", labelMap);

reader.skipNodeMap<QuotedStringReader>("description");

reader.readNodeMap("color", colorMap);

\endcode

With the \ref lemon::GraphReader::readEdgeMap() "readEdgeMap()"

member function you can give an edge map

reading command similar to the NodeMaps. 

\code

reader.readEdgeMap("weight", weightMap);

reader.readEdgeMap("label", labelMap);

\endcode

With \ref lemon::GraphReader::readNode() "readNode()"

and \ref lemon::GraphReader::readEdge() "readEdge()"

functions you can read labeled Nodes and

Edges.

\code

reader.readNode("source", sourceNode);

reader.readNode("target", targetNode);

reader.readEdge("observed", edge);

\endcode

With \ref lemon::GraphReader::readAttribute() "readAttribute()"

function you can read an attribute from the file.

\code

std::string author;

writer.readAttribute("author", author);

int version;

writer.writeAttribute("version", version);

\endcode

After you give all read commands you must call the

\ref lemon::GraphReader::run() "run()" member

function, which executes all the commands.

\code

reader.run();

\endcode

\anchor rwbackground

\section types Background of Reading and Writing

To read a map (on the nodes or edges)

the \ref lemon::GraphReader "GraphReader"

should know how to read a Value from the given map.

By the default implementation the input operator reads a value from

the stream and the type of the read value is the value type of the given map.

When the reader should skip a value in the stream, because you do not

want to store it in a map, the reader skips a character sequence without 

whitespaces. 

If you want to change the functionality of the reader, you can use

template parameters to specialize it. When you give a reading

command for a map you can give a Reader type as template parameter.

With this template parameter you can control how the Reader reads

a value from the stream.

The reader has the next structure: 

\code

struct TypeReader {

  typedef TypeName Value;

  void read(std::istream& is, Value& value);

};

\endcode

For example, the \c "strings" nodemap contains strings and you do not need

the value of the string just the length. Then you can implement an own Reader

struct.

\code

struct LengthReader {

  typedef int Value;

  void read(std::istream& is, Value& value) {

    std::string tmp;

    is >> tmp;

    value = tmp.length();

  }

};

...

reader.readNodeMap<LengthReader>("strings", lengthMap);

\endcode  

The global functionality of the reader class can be changed by giving a

special template parameter to the GraphReader class. By default, the

template parameter is \c DefaultReaderTraits. A reader traits class 

should provide a nested template class Reader for each type, and a 

DefaultReader for skipping a value.

The specialization of writing is very similar to that of reading.

\section u Undirected graphs

In a file describing an undirected graph (ugraph, for short) you find an

\c uedgeset section instead of the \c edgeset section. The first line of

the section describes the names of the maps on the undirected egdes and all

next lines describe one undirected edge with the the incident nodes and the

values of the map.

The format handles directed edge maps as a syntactical sugar???, if there

are two maps with names being the same with a \c '+' and a \c '-' prefix

then this will be read as a directed map.

\code

@uedgeset

             label      capacity        +flow   -flow

32   2       1          4.3             2.0     0.0

21   21      5          2.6             0.0     2.6

21   12      8          3.4             0.0     0.0

\endcode

The \c edges section is changed to \c uedges section. This section

describes labeled edges and undirected edges. The directed edge label

should start with a \c '+' or a \c '-' prefix to decide the direction

of the edge. 

\code

@uedges

uedge 1

+edge 5

-back 5

\endcode

There are similar classes to the \ref lemon::GraphReader "GraphReader" and

\ref lemon::GraphWriter "GraphWriter" which

handle the undirected graphs. These classes are

the \ref lemon::UGraphReader "UGraphReader"

and \ref lemon::UGraphWriter "UGraphWriter".

The \ref lemon::UGraphReader::readUEdgeMap() "readUEdgeMap()"

function reads an undirected map and the

\ref lemon::UGraphReader::readUEdge() "readUEdge()"

reads an undirected edge from the file, 

\code

reader.readUEdgeMap("capacity", capacityMap);

reader.readEdgeMap("flow", flowMap);

...

reader.readUEdge("u_edge", u_edge);

reader.readEdge("edge", edge);

\endcode

\section advanced Advanced features

The graph reader and writer classes give an easy way to read and write

graphs. But sometimes we want more advanced features. In this case we can

use the more general <tt>lemon reader and writer</tt> interface.

The LEMON file format is a section oriented file format. It contains one or

more sections, each starting with a line identifying its type 

(the word starting with the \c \@  character).

The content of the section this way cannot contain line with \c \@ first

character. The file may contains comment lines with \c # first character.

The \ref lemon::LemonReader "LemonReader"

and \ref lemon::LemonWriter "LemonWriter"

gives a framework to read and

write sections. There are various section reader and section writer

classes which can be attached to a \ref lemon::LemonReader "LemonReader"

or a \ref lemon::LemonWriter "LemonWriter".

There are default section readers and writers for reading and writing

item sets, and labeled items in the graph. These read and write

the format described above. Other type of data can be handled with own

section reader and writer classes which are inherited from the

\c LemonReader::SectionReader or the

\ref lemon::LemonWriter::SectionWriter "LemonWriter::SectionWriter"

classes.

The next example defines a special section reader which reads the

\c \@description sections into a string:

\code 

class DescriptionReader : LemonReader::SectionReader {

protected:

  virtual bool header(const std::string& line) {

    std::istringstream ls(line);

    std::string head;

    ls >> head;

    return head == "@description";

  }

  virtual void read(std::istream& is) {

    std::string line;

    while (getline(is, line)) {

      desc += line;

    }

  }

public:

  typedef LemonReader::SectionReader Parent;

  DescriptionReader(LemonReader& reader) : Parent(reader) {}

  const std::string& description() const {

    return description;

  }

private:

  std::string desc;

};

\endcode

The other advanced stuff of the generalized file format is that 

multiple edgesets can be stored to the same nodeset. It can be used 

for example as a network traffic matrix.

In our example there is a network with symmetric links and there are assymetric

traffic request on the network. This construction can be stored in an

undirected graph and in a directed \c ListEdgeSet class. The example

shows the input with the \ref lemon::LemonReader "LemonReader" class:

\code

ListUGraph network;

ListUGraph::UEdgeMap<double> capacity;

ListEdgeSet<ListUGraph> traffic(network);

ListEdgeSet<ListUGraph>::EdgeMap<double> request(network);

LemonReader reader(std::cin);

NodeSetReader<ListUGraph> nodesetReader(reader, network);

UEdgeSetReader<ListUGraph> 

  uEdgesetReader(reader, network, nodesetReader);

uEdgesetReader.readEdgeMap("capacity", capacity);

EdgeSetReader<ListEdgeSet<ListUGraph> > 

  edgesetReader(reader, traffic, nodesetReader, "traffic");

edgesetReader.readEdgeMap("request", request);

reader.run();

\endcode

Because both the \ref lemon::GraphReader "GraphReader"

and the \ref lemon::UGraphReader "UGraphReader" can be converted

to \ref lemon::LemonReader "LemonReader"

and it can resolve the label's of the items, the previous

result can be achived with the \ref lemon::UGraphReader "UGraphReader"

class, too.

\code

ListUGraph network;

ListUGraph::UEdgeSet<double> capacity;

ListEdgeSet<ListUGraph> traffic(network);

ListEdgeSet<ListUGraph>::EdgeMap<double> request(network);

UGraphReader<ListUGraph> reader(std::cin, network);

reader.readEdgeMap("capacity", capacity);

EdgeSetReader<ListEdgeSet<ListUGraph> > 

  edgesetReader(reader, traffic, reader, "traffic");

edgesetReader.readEdgeMap("request", request);

reader.run();

\endcode

\author Balazs Dezso

*/

}