jrtechs-vis/lib/network/dotparser.js

/**
 * Parse a text source containing data in DOT language into a JSON object.
 * The object contains two lists: one with nodes and one with edges.
 *
 * DOT language reference: http://www.graphviz.org/doc/info/lang.html
 *
 * @param {String} data     Text containing a graph in DOT-notation
 * @return {Object} graph   An object containing two parameters:
 *                          {Object[]} nodes
 *                          {Object[]} edges
 */
function parseDOT (data) {
  dot = data;
  return parseGraph();
}

// token types enumeration
var TOKENTYPE = {
  NULL : 0,
  DELIMITER : 1,
  IDENTIFIER: 2,
  UNKNOWN : 3
};

// map with all delimiters
var DELIMITERS = {
  '{': true,
  '}': true,
  '[': true,
  ']': true,
  ';': true,
  '=': true,
  ',': true,

  '->': true,
  '--': true
};

var dot = '';                   // current dot file
var index = 0;                  // current index in dot file
var c = '';                     // current token character in expr
var token = '';                 // current token
var tokenType = TOKENTYPE.NULL; // type of the token

/**
 * Get the first character from the dot file.
 * The character is stored into the char c. If the end of the dot file is
 * reached, the function puts an empty string in c.
 */
function first() {
  index = 0;
  c = dot.charAt(0);
}

/**
 * Get the next character from the dot file.
 * The character is stored into the char c. If the end of the dot file is
 * reached, the function puts an empty string in c.
 */
function next() {
  index++;
  c = dot.charAt(index);
}

/**
 * Preview the next character from the dot file.
 * @return {String} cNext
 */
function nextPreview() {
  return dot.charAt(index + 1);
}

/**
 * Test whether given character is alphabetic or numeric
 * @param {String} c
 * @return {Boolean} isAlphaNumeric
 */
var regexAlphaNumeric = /[a-zA-Z_0-9.:#]/;
function isAlphaNumeric(c) {
  return regexAlphaNumeric.test(c);
}

/**
 * Merge all options of object b into object b
 * @param {Object} a
 * @param {Object} b
 * @return {Object} a
 */
function merge (a, b) {
  if (!a) {
    a = {};
  }

  if (b) {
    for (var name in b) {
      if (b.hasOwnProperty(name)) {
        a[name] = b[name];
      }
    }
  }
  return a;
}

/**
 * Set a value in an object, where the provided parameter name can be a
 * path with nested parameters. For example:
 *
 *     var obj = {a: 2};
 *     setValue(obj, 'b.c', 3);     // obj = {a: 2, b: {c: 3}}
 *
 * @param {Object} obj
 * @param {String} path  A parameter name or dot-separated parameter path,
 *                      like "color.highlight.border".
 * @param {*} value
 */
function setValue(obj, path, value) {
  var keys = path.split('.');
  var o = obj;
  while (keys.length) {
    var key = keys.shift();
    if (keys.length) {
      // this isn't the end point
      if (!o[key]) {
        o[key] = {};
      }
      o = o[key];
    }
    else {
      // this is the end point
      o[key] = value;
    }
  }
}

/**
 * Add a node to a graph object. If there is already a node with
 * the same id, their attributes will be merged.
 * @param {Object} graph
 * @param {Object} node
 */
function addNode(graph, node) {
  var i, len;
  var current = null;

  // find root graph (in case of subgraph)
  var graphs = [graph]; // list with all graphs from current graph to root graph
  var root = graph;
  while (root.parent) {
    graphs.push(root.parent);
    root = root.parent;
  }

  // find existing node (at root level) by its id
  if (root.nodes) {
    for (i = 0, len = root.nodes.length; i < len; i++) {
      if (node.id === root.nodes[i].id) {
        current = root.nodes[i];
        break;
      }
    }
  }

  if (!current) {
    // this is a new node
    current = {
      id: node.id
    };
    if (graph.node) {
      // clone default attributes
      current.attr = merge(current.attr, graph.node);
    }
  }

  // add node to this (sub)graph and all its parent graphs
  for (i = graphs.length - 1; i >= 0; i--) {
    var g = graphs[i];

    if (!g.nodes) {
      g.nodes = [];
    }
    if (g.nodes.indexOf(current) == -1) {
      g.nodes.push(current);
    }
  }

  // merge attributes
  if (node.attr) {
    current.attr = merge(current.attr, node.attr);
  }
}

/**
 * Add an edge to a graph object
 * @param {Object} graph
 * @param {Object} edge
 */
function addEdge(graph, edge) {
  if (!graph.edges) {
    graph.edges = [];
  }
  graph.edges.push(edge);
  if (graph.edge) {
    var attr = merge({}, graph.edge);     // clone default attributes
    edge.attr = merge(attr, edge.attr); // merge attributes
  }
}

/**
 * Create an edge to a graph object
 * @param {Object} graph
 * @param {String | Number | Object} from
 * @param {String | Number | Object} to
 * @param {String} type
 * @param {Object | null} attr
 * @return {Object} edge
 */
function createEdge(graph, from, to, type, attr) {
  var edge = {
    from: from,
    to: to,
    type: type
  };

  if (graph.edge) {
    edge.attr = merge({}, graph.edge);  // clone default attributes
  }
  edge.attr = merge(edge.attr || {}, attr); // merge attributes

  return edge;
}

/**
 * Get next token in the current dot file.
 * The token and token type are available as token and tokenType
 */
function getToken() {
  tokenType = TOKENTYPE.NULL;
  token = '';

  // skip over whitespaces
  while (c == ' ' || c == '\t' || c == '\n' || c == '\r') {  // space, tab, enter
    next();
  }

  do {
    var isComment = false;

    // skip comment
    if (c == '#') {
      // find the previous non-space character
      var i = index - 1;
      while (dot.charAt(i) == ' ' || dot.charAt(i) == '\t') {
        i--;
      }
      if (dot.charAt(i) == '\n' || dot.charAt(i) == '') {
        // the # is at the start of a line, this is indeed a line comment
        while (c != '' && c != '\n') {
          next();
        }
        isComment = true;
      }
    }
    if (c == '/' && nextPreview() == '/') {
      // skip line comment
      while (c != '' && c != '\n') {
        next();
      }
      isComment = true;
    }
    if (c == '/' && nextPreview() == '*') {
      // skip block comment
      while (c != '') {
        if (c == '*' && nextPreview() == '/') {
          // end of block comment found. skip these last two characters
          next();
          next();
          break;
        }
        else {
          next();
        }
      }
      isComment = true;
    }

    // skip over whitespaces
    while (c == ' ' || c == '\t' || c == '\n' || c == '\r') {  // space, tab, enter
      next();
    }
  }
  while (isComment);

  // check for end of dot file
  if (c == '') {
    // token is still empty
    tokenType = TOKENTYPE.DELIMITER;
    return;
  }

  // check for delimiters consisting of 2 characters
  var c2 = c + nextPreview();
  if (DELIMITERS[c2]) {
    tokenType = TOKENTYPE.DELIMITER;
    token = c2;
    next();
    next();
    return;
  }

  // check for delimiters consisting of 1 character
  if (DELIMITERS[c]) {
    tokenType = TOKENTYPE.DELIMITER;
    token = c;
    next();
    return;
  }

  // check for an identifier (number or string)
  // TODO: more precise parsing of numbers/strings (and the port separator ':')
  if (isAlphaNumeric(c) || c == '-') {
    token += c;
    next();

    while (isAlphaNumeric(c)) {
      token += c;
      next();
    }
    if (token == 'false') {
      token = false;   // convert to boolean
    }
    else if (token == 'true') {
      token = true;   // convert to boolean
    }
    else if (!isNaN(Number(token))) {
      token = Number(token); // convert to number
    }
    tokenType = TOKENTYPE.IDENTIFIER;
    return;
  }

  // check for a string enclosed by double quotes
  if (c == '"') {
    next();
    while (c != '' && (c != '"' || (c == '"' && nextPreview() == '"'))) {
      token += c;
      if (c == '"') { // skip the escape character
        next();
      }
      next();
    }
    if (c != '"') {
      throw newSyntaxError('End of string " expected');
    }
    next();
    tokenType = TOKENTYPE.IDENTIFIER;
    return;
  }

  // something unknown is found, wrong characters, a syntax error
  tokenType = TOKENTYPE.UNKNOWN;
  while (c != '') {
    token += c;
    next();
  }
  throw new SyntaxError('Syntax error in part "' + chop(token, 30) + '"');
}

/**
 * Parse a graph.
 * @returns {Object} graph
 */
function parseGraph() {
  var graph = {};

  first();
  getToken();

  // optional strict keyword
  if (token == 'strict') {
    graph.strict = true;
    getToken();
  }

  // graph or digraph keyword
  if (token == 'graph' || token == 'digraph') {
    graph.type = token;
    getToken();
  }

  // optional graph id
  if (tokenType == TOKENTYPE.IDENTIFIER) {
    graph.id = token;
    getToken();
  }

  // open angle bracket
  if (token != '{') {
    throw newSyntaxError('Angle bracket { expected');
  }
  getToken();

  // statements
  parseStatements(graph);

  // close angle bracket
  if (token != '}') {
    throw newSyntaxError('Angle bracket } expected');
  }
  getToken();

  // end of file
  if (token !== '') {
    throw newSyntaxError('End of file expected');
  }
  getToken();

  // remove temporary default options
  delete graph.node;
  delete graph.edge;
  delete graph.graph;

  return graph;
}

/**
 * Parse a list with statements.
 * @param {Object} graph
 */
function parseStatements (graph) {
  while (token !== '' && token != '}') {
    parseStatement(graph);
    if (token == ';') {
      getToken();
    }
  }
}

/**
 * Parse a single statement. Can be a an attribute statement, node
 * statement, a series of node statements and edge statements, or a
 * parameter.
 * @param {Object} graph
 */
function parseStatement(graph) {
  // parse subgraph
  var subgraph = parseSubgraph(graph);
  if (subgraph) {
    // edge statements
    parseEdge(graph, subgraph);

    return;
  }

  // parse an attribute statement
  var attr = parseAttributeStatement(graph);
  if (attr) {
    return;
  }

  // parse node
  if (tokenType != TOKENTYPE.IDENTIFIER) {
    throw newSyntaxError('Identifier expected');
  }
  var id = token; // id can be a string or a number
  getToken();

  if (token == '=') {
    // id statement
    getToken();
    if (tokenType != TOKENTYPE.IDENTIFIER) {
      throw newSyntaxError('Identifier expected');
    }
    graph[id] = token;
    getToken();
    // TODO: implement comma separated list with "a_list: ID=ID [','] [a_list] "
  }
  else {
    parseNodeStatement(graph, id);
  }
}

/**
 * Parse a subgraph
 * @param {Object} graph    parent graph object
 * @return {Object | null} subgraph
 */
function parseSubgraph (graph) {
  var subgraph = null;

  // optional subgraph keyword
  if (token == 'subgraph') {
    subgraph = {};
    subgraph.type = 'subgraph';
    getToken();

    // optional graph id
    if (tokenType == TOKENTYPE.IDENTIFIER) {
      subgraph.id = token;
      getToken();
    }
  }

  // open angle bracket
  if (token == '{') {
    getToken();

    if (!subgraph) {
      subgraph = {};
    }
    subgraph.parent = graph;
    subgraph.node = graph.node;
    subgraph.edge = graph.edge;
    subgraph.graph = graph.graph;

    // statements
    parseStatements(subgraph);

    // close angle bracket
    if (token != '}') {
      throw newSyntaxError('Angle bracket } expected');
    }
    getToken();

    // remove temporary default options
    delete subgraph.node;
    delete subgraph.edge;
    delete subgraph.graph;
    delete subgraph.parent;

    // register at the parent graph
    if (!graph.subgraphs) {
      graph.subgraphs = [];
    }
    graph.subgraphs.push(subgraph);
  }

  return subgraph;
}

/**
 * parse an attribute statement like "node [shape=circle fontSize=16]".
 * Available keywords are 'node', 'edge', 'graph'.
 * The previous list with default attributes will be replaced
 * @param {Object} graph
 * @returns {String | null} keyword Returns the name of the parsed attribute
 *                                  (node, edge, graph), or null if nothing
 *                                  is parsed.
 */
function parseAttributeStatement (graph) {
  // attribute statements
  if (token == 'node') {
    getToken();

    // node attributes
    graph.node = parseAttributeList();
    return 'node';
  }
  else if (token == 'edge') {
    getToken();

    // edge attributes
    graph.edge = parseAttributeList();
    return 'edge';
  }
  else if (token == 'graph') {
    getToken();

    // graph attributes
    graph.graph = parseAttributeList();
    return 'graph';
  }

  return null;
}

/**
 * parse a node statement
 * @param {Object} graph
 * @param {String | Number} id
 */
function parseNodeStatement(graph, id) {
  // node statement
  var node = {
    id: id
  };
  var attr = parseAttributeList();
  if (attr) {
    node.attr = attr;
  }
  addNode(graph, node);

  // edge statements
  parseEdge(graph, id);
}

/**
 * Parse an edge or a series of edges
 * @param {Object} graph
 * @param {String | Number} from        Id of the from node
 */
function parseEdge(graph, from) {
  while (token == '->' || token == '--') {
    var to;
    var type = token;
    getToken();

    var subgraph = parseSubgraph(graph);
    if (subgraph) {
      to = subgraph;
    }
    else {
      if (tokenType != TOKENTYPE.IDENTIFIER) {
        throw newSyntaxError('Identifier or subgraph expected');
      }
      to = token;
      addNode(graph, {
        id: to
      });
      getToken();
    }

    // parse edge attributes
    var attr = parseAttributeList();

    // create edge
    var edge = createEdge(graph, from, to, type, attr);
    addEdge(graph, edge);

    from = to;
  }
}

/**
 * Parse a set with attributes,
 * for example [label="1.000", shape=solid]
 * @return {Object | null} attr
 */
function parseAttributeList() {
  var attr = null;

  while (token == '[') {
    getToken();
    attr = {};
    while (token !== '' && token != ']') {
      if (tokenType != TOKENTYPE.IDENTIFIER) {
        throw newSyntaxError('Attribute name expected');
      }
      var name = token;

      getToken();
      if (token != '=') {
        throw newSyntaxError('Equal sign = expected');
      }
      getToken();

      if (tokenType != TOKENTYPE.IDENTIFIER) {
        throw newSyntaxError('Attribute value expected');
      }
      var value = token;
      setValue(attr, name, value); // name can be a path

      getToken();
      if (token ==',') {
        getToken();
      }
    }

    if (token != ']') {
      throw newSyntaxError('Bracket ] expected');
    }
    getToken();
  }

  return attr;
}

/**
 * Create a syntax error with extra information on current token and index.
 * @param {String} message
 * @returns {SyntaxError} err
 */
function newSyntaxError(message) {
  return new SyntaxError(message + ', got "' + chop(token, 30) + '" (char ' + index + ')');
}

/**
 * Chop off text after a maximum length
 * @param {String} text
 * @param {Number} maxLength
 * @returns {String}
 */
function chop (text, maxLength) {
  return (text.length <= maxLength) ? text : (text.substr(0, 27) + '...');
}

/**
 * Execute a function fn for each pair of elements in two arrays
 * @param {Array | *} array1
 * @param {Array | *} array2
 * @param {function} fn
 */
function forEach2(array1, array2, fn) {
  if (Array.isArray(array1)) {
    array1.forEach(function (elem1) {
      if (Array.isArray(array2)) {
        array2.forEach(function (elem2)  {
          fn(elem1, elem2);
        });
      }
      else {
        fn(elem1, array2);
      }
    });
  }
  else {
    if (Array.isArray(array2)) {
      array2.forEach(function (elem2)  {
        fn(array1, elem2);
      });
    }
    else {
      fn(array1, array2);
    }
  }
}

/**
 * Convert a string containing a graph in DOT language into a map containing
 * with nodes and edges in the format of graph.
 * @param {String} data         Text containing a graph in DOT-notation
 * @return {Object} graphData
 */
function DOTToGraph (data) {
  // parse the DOT file
  var dotData = parseDOT(data);
  var graphData = {
    nodes: [],
    edges: [],
    options: {}
  };

  // copy the nodes
  if (dotData.nodes) {
    dotData.nodes.forEach(function (dotNode) {
      var graphNode = {
        id: dotNode.id,
        label: String(dotNode.label || dotNode.id)
      };
      merge(graphNode, dotNode.attr);
      if (graphNode.image) {
        graphNode.shape = 'image';
      }
      graphData.nodes.push(graphNode);
    });
  }

  // copy the edges
  if (dotData.edges) {
    /**
     * Convert an edge in DOT format to an edge with VisGraph format
     * @param {Object} dotEdge
     * @returns {Object} graphEdge
     */
    var convertEdge = function (dotEdge) {
      var graphEdge = {
        from: dotEdge.from,
        to: dotEdge.to
      };
      merge(graphEdge, dotEdge.attr);
      graphEdge.style = (dotEdge.type == '->') ? 'arrow' : 'line';
      return graphEdge;
    }

    dotData.edges.forEach(function (dotEdge) {
      var from, to;
      if (dotEdge.from instanceof Object) {
        from = dotEdge.from.nodes;
      }
      else {
        from = {
          id: dotEdge.from
        }
      }

      if (dotEdge.to instanceof Object) {
        to = dotEdge.to.nodes;
      }
      else {
        to = {
          id: dotEdge.to
        }
      }

      if (dotEdge.from instanceof Object && dotEdge.from.edges) {
        dotEdge.from.edges.forEach(function (subEdge) {
          var graphEdge = convertEdge(subEdge);
          graphData.edges.push(graphEdge);
        });
      }

      forEach2(from, to, function (from, to) {
        var subEdge = createEdge(graphData, from.id, to.id, dotEdge.type, dotEdge.attr);
        var graphEdge = convertEdge(subEdge);
        graphData.edges.push(graphEdge);
      });

      if (dotEdge.to instanceof Object && dotEdge.to.edges) {
        dotEdge.to.edges.forEach(function (subEdge) {
          var graphEdge = convertEdge(subEdge);
          graphData.edges.push(graphEdge);
        });
      }
    });
  }

  // copy the options
  if (dotData.attr) {
    graphData.options = dotData.attr;
  }

  return graphData;
}

// exports
exports.parseDOT = parseDOT;
exports.DOTToGraph = DOTToGraph;