Implement graph IO in gml format

This includes a spirit parser for the gml syntax, and a graph writer.

Additionaly, the general graph io code also been cleaned up.

doc/quickstart.rst

@@ -388,13 +388,16 @@ type):
 Graph I/O
 ---------
 
-Graphs can be saved and loaded in two formats: `graphml
-<http://graphml.graphdrawing.org/>`_ and `dot
-<http://www.graphviz.org/doc/info/lang.html>`_. Graphml is the default and
-preferred format. The dot format is also supported, but since it contains no
-type information, all properties are read later as strings, and must be
-converted per hand. Therefore you should always use graphml, except when
-interfacing with another software which expects dot format.
+Graphs can be saved and loaded in three formats: `graphml
+<http://graphml.graphdrawing.org/>`_, `dot
+<http://www.graphviz.org/doc/info/lang.html>`_ and
+`gml <http://www.fim.uni-passau.de/en/fim/faculty/chairs/theoretische-informatik/projects.html>_`.
+``Graphml`` is the default and preferred format. The ``dot`` and ``gml``
+formats are fully supported, but since they contains no precise type
+information, all properties are read later as strings (or also as
+double, in the case of ``gml``), and must be converted per
+hand. Therefore you should always use graphml, except when interfacing
+with another software which expects ``dot`` or ``gml`` format.
 
 A graph can be saved or loaded to a file with the :attr:`~graph_tool.Graph.save`
 and :attr:`~graph_tool.Graph.load` methods, which take either a file name or a

src/boost-workaround/boost/graph/detail/read_graphviz_new.hpp

@@ -86,7 +86,7 @@ namespace read_graphviz_detail {
   };
 
   // The actual parser, from libs/graph/src/read_graphviz_new.cpp
-  void parse_graphviz_from_string(const std::string& str, parser_result& result, bool want_directed);
+  void parse_graphviz_from_string(const std::string& str, parser_result& result, int want_directed);
 
   // Translate from those results to a graph
   void translate_results_to_graph(const parser_result& r, ::boost::detail::graph::mutate_graph* mg);
@@ -96,23 +96,27 @@ namespace read_graphviz_detail {
 // This is also in boost/graph/graphviz.hpp
 namespace detail {
   namespace graph {
-    BOOST_GRAPH_DECL bool read_graphviz(const std::string& str, boost::detail::graph::mutate_graph* mg);
+    BOOST_GRAPH_DECL bool read_graphviz(const std::string& str, boost::detail::graph::mutate_graph* mg,
+                                        bool ignore_directedness);
   } // end namespace graph
 } // end namespace detail
 
 template <typename MutableGraph>
 bool read_graphviz(const std::string& str,
                    MutableGraph& graph, boost::dynamic_properties& dp,
-                   std::string const& node_id = "node_id") {
+                   std::string const& node_id = "node_id",
+                   bool ignore_directedness = false) {
   boost::detail::graph::mutate_graph_impl<MutableGraph> mg(graph, dp, node_id);
-  return detail::graph::read_graphviz(str, &mg);
+  return detail::graph::read_graphviz(str, &mg, ignore_directedness);
 }
 
 template <typename InputIter, typename MutableGraph>
 bool read_graphviz(InputIter begin, InputIter end,
                    MutableGraph& graph, boost::dynamic_properties& dp,
-                   std::string const& node_id = "node_id") {
-  return read_graphviz(std::string(begin, end), graph, dp, node_id);
+                   std::string const& node_id = "node_id",
+                   bool ignore_directedness = false) {
+   return read_graphviz(std::string(begin, end), graph, dp, node_id,
+                        ignore_directedness);
 }
 
 } // namespace boost

src/boost-workaround/boost/graph/graphml.hpp

@@ -49,7 +49,9 @@ class mutate_graph
 {
 public:
     virtual ~mutate_graph() {}
-    virtual bool is_directed() const = 0;
+    virtual int is_directed() const = 0;
+    virtual void flip_directed(bool) = 0;
+    virtual bool get_directed() const = 0;
 
     virtual boost::any do_add_vertex() = 0;
     virtual std::pair<boost::any,bool> do_add_edge(boost::any source,
@@ -79,16 +81,31 @@ class mutate_graph_impl : public mutate_graph
         edge_descriptor;
 
 public:
-    mutate_graph_impl(MutableGraph& g, dynamic_properties& dp)
-        : m_g(g), m_dp(dp) { }
+    mutate_graph_impl(MutableGraph& g, dynamic_properties& dp,
+                      bool ignore_directedness)
+        : m_g(g), m_dp(dp), m_ignore_directedness(ignore_directedness),
+          m_is_directed(false) { }
 
-    bool is_directed() const
+    virtual int is_directed() const
     {
+        if (m_ignore_directedness)
+            return 2;
         return is_convertible
             <typename graph_traits<MutableGraph>::directed_category,
             directed_tag>::value;
     }
 
+    virtual void flip_directed(bool directed)
+    {
+        if (!m_is_directed)
+            m_is_directed = directed;
+    }
+
+    virtual bool get_directed() const
+    {
+        return m_is_directed;
+    }
+
     virtual any do_add_vertex()
     {
         return any(add_vertex(m_g));
@@ -228,6 +245,8 @@ public:
 protected:
     MutableGraph& m_g;
     dynamic_properties& m_dp;
+    bool m_ignore_directedness;
+    bool m_is_directed;
     typedef mpl::vector<uint8_t, int32_t, int64_t, double, long double,
                         std::vector<uint8_t>, std::vector<int32_t>,
                         std::vector<int64_t>, std::vector<double>,
@@ -247,11 +266,13 @@ void
 read_graphml(std::istream& in, mutate_graph& g, bool store_ids);
 
 template<typename MutableGraph>
-void
-read_graphml(std::istream& in, MutableGraph& g, dynamic_properties& dp, bool store_ids)
+bool
+read_graphml(std::istream& in, MutableGraph& g, dynamic_properties& dp,
+             bool store_ids, bool ignore_directedness)
 {
-    mutate_graph_impl<MutableGraph> mg(g,dp);
+    mutate_graph_impl<MutableGraph> mg(g,dp,ignore_directedness);
     read_graphml(in, mg, store_ids);
+    return mg.get_directed();
 }
 
 template <typename Types>

src/boost-workaround/boost/graph/graphviz.hpp

@@ -799,14 +799,15 @@ bool read_graphviz(std::istream& in, mutate_graph& graph);
 template <typename MutableGraph>
 bool read_graphviz(std::istream& in, MutableGraph& graph,
                    dynamic_properties& dp,
-                   std::string const& node_id = "node_id")
+                   std::string const& node_id = "node_id",
+                   bool ignore_directedness = false)
 {
   std::string data;
   in >> std::noskipws;
   std::copy(std::istream_iterator<char>(in),
             std::istream_iterator<char>(),
             std::back_inserter(data));
-  return read_graphviz(data,graph,dp,node_id);
+  return read_graphviz(data,graph,dp,node_id,ignore_directedness);
 }
 
 } // namespace boost

src/graph/Makefile.am

@@ -45,6 +45,7 @@ libgraph_tool_core_la_includedir = $(pythondir)/graph_tool/include
 libgraph_tool_core_la_include_HEADERS = \
     ../../config.h \
     fast_vector_property_map.hh \
+    gml.hh \
     graph.hh \
     graph_adaptor.hh \
     graph_exceptions.hh \

src/graph/gml.hh

+#include <boost/config/warning_disable.hpp>
+#include <boost/spirit/include/qi.hpp>
+#include <boost/spirit/include/phoenix_core.hpp>
+#include <boost/spirit/include/phoenix_operator.hpp>
+#include <boost/spirit/include/phoenix_fusion.hpp>
+#include <boost/spirit/include/phoenix_stl.hpp>
+#include <boost/fusion/include/adapt_struct.hpp>
+#include <boost/variant/recursive_variant.hpp>
+#include <boost/spirit/include/support_istream_iterator.hpp>
+#include <boost/foreach.hpp>
+#include <boost/type_traits.hpp>
+
+#include <iostream>
+#include <fstream>
+#include <string>
+#include <vector>
+
+namespace graph_tool{
+
+using namespace std;
+using namespace boost;
+
+class gml_parse_error: public std::exception
+{
+public:
+    gml_parse_error(const string& w): _what(w) {}
+    ~gml_parse_error() throw() {}
+    std::string what() {return _what;}
+    
+private:
+    std::string _what;
+};
+
+
+template <class Graph>
+class gml_state
+{
+public:
+    gml_state(Graph& g, dynamic_properties& dp)
+        : _g(g), _dp(dp), _directed(false) {}
+
+    typedef boost::variant<std::string, int, double> val_t;
+    
+    // key / value mechanics
+    void push_key(const std::string& key)
+    {
+        _stack.push_back(make_pair(key, prop_list_t()));
+    }
+
+    void push_value(const val_t& value)
+    {
+        if (_stack.empty())
+            return;
+        std::string k = _stack.back().first;
+        _stack.pop_back();
+        if (!_stack.empty())
+            _stack.back().second[k] = value;
+    }
+
+    // actual parsing
+    void finish_list()
+    {
+        if (_stack.empty())
+            return;
+
+        std::string &k = _stack.back().first;
+        if (k == "node")
+        {
+            int id;
+            if (_stack.back().second.find("id") == _stack.back().second.end())
+                throw gml_parse_error("node does not have an id");
+            try
+            {
+                id = get<double>(_stack.back().second["id"]);
+            }
+            catch (bad_get)
+            {
+                throw gml_parse_error("invalid node id");    
+            }
+
+            typename graph_traits<Graph>::vertex_descriptor v = get_vertex(id);
+
+            // put properties
+            for (typeof(_stack.back().second.begin()) iter = _stack.back().second.begin();
+                 iter != _stack.back().second.end(); ++iter)
+            {
+                if (iter->first == "id")
+                    continue;
+                try
+                {    
+                    put(iter->first, _dp, v, get<string>(iter->second));
+                }
+                catch (bad_get)
+                {
+                    put(iter->first, _dp, v, get<double>(iter->second));
+                }
+            }
+        }
+        else if (k == "edge")
+        {
+            int source, target;
+            if (_stack.back().second.find("source") ==
+                _stack.back().second.end() ||
+                _stack.back().second.find("target") ==
+                _stack.back().second.end())
+                throw gml_parse_error("edge does not have source and target ids");
+            try
+            {
+                source = get<double>(_stack.back().second["source"]);
+                target = get<double>(_stack.back().second["target"]);
+            }
+            catch (bad_get)
+            {
+                throw gml_parse_error("invalid source and target ids");
+            }
+
+            typename graph_traits<Graph>::vertex_descriptor s, t;
+            s = get_vertex(source);
+            t = get_vertex(target);
+
+            typename graph_traits<Graph>::edge_descriptor e =
+                add_edge(s, t, _g).first;
+
+            // put properties
+            for (typeof(_stack.back().second.begin()) iter = _stack.back().second.begin();
+                 iter != _stack.back().second.end(); ++iter)
+            {
+                if (iter->first == "id" || iter->first == "source" || iter->first == "target")
+                    continue;
+                try
+                {    
+                    put(iter->first, _dp, e, get<string>(iter->second));
+                }
+                catch (bad_get)
+                {
+                    put(iter->first, _dp, e, get<double>(iter->second));
+                }
+            }
+            
+        }
+        else if (k == "graph")
+        {
+            // put properties
+            for (typeof(_stack.back().second.begin()) iter = _stack.back().second.begin();
+                 iter != _stack.back().second.end(); ++iter)
+            {
+                if (iter->first == "directed")
+                    _directed = get<double>(iter->second);
+                try
+                {    
+                    put(iter->first, _dp, graph_property_tag(), get<string>(iter->second));
+                }
+                catch (bad_get)
+                {
+                    put(iter->first, _dp, graph_property_tag(), get<double>(iter->second));
+                }
+            }
+            
+        }
+        _stack.pop_back();
+    }
+
+    typename graph_traits<Graph>::vertex_descriptor get_vertex(size_t index)
+    {
+        if (_vmap.find(index) == _vmap.end())
+            _vmap[index] = add_vertex(_g);
+        return _vmap[index];
+    }
+    
+
+    bool is_directed() 
+    {
+        return _directed;
+    }
+    
+
+private:
+    Graph& _g;
+    dynamic_properties& _dp;
+    bool _directed;
+    unordered_map<int, typename graph_traits<Graph>::vertex_descriptor> _vmap;
+
+    // the stack holds the keys, and its properties (but omits nested lists)
+    typedef unordered_map<std::string, val_t> prop_list_t;
+    vector<pair<std::string,  prop_list_t> > _stack;
+};
+
+
+template <class Iterator, class Graph, class Skipper>
+struct gml : spirit::qi::grammar<Iterator, void(), Skipper>
+{
+    gml(Graph& g, dynamic_properties& dp) : gml::base_type(start), _state(g, dp)
+    {
+        using namespace spirit;
+        using spirit::ascii::char_;
+
+        unesc_str = spirit::lexeme['"' >> *(unesc_char | (spirit::qi::char_ - "\"") | "\\x" >> qi::hex) >> '"'];
+        unesc_char.add("\\a", '\a')("\\b", '\b')("\\f", '\f')("\\n", '\n')
+            ("\\r", '\r')("\\t", '\t')("\\v", '\v')("\\\\", '\\')
+            ("\\\'", '\'')("\\\"", '\"');
+        key_identifier %= spirit::lexeme[((+spirit::qi::alnum) >> *spirit::qi::alnum)];
+        key = key_identifier
+            [boost::bind(&gml_state<Graph>::push_key, &_state, ::_1)];
+        value_identifier %= (spirit::lexeme[spirit::qi::double_] | unesc_str);
+        value %= value_identifier
+            [boost::bind(&gml_state<Graph>::push_value, &_state, ::_1)];
+        list_identifier = *(key >> (value | "[" >> list >> "]"));
+        list = list_identifier
+            [boost::bind(&gml_state<Graph>::finish_list, &_state)];
+        start = list;
+    }
+
+    typedef boost::variant<std::string, double> val_t;
+
+    spirit::qi::rule<Iterator, std::string(), Skipper> unesc_str;
+    spirit::qi::symbols<char const, char const> unesc_char;
+    spirit::qi::rule<Iterator, std::string(), Skipper> key, key_identifier;
+    spirit::qi::rule<Iterator, val_t(), Skipper> value, value_identifier;
+    spirit::qi::rule<Iterator, void(), Skipper> list, list_identifier;
+    spirit::qi::rule<Iterator, void(), Skipper> start;
+    
+    gml_state<Graph> _state;
+};
+
+template <class Iterator, class Graph, class Skipper>
+bool parse_grammar(Iterator begin, Iterator end, Graph& g,
+                   dynamic_properties& dp, Skipper skip)
+{
+    using namespace spirit;
+    gml<spirit::istream_iterator, Graph, Skipper> parser(g, dp);
+    bool ok = qi::phrase_parse(begin, end, parser, skip);
+    if (!ok)
+        throw gml_parse_error("invalid syntax");
+    return parser._state.is_directed();
+}
+
+
+template <class Graph>
+bool read_gml(istream& in, Graph& g, dynamic_properties& dp)
+{
+    using namespace spirit;
+
+    in >> std::noskipws;
+    spirit::istream_iterator begin(in);
+    spirit::istream_iterator end;
+
+    bool directed =
+        parse_grammar(begin, end, g, dp,
+                      (ascii::space |'#' >> *(ascii::char_ - qi::eol) >> qi::eol));
+
+    in >> std::noskipws;
+    std::stringstream input;
+    input << in.rdbuf();
+    return directed;
+}
+
+struct get_str
+{
+    template <typename ValueType>
+    void operator()(const boost::any& val, std::string& sval, ValueType) const
+    {
+        try 
+        {
+            ValueType v = any_cast<ValueType>(val);
+            if (is_same<ValueType, python::object>::value)
+            {
+                sval = lexical_cast<string>(v);
+            }
+            else
+            {
+                stringstream s;
+                s << v;
+                sval = s.str();
+            }
+
+            if (!is_scalar<ValueType>::value)
+            {
+                replace_all(sval, "\"", "\\\"");
+                sval = "\"" + sval + "\"";
+            }            
+        }
+        catch (bad_any_cast)
+        {
+        }
+    }
+};
+
+template <typename ValueTypes, typename Descriptor>
+std::string print_val(dynamic_property_map& pmap, const Descriptor& v)
+{
+    std::string val;
+    boost::any oval = pmap.get(v);
+    mpl::for_each<ValueTypes>(bind<void>(get_str(), ref(oval),
+                                         ref(val), _1));
+    return val;
+}
+
+
+template <typename Graph, typename VertexIndexMap>
+void write_gml(std::ostream& out, const Graph& g, VertexIndexMap vertex_index,
+               const dynamic_properties& dp)
+{
+    typedef typename graph_traits<Graph>::directed_category directed_category;
+    typedef typename graph_traits<Graph>::edge_descriptor edge_descriptor;
+    typedef typename graph_traits<Graph>::vertex_descriptor vertex_descriptor;
+
+    typedef mpl::vector<bool, uint8_t, int8_t, uint32_t, int32_t,
+                        uint64_t, int64_t, float, double, long double,
+                        std::vector<uint8_t>, std::vector<int32_t>,
+                        std::vector<int64_t>, std::vector<double>,
+                        std::vector<long double>, std::vector<std::string>,
+                        std::string, python::object> value_types;
+
+    BOOST_STATIC_CONSTANT(bool, graph_is_directed =
+                          (is_convertible<directed_category*,
+                                          directed_tag*>::value));
+
+    out << "graph [" << endl;
+
+    if (graph_is_directed)
+        out << "   directed " << 1 << endl;
+
+    for (dynamic_properties::const_iterator i = dp.begin(); i != dp.end();
+         ++i)
+    {
+        if (i->second->key() == typeid(graph_property_tag))
+        {
+            std::string val = print_val<value_types>(*i->second,
+                                                     graph_property_tag());
+            if (val.empty())
+                continue;
+            out << "   " << i->first << " " << val << endl;
+        }
+    }
+
+    typedef typename graph_traits<Graph>::vertex_iterator vertex_iterator;
+    vertex_iterator v, v_end;
+    for (tie(v, v_end) = vertices(g); v != v_end; ++v)
+    {
+        out << "   node [" << endl;
+        out << "      id " << get(vertex_index, *v) << endl;
+
+        for (dynamic_properties::const_iterator i = dp.begin(); i != dp.end();
+             ++i)
+        {
+            if (i->second->key() == typeid(vertex_descriptor))
+            {
+                std::string val = print_val<value_types>(*i->second, *v);
+                if (val.empty())
+                    continue;
+                out << "      " << i->first << " " << val << endl;
+            }
+        }
+        out << "   ]" << endl;
+    }
+
+    typedef typename graph_traits<Graph>::edge_iterator edge_iterator;
+    edge_iterator e, e_end;
+    typename graph_traits<Graph>::edges_size_type edge_count = 0;
+    for (tie(e, e_end) = edges(g); e != e_end; ++e)
+    {
+        out << "   edge [" << endl;
+        out << "      id " << edge_count++ << endl;
+        out << "      source " << get(vertex_index, source(*e, g)) << endl;
+        out << "      target " << get(vertex_index, target(*e, g)) << endl;
+
+        for (dynamic_properties::const_iterator i = dp.begin(); i != dp.end();
+             ++i)
+        {
+            if (i->second->key() == typeid(edge_descriptor))
+            {
+                std::string val = print_val<value_types>(*i->second, *e);
+                if (val.empty())
+                    continue;
+                out << "      " << i->first << " " << val << endl;
+            }
+        }
+        out << "   ]" << endl;
+    }
+    out << "]" << endl;
+}
+
+
+template <typename Graph>
+void write_gml(std::ostream& out, const Graph& g, const dynamic_properties& dp)
+{
+    write_gml(out, g, get(vertex_index, g), dp);
+}
+
+} // namespace graph_tool

src/graph/graph_io.cc

@@ -35,6 +35,8 @@
 #include <boost/lexical_cast.hpp>
 #include <boost/xpressive/xpressive.hpp>
 
+#include "gml.hh"
+
 #include "graph_python_interface.hh"
 #include "str_repr.hh"
 
@@ -290,6 +292,24 @@ class graph_traits<GraphEdgeIndexWrap<Graph,EdgeIndexMap> >
     : public graph_traits<Graph> {};
 }
 
+template <class Graph, class EdgeIndexMap>
+inline
+typename graph_traits
+    <GraphEdgeIndexWrap<Graph,EdgeIndexMap> >::vertex_descriptor
+num_vertices(GraphEdgeIndexWrap<Graph,EdgeIndexMap>& g)
+{
+    return num_vertices(g._g);
+}
+
+template <class Graph, class EdgeIndexMap>
+inline
+typename graph_traits
+    <GraphEdgeIndexWrap<Graph,EdgeIndexMap> >::vertex_descriptor
+vertex(size_t i, GraphEdgeIndexWrap<Graph,EdgeIndexMap>& g)
+{
+    return vertex(i, g._g);
+}
+
 // this graph wraps an UndirectedAdaptor, but overrides the underlying
 // edge_descriptor type with the original type. This will make the edge property
 // maps compatible with the original graph, but will break some things which
@@ -319,6 +339,7 @@ struct FakeEdgeIterator:
 
 };
 
+
 namespace boost {
 template <class Graph>
 struct graph_traits<FakeUndirGraph<Graph> >
@@ -367,7 +388,7 @@ void build_stream
 python::tuple GraphInterface::ReadFromFile(string file, python::object pfile,
                                            string format)
 {
-    if (format != "dot" && format != "xml" && format != "auto")
+    if (format != "dot" && format != "xml" && format != "gml")
         throw ValueException("error reading from file '" + file +
                              "': requested invalid format '" + format + "'");
     try
@@ -384,31 +405,13 @@ python::tuple GraphInterface::ReadFromFile(string file, python::object pfile,