GML files in MATLAB

Hi All,

I'm looking to start playing around with graphs in MATLAB - and I keep finding data sets in .gml format.

I can read it fine in a text editor - but we're talking thousands of nodes/edges - does anyone know of a program I can use to convert this into a MATLAB friendly format? I just want an adjacency matrix, and I don't know how I would convert it myself.

Here is a link to a small example of the kind of file I'm talking about: http://www-personal.umich.edu/~mejn/netdata/dolphins.zip

thanks.

Neil.

Answer 1

[Walter Roberson]

S = fileread('dolphins.gml');

nodes = regexp(S, 'node.*?id (?<id>\d+).*?label\s*"(?<label>[^"]*)"', 'names');
edges = regexp(S, 'edge.*?source\s*(?<source>\d+).*?target\s*(?<target>\d+)', 'names');

all_ids = {nodes.id};
all_names = {nodes.label};
all_sources = {edges.source};
all_targets = {edges.target};

[source_found, s] = ismember(all_sources, all_ids);
nfidx = find(~source_found);
if ~isempty(nfidx)
   error('Source ids not found in node list, starting from "%s"', edges(nfidx(1).source));
end
[target_found, t] = ismember(all_targets, all_ids);
nfidx = find(~target_found);
if ~isempty(nfidx)
   error('Target ids not found in node list, starting from "%s"', edges(nfidx(1).target));
end

EdgeTable = table([s.', t.'], ones(length(s),1), 'VariableNames', {'EndNodes' 'Weight'});
NodeTable = table(all_names.', 'VariableNames',{'Name'});
G = graph(EdgeTable,NodeTable);
plot(G);

[M Abedi]

thanx alot. this cod work for me. but my graph has longitud and latitude for each node. how i can access them in matlab?

[Walter Roberson]

At the time you do the plot(), pass in 'XData' and 'YData' name/value pairs. You could have previously stored the data into the NodeData property; https://www.mathworks.com/help/matlab/ref/graph.html#bun70n7-NodeTable

See also

https://www.mathworks.com/help/matlab/math/add-node-properties-to-graph-plot-data-cursor.html

[M Abedi]

but my problem is about reading Xdata and Ydata frome gml file.

a part of my gml file is like this:

node [

id 1

label "Brisbane2"

Country "Australia"

Longitude 153.02809

Internal 1

Latitude -27.46794

]

node [

id 2

label "Canberra1"

Country "Australia"

Longitude 149.12807

Internal 1

Latitude -35.28346

]

with the above cod i could read nodes IDs and labels... but dont read nodes Longitudes and Latitudes yet.

[Walter Roberson]

nodes = regexp(S, 'node.*?id (?<id>\d+).*?label\s*"(?<label>[^"]*)"', 'names');

can be extended to parse the lat and long as well using the same .*?KEYWORD\s*(<TAG>\S+) format.

Answer 2

[Mahmoud]

You can use the following naive code to extract the edges from the graph and put them in a matrix which can easily be changed to adjacency matrix.

   %Extracting edges from gml file graph

fileName = 'dolphins.gml';
inputfile = fopen(fileName);
A=[];

l=0;
k=1;
while 1

      % Get a line from the input file
      tline = fgetl(inputfile);

      % Quit if end of file
      if ~ischar(tline)
          break
      end

      nums = regexp(tline,'\d+','match');
      if length(nums)
          if l==1
              l=0;
              A(k,2)=str2num(nums{1});  
              k=k+1;
              continue;
          end
          A(k,1)=str2num(nums{1});
          l=1;
      else
          l=0;
          continue;
      end

end

A[] will be the [m x 2] matrix containing all edges.

Best,

 Mahmoud

[Walter Roberson]

str2double() instead of str2num() would be better programming practice (more secure, faster)

[mary]

sorry I have used this solution but gives these 3 errors: error in fgets,fgetl and tline= fgetl(inputfile); what is the reason? thank you

[Yasaman]

thanks a lot!

[Karen Sachs]

This is great thank you! Do you know how to extract the names of the variables? This tells me the indicies of the variables but not the variables themselves..

[Walter Roberson]

See my (new) answer, which graphs the node ids and labels and edge information, and builds a complete Graph object from it.

https://www.mathworks.com/matlabcentral/answers/15318-gml-files-in-matlab#answer_420804

Answer 3

[Antonio]

function G = importgml(fileName)
inputfile = fopen(fileName);
A=[];
l=0; k=1; while 1
    % Get a line from the input file
    tline = fgetl(inputfile);

    % Quit if end of file
    if ~ischar(tline)
        break
    end

    nums = regexp(tline,'\d+','match'); %get number from string
    if length(nums)
        if l==1
            l=0;
            A(k,2)=str2num(nums{1});
            k=k+1;
            continue;
        end
        A(k,1)=str2num(nums{1});
        l=1;
    else
        l=0;
        continue;
    end
end
G=[];
for i=1:length(A)
    G(A(i,1)+1,A(i,2)+1) = 1;
end

[Antonio]

A slight enhancement which returns the adjacency matrix.

[hossein]

Adjacency matrix must be symmetric.

 for i=1:length(A)
     G(A(i,1)+1,A(i,2)+1) = 1;
     G(A(i,2)+1,A(i,1)+1) = 1;
 end

[Alexander Kocian]

@hossein: this is only true for undirected graphs

Answer 4

[Mohammad Bhat]

How can I convert an Adjacency Matrix into .gml file