Hello Pankaj Malhotra,

Yes, I need to assign more than values to the edge. I need the edge has [flow, 
capacity, distinction, source] 
You says I need to create a new class type for edge, How can I pass it to json 
class and how can I make json class read my new edge object.

I need to submit my code next week, so any answer or recommendation will 
definitely help.

thanks
Rana

On Mar 24, 2014, at 7:29 AM, Pankaj Malhotra <pankajiit...@gmail.com> wrote:

> If I get your point-1 right, you want to have a vertex class type other than 
> the ones available, for example: Text. If this is what you want, you can have 
> a class having as many fields as you want. Then convert the objects of that 
> class to Json objects using Gson for example. Then, you can use the Json 
> string to set the fields of the Text class.
> 
> 
> On 22 March 2014 04:50, Rana Althunyan <ranoo...@gmail.com> wrote:
> Hello experts,
> > I really need your help. I am new to Giraph. I have project on pregel 
> > paper, and I need to implement One algorithm via Apache Giraph.The 
> > algorithm is push-relabel flow algorithm. Actually when I started to adapt 
> > the algorithm to fit to Giraph I found some problems. I will list them.
> >
> > 1- Because any type of data structure are not allowed in Giraph how can I 
> > store excess flow and distance value for each vertex.
> > 2- When the excess value becomes zero for any vertex, I wanna keep vertex 
> > flow  during future supersteps even if the vertex received messages.
> > 3- The last point, when I want to send a flow to a vertex, i need store two 
> > values of residual graph  the remaining value from the edge capacity and 
> > the amount of flow that we use so far.  How can I do that. How can I keep 
> > these two values for each vertex in the graph.
> >
> > these following questions may are easy for you ..
> > 4- How can I determine the last vertex as a sink.
> > 5- How can I get the number of all vertexes from the input file .
> >
> > I know I have written many question.I searched a lot in the Internet, but I 
> > did not find any discussion regards to this algorithm
> >
> > sorry and thank you.I appreciate any help or recommendations
> > Rana
> 
> 

Reply via email to