OpenRouter Network builder #3 was: Re: [SoC] Mapbender CSW - Week 3

Stephen Woodbridge woodbri at swoodbridge.com
Mon Jun 15 10:25:41 EDT 2009


Hi Wolf, Ashraf,

You can view the page from the same url by changing https to http. Here 
is a link to Ashraf's status reports:
http://sourceforge.net/apps/trac/opengraphrouter/wiki/WeeklySoCStatusReports

With regards to caching the graph, one of the things that I did in C on 
a past project was to create the graph structure in memory that was 
mmap'ed to a file. This made it possible to prepare the graph and save 
it and later allowed solver processes to mmap the graph into memory 
using the system paging. This was very fast and effective. That said, it 
is a lot tricker to do this in C++, and I believe it is still possible 
to achieve the same results and benefits in C++, but we thought it best 
to leave this outside of SoC as Ashraf plans to continue working on the 
project.

Ashraf,

You will need to remember to edit the url when you post to the list.

Thanks,
   -Steve

Ashraf Hossain wrote:
> Hi Wolf,
> I am very very sorry for my unwanted mistake of the subject.
> 
> The docs will be publicly available very soon.
> 
> Generally a poly line feature is treated as a edge and its start and
> end point as vertex(as shown in the figure).
> In case of the edge no 2 the middle point is not taken as graph input
> as this is not a starting or ending node.
> In case of the point connected edge no 1 and 3 is taken as vertex
> though in figure it seems that is should not taken as vertex.
> But for every feature in the polyline shape there is a database row,
> that may be contain many fields.
> 
> Like Length,traveltime, routetime, traffic jam factors and many others.
> We may add 2 lengths but how can we determine the traffic jam factors
> of the total additions or other fields.
> 
> So only the intersection point of the two feature(s) are treated as
> vertex entry in graph.
> 
> And yes there are lots of ways to cache graph files.
> 
> But SoC is for three months, So I want to complete a library which
> will be memory based but functional and usable.
> 
> I will start work on caching and other technique after GSoC tasks complete.
> 
> Thank you again for this comments.
> 
> Regards
> Ashraf Roni
> 
> 
> 
> 
> 
> 
> 
> 
> On Mon, Jun 15, 2009 at 12:44 PM, Wolf
> Bergenheim<wolf+grass at bergenheim.net> wrote:
>> Hi Ashraf,
>>
>> Thanks for your report. Two requests though. Please don't hijack other
>> students' threads! (i.e. don't hit the reply button [or at the very
>> least change the subject!]). And the second is that the page with the
>> link requires authorisation. Please make the report pages publicly
>> available. Thanks!
>>
>> As for the vertexes, how about making only intersections into nodes,
>> or will that still result in a graph that is too huge? Is there a way
>> that the graph could be cached to the disk somehow?
>>
>> Regards,
>> Wolf Bergenheim
>> OSGeo GSoc Administrator
>>
>> On Sat, Jun 13, 2009 at 01:36, Ashraf Hossain<ashraf10bd at gmail.com> wrote:
>>> Hi,
>>>
>>> I was thinking about the navigation service and accurate routing and
>>> got this method(in the link) is the best so far.
>>> Because we can not allow every point in a polyline feature as vertex.
>>> If we allow then the size of the graph will be huge.
>>> So if we think about virtual vertex and virtual edge then the problem
>>> will be solved.
>>> The virtual vertex and virtual edges will not be added with the main
>>> vertex and edge list in the graph.
>>> This will be calculated locally in the shortest path algorithm functions.
>>>
>>> I will be glad if any one has better idea regarding this.
>>>
>>> The report link is below with 2 attached figure.
>>>
>>> https://sourceforge.net/apps/trac/opengraphrouter/wiki/OpenRouter_Network_builder_report_20090612
>>>
>>>
>>> Regards
>>> Roni
>>> _______________________________________________
>>> SoC mailing list
>>> SoC at lists.osgeo.org
>>> http://lists.osgeo.org/mailman/listinfo/soc
>>>
>>
>>
>> --
>>
>> <:3 )---- Wolf Bergenheim ----( 8:>
>>
>>
>> ------------------------------------------------------------------------
>>



More information about the SoC mailing list