[pgrouting-users] tsp_dijkstra

pgRouting pgrouting-users
1

I am trying to use tsp_dijkstra solution but it turned out that this postgresql function contains a table name called tsp_test. I did not see any query that creates even a temporary table named tsp_test. So what is the use of this function ? Is there a specific need for a temporary table ?

/emre

2

Hi Emre,

Oh my! Thank you for reporting this. It must be fixed.
I don't remember now what I meant by this, so please give me few days to fix it.

Sorry,
Anton.

3

Hi Anton,

There is also something that I did not understand. We are using both cost and reverse_cost in tsp_dijkstra_directed when we call dijkstra. but we only define reverse_cost after calling tsp in the query

query := ‘SELECT vertex_id FROM tsp(’'select distinct source::integer as source_id, '||
‘x(startpoint(the_geom)), y(startpoint(the_geom))’;

IF rc THEN query := query || ’ , reverse_cost ';

why do we use only reverse_cost and not the cost ?

/emre

On Wed, Jan 26, 2011 at 5:08 AM, Anton Patrushev <anton.patrushev@georepublic.de> wrote:

Hi Emre,

Oh my! Thank you for reporting this. It must be fixed.
I don’t remember now what I meant by this, so please give me few days to fix it.

Sorry,
Anton.

Pgrouting-users mailing list
Pgrouting-users@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/pgrouting-users

4

Hi Emre,

Yes, it looks wrong. For current TSP implementation we use Euclidean
distance, which means need only vertex IDs, x and y. No cost or
reverse_cost are involved.
I will fix it as soon as I can.

Anton.

5

Hi Anton,

So using tsp_dijkstra_directed function does not actually use dijkstra for directed cost calculation right ? I know you probably have things to do but I need to have a tsp implementation with dijkstra. Since I dont know much about developing on pgrouting its very hard for me to solve the issue. So I would be more than happy if we can solve this issue in a short time.

Thanks a lot,

/emre

On Wed, Jan 26, 2011 at 11:47 AM, Anton Patrushev <anton.patrushev@georepublic.de> wrote:

Hi Emre,

Yes, it looks wrong. For current TSP implementation we use Euclidean
distance, which means need only vertex IDs, x and y. No cost or
reverse_cost are involved.
I will fix it as soon as I can.

Anton.

Pgrouting-users mailing list
Pgrouting-users@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/pgrouting-users