I need to create a data model for routing data between points. For example, path between two cities which will include the two start and destination cities, way points (like smaller towns along the way), total distance between the two cities (distances between routing points do not need to be modelled though it would be good if that can be done too).
My current idea is like below.
- Create one table containing the cities:
city
. Columns:city_id
(primary key),city_name
, etc. - Create a many-to-many table
path
with 4 columns:from_city
(referencescity.city_id
,to_city
(referencescity.city_id
), calculated columnpath
=MD5(CONCAT(from_city, to_city))
,distance
to store the total distance for the route. - Create a table to store route points called
town
similar to thecity
table. - Create a man-to-many table
route
betweenpath
andtown
which will have three columns:town_id
(referencestown.town_id
),path_id
(referencespath.path
) androute_position
which will be a value 1-n which will indicate the position of the route point in the actual path. For example, first route point will have number 1, second 2 and so on.
Questions:
- Do you guys think this is a feasible model? Is there a better approach to this?
- In the
path
table, is it better to use aMD5
hash or create a multiple-column index onfrom_city
andto_city
?
EDIT:
To give some context, I'm currently using a graph database to store this data and have a user-defined function that uses A* search to find the shortest path between points. Problem is as the graph gets denser, this computation becomes too slow. So, I think a lookup-based approach would be faster.
path
table, is it? 3) It's possible more than 1 path "from CityA to CityB" - how You'll store them?