[pgrouting-dev] GID vs ID field name

[Stephen Woodbridge]

On 6/21/2012 11:42 PM, Mario Basa wrote:
> Hello,
>
> Is there any particular reason why the sql functions expect the unique
> identifier have a "GID" of the table that contains the network data. I
> am asking since some sql functions expect an "ID" unique identifier
> while the functions in the routing_dd_wrapper expects a "GID"
> identifier in which it is just renamed to "ID".
>
> This is a bit confusing for me.
>
> Anyway, will there be any objections if I will just expect that the
> table that contains the network data have a "ID" field name as its
> record unique identifier.

Hi Mario,

I think the origin of this has something to do with the fact that some 
of the algorithms allocate space based on a vector that is more 
efficient if the ID values are contiguous. The assumption was the vendor 
data ID values might be more random and therefore less contiguous than 
GID values which are assigned when the shapefile loader loads the data 
and are contiguous.

This issue is mostly with the Boost based algorithms. I think it is fine 
to make the change you are suggesting, the user should be able to select 
any column as a unique integer identifier and alias it to ID.

-Steve