We haven't gotten started on our own geocoder. In the meantime, we
support pluggable geocoders -- see the opentripplanner-geocoder
directory.
Josh: you can modify the URL of the Nominatim server you want to use in
the application-context.xml file. By default it uses the Mapquest one.
Wojcieh: I am not sure what you mean. The request is always sent as an
address (streetname + house number or POI...) ?
Viv
I think I understand what you mean now and it seems that it comes down
to user interaction and how the modules are plugged together.
As it works currently, the user types in a start address that is
requested to the geocoder module and then he gets a result/coordinates
(or not, in this case he can retype) and then he types the to address
that is requested to the geocoder module and gets a result/coordinates
(or not). Then, the user click on "Plan" and the start and end
coordinates are sent to the routing API.
If I understand correctly, you suggest that the user would type in the
two addresses then click on "plan" to send the request to the routing
API and then, the routing API would request the coordinate from the
geocoder module (nominatim or another plugin) and then it plans the
route directly if there are only one answer for each address or it would
request the user to choose from multiple option if there is an ambiguity
on the addresses? In terms of user interaction, I think it is actually
kind of like Google maps behaves except that there is auto-completion of
addresses that is not there yet in OTP (maybe once there is an
implementation with Apache SOLR).
I am not sure what the best, but since OTP was and is working like
option 1, I just followed that and I added a plugin to the geocoder
module to support Nominatim so the quick answer would be that no I had
not planned to support the option 2 as it involves a bit more changes to
the way OTP currently works and I am not sure how well it would work in
terms of user interaction without auto-completion since Nominatim tends
to often return several results for a request. It also true that there
are a few oddities in the way the input form interact with geocoding
right now but that could be fixed.
What do you think are the advantages of option 2? Had anyone planned to
implement geocoding this way?
On this topic, there are actually two params "from" and "to" that are
sent in the URL and contain the addresses (fromplace and toplace contain
the coordinates) but I am not sure how and when they are used by the API
module.
Viv