Nodes (ways and relations) by ID

[wildwalks]

Hi, thanks for geodesk, it looks amazing.
I have been using osm2postgres -- hoping this will make life easier and faster :)

Getting my head around it with the Python examples.
But I am stuck trying to create an object of nodes, based on their OSM ID. I assumed this would be easy, but I am missing something.
(My plan after that is to get info on each, then query based on a line between them and a buffer.)
Can you point by in the right direction?
thanks

[wildwalks]

So my two findings.
i can find nodes using this function (but I assume there is a better way - like using an index :)
def find_node_by_id(features, target_id):
for node in features.nodes:
if node.id == target_id:
return node
return None

Usage:

target_node = find_node_by_id(features, 2762334883)
if target_node:
print(f"Node ID: {target_node.id}")
print(f"Type: {target_node.osm_type}")
print(f"Coordinates: ({target_node.lat}, {target_node.lon})")
print(f"Tags: {target_node.tags}")

But using this, I can find node - 7733744933
target_node = find_node_by_id(features, 7733744933)
if target_node:
print(f"Node ID: {target_node.id}")
print(f"Type: {target_node.osm_type}")
print(f"Coordinates: ({target_node.lat}, {target_node.lon})")
print(f"Tags: {target_node.tags}")

This node has no tags. It is used to form a way. I get why it might not be imported by default, but is there an option to force them to be included?

thanks again

[clarisma]

Hi, thanks for opening this discussion!

Unfortunately, there's currently no way to quickly query features solely based on their ID. (We're working on this for the next major version, because this is a common complaint). See also clarisma/geodesk-py#24

So for now, you would need to perform a linear search over the set of nodes (i.e. your approach above) -- but of course that is quite slow. One way to improve performance significantly is to narrow down the search space -- either spatially or via a GOQL query (e.g. if you know your node is in the Sydney area, you can use a bbox query).

However, this still will not give you anonymous nodes (nodes that have no tags, whose whole purpose is to provide the geometry of ways), since these are not stored as features.

Instead, you could query the ways of interest (e.g. w[highway=path,footway]), then iterate over the nodes of each way (for node in way.nodes), which is very fast.

If you want to create a search envelope (e.g. to determine what's within 50 meters of a path), you can also turn the way into a Shapely geometry (way.shape), then apply a buffer around that shape. You can then use the resulting polygon to issue a GeoDesk query (e.g. world.nodes("n[tourism=information][information=guidepost]")(buffered_shape) to get any nearby guideposts). For buffering, keep in mind that all geometries returned by GeoDesk use Mercator units -- you can use to_mercator(m=50, lat=-33) to convert 50 meters into the equivalent number of units (which you can then pass to Shapely's buffer() function) at -33 degrees latitude.

Do you have a specific use case where you will have to be able to identify untagged nodes by their ID?

[wildwalks]

Ohh thanks -- great to see it is on the road map, sorry I should have seen that issue, missed it.
Ohh yes, narrowing the list to a set based on an area, then iterating will absolutely work for me in the meantime -- good tiip.
Yes, I read the project stuff, makes sense, but still got it wrong the first time I tried -- hehe :)

At this stage I am just playing with it to get a sense of its capability. But I have a bushwalking (hiking) site that I am using and OSM DB and a set of queries to get details for specific tracks. It works well, but the queries are cumbersome, difficult to maintain and take a lot of computing resources. I had started working on a solution using spatialite to tile the data I need to speed things up and make it more scalable. Then I found geodesk and was blown away with the work you have done. Amazing. It is crazy fast, especially for area-based searches (this is a real resource killer for me on PostgreSQL, yes indexed). Geodesk looks like it solves my issues, and way better than my dodgy plan :)
The only things at this stage not out of the box are
Looking items up by ID
finding parallel items (Like crossing, but 90 ish-degrees out - eg is there a creek on left or right of the track)
getting coordinates of nodes (and I assume ways) that have no tags (I assume they are not included in the import - at least it seems I can find them)
And the other features I would need are best handled in shapely/geopandas (although I am still learning).

Why the untaged Nodes?
Currently i am use a rouing tool to create the initial walking route, that gives me a list of OSM node ID's and lats and longs. But the lats and longs are simplified, so I get the geoms ID from my DB. In Geodesk my plan was to to create an object of the nodes, then create a line, buffer etc. I can work around it okay, it is not a show stopper for me - I think I just assumed with all the filters that ID would be one :)
I suspect I could probably write a function that gets a nodes point, by finding a parent way and matches the node id (Have not tried it, but maybe the data is there). I suspect other people wanting to work with parts of ways will have a similar issue. Hope that makes sense. Thanks again, I am really impressed with geodesk, it makes OSM data so much easier to interact with.

[wildwalks]

Okay -- So it seems I should have read the instructions ;)
https://docs.geodesk.com/gol/build
Perfect, and updateable soon too -- you made my day -- thanks :)