gdf2osmnx #545
gdf2osmnx #545
Conversation
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #545 +/- ##
=======================================
- Coverage 97.4% 97.3% -0.1%
=======================================
Files 26 26
Lines 4317 4358 +41
=======================================
+ Hits 4203 4240 +37
- Misses 114 118 +4
|
There was a problem hiding this comment.
Looking very good @anastassiavybornova.
- break each linestring in gdf into two-point linestrings
- for each two-point linestring, find corresponding node IDs (u,v)
I am curious, does this approach handle loops (e.g.,
| blg.loc[ | ||
| blg.loc[blg["mm_uid"] == key].index[0], blg.geometry.name | ||
| ] = new_geom | ||
| blg.loc[blg.loc[blg["mm_uid"] == key].index[0], blg.geometry.name] = ( |
There was a problem hiding this comment.
Is this extra indexing needed due to an update in Pandas/GeoPandas?
There was a problem hiding this comment.
haven't touched that one (except for running black on it) i'll leave the question to @martinfleis
There was a problem hiding this comment.
Did only a partial review as I need to clarify some bits.
What exactly is the difference between the two options? Assuming both are MultiDiGraphs. As far as I can tell, the only difference is that the original uses tuple (x, y) to encode node with no attributes, the osmnx-like encodes nodes to integers and adds x and y as attributes. Is there anything else I don't see?
If I am right, then my suggestion would be to adapt our original function to add {'x': 1, 'y': 2} attributes to each node as osmnx-like version do and if osmnx_like=True, call convert_node_labels_to_integers and be done with it. Am I wrong?
| else: | ||
| graph.add_edge(first, last, **attributes) | ||
| gdf_network["geometry"] = gdf_network["geometry"].apply( | ||
| lambda x: linemerge(x) if x.geom_type == "MultiLineString" else x |
There was a problem hiding this comment.
Why are you doing this? If it is a pre-processing step, than a user should do it themselves. Otherwise this does not round-trip as it should.
There was a problem hiding this comment.
_generate_primal already checks for some potential non-linestring geometries, so i thought i'd add this one as well - but yes it is a pre-processing step indeed. fine to leave it out of the function.
|
Looking into it a bit further, it seems that the only change we need to do is to add the attributes with coordinates to nodes. Adding these two lines results in no issue on a small sample dataset. node_attrs = {node: {"x": node[0], "y": node[1]} for node in graph.nodes}
nx.set_node_attributes(graph, node_attrs) |
Co-authored-by: James Gaboardi <[email protected]>
Co-authored-by: James Gaboardi <[email protected]>
|
@anastassiavybornova maybe have a look at #546 before spending more time on this. |
Added the
osmnx_likekeyword togdf_to_nx(). Ifosmnx_like=True, within_generate_primal:netwith unique dummy IDsgdf_to_nx()now works as input forosmnx.simplify_graph()