Issue when importing a network instead of using it within the function

[csebastiao]

Describe the bug
Sometimes when using the main growbikenet function, I receive a KeyError when importing a street network, even if it is the same as the one used in the function. This is not systematically the case, sometimes it works and sometimes it doesn't, with no obvious pattern.

To Reproduce
If I do:

edges_ranked = gbn.growbikenet(
"Asti",
ranking="betweenness_centrality",
export_data=False,
)

It works, but if instead I do:

file_out = "gdf_osm.gpkg"
city_boundary_gdf = ox.geocoder.geocode_to_gdf("Asti")
city_boundary_geometry = city_boundary_gdf.geometry[0]
G = ox.graph_from_polygon(city_boundary_geometry, network_type="drive")
ox.io.save_graph_geopackage(G, file_out)
edges_ranked = gbn.growbikenet(
"Asti",
ranking="betweenness_centrality",
export_data=False,
import_files={"street_network": file_out},
)

I receive a KeyError.

Expected behavior
I should not receive the KeyError when I import the street network.

Desktop (please complete the following information):

• OS: MacOS Version 15.7.7

[mszell]

Hi Clement, thx!
I think this is a bug that was just introduced in a PR today.
Which growbikenet version are you using?
print(gbn.__version__)
I suspect if you just use version 0.11.1 installed from conda-forge, you wouldn't encounter this problem. Let me know.

[mszell]

In any case, if I use the latest version from the main branch, I get this error:

==============================================
RUNNING GROWBIKENET FOR CITY: Asti
betweenness_centrality | auto | from scratch
----------------------------------------------╮
Importing network data : 100%|████████████████| 1/1 [00:01<00:00,  1.11s/network]
Creating seed points   : 100%|████████████████| 3/3 [00:01<00:00,  1.97step/s]
Triangulation          : 100%|████████████████| 1/1 [00:00<00:00, 11.00step/s]
Routing                :  33%|█████▎          | 1/3 [00:00<00:01,  1.37step/s]Traceback (most recent call last):
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/examples/development.py", line 16, in <module>
    edges_ranked = gbn.growbikenet(
    "Asti",
    ...<2 lines>...
    import_files={"street_network": file_out},
    )
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/growbikenet/growbikenet.py", line 307, in growbikenet
    grown_bikenet_edges = create_gdf_with_geoms(grown_bikenet_edges_abstract, edges)
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/growbikenet/functions.py", line 1149, in create_gdf_with_geoms
    df["geometry"] = df.path_edges.apply(lambda x: MultiLineString(list(edges.loc[x].geometry)))
                     ~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/series.py", line 5084, in apply
    ).apply()
      ~~~~~^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/apply.py", line 1520, in apply
    return self.apply_standard()
           ~~~~~~~~~~~~~~~~~~~^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/apply.py", line 1578, in apply_standard
    mapped = obj._map_values(mapper=curried)
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/base.py", line 1022, in _map_values
    return algorithms.map_array(arr, mapper, na_action=na_action)
           ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/algorithms.py", line 1715, in map_array
    return lib.map_infer(values, mapper)
           ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "pandas/_libs/lib.pyx", line 3071, in pandas._libs.lib.map_infer
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/growbikenet/functions.py", line 1149, in <lambda>
    df["geometry"] = df.path_edges.apply(lambda x: MultiLineString(list(edges.loc[x].geometry)))
                                                                        ~~~~~~~~~^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexing.py", line 1207, in __getitem__
    return self._getitem_axis(maybe_callable, axis=axis)
           ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexing.py", line 1438, in _getitem_axis
    return self._getitem_iterable(key, axis=axis)
           ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexing.py", line 1378, in _getitem_iterable
    keyarr, indexer = self._get_listlike_indexer(key, axis)
                      ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexing.py", line 1576, in _get_listlike_indexer
    keyarr, indexer = ax._get_indexer_strict(key, axis_name)
                      ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexes/multi.py", line 3242, in _get_indexer_strict
    return super()._get_indexer_strict(key, axis_name)
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexes/base.py", line 6302, in _get_indexer_strict
    self._raise_if_missing(keyarr, indexer, axis_name)
    ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexes/multi.py", line 3262, in _raise_if_missing
    return super()._raise_if_missing(key, indexer, axis_name)
           ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mszell/Github/BikeNetKit/GrowBikeNet/.pixi/envs/default/lib/python3.14/site-packages/pandas/core/indexes/base.py", line 6355, in _raise_if_missing
    raise KeyError(f"{not_found} not in index")
KeyError: '[(8390353889, 8390353883)] not in index'
Routing                :  33%|█████▎          | 1/3 [00:00<00:01,  1.19step/s]

@marianamirandapessoa , the problem is the line

df["geometry"] = df.path_edges.apply(lambda x: MultiLineString(list(edges.loc[x].geometry)))

which was just changed in your PR #205 . So I suspect, while this PR fixed some previous bug, it also introduced a new one. Could you investigate?

[csebastiao]

Hi Clement, thx! I think this is a bug that was just introduced in a PR today. Which growbikenet version are you using? print(gbn.__version__) I suspect if you just use version 0.11.1 installed from conda-forge, you wouldn't encounter this problem. Let me know.

I am using 0.11.1 installed via conda-forge. The bug was exactly at 1/3 on routing like you, in this example and some other cities.

[mszell]

Thx!

Note for devs: Apart from investigating and fixing, we should use offline Asti as an additional future test case in test_main.

[mszell]

Ok I just added the Asti test #211 , which currently fails but should succeed once the bug is fixed. (Note one has to use "Asti 16" as city name to get the city and not the big municipality.)
@marianamirandapessoa , whenever you can look at it, @Manuel-Knepper might be able to also help since he wrote the original function.

[marianamirandapessoa]

Hi!

I started an initial investigation, that turned into a little bit of a rabbit hole 😅.

First, it seems that the problem is at this step: edges.loc[x].geometry, which was already in the function before the changes made by PR #205.

I checked by running the second code snippet Clement shared (where the network is saved and then directly imported to growbikenet) with the code before and after the changes made, and the KeyError appears in both. The fact that it is a KeyError, seems to confirm that the error is in the .loc[].

Looking at the error output from the added Asti test #211, it seems that for some reason the edge (4079292976, 4079293055) is not exported to the geopackage.

I ran the two approaches:

Approach 1	Approach 2
The street network is extracted within the growbikenet function	The street network is extracted outside the growbikenet function and then imported
Code to generate this approach:	Code to generate this approach (like the test added in #211):

The only difference between the street networks of the two approaches is this specific edge:

To understand why this edge could be a problem, I looked for it in OSM directly (osmid = 405778748) → It's a loop edge.

Taking the graph produced by:

city_boundary_gdf = ox.geocoder.geocode_to_gdf("Asti 16")
city_boundary_geometry = city_boundary_gdf.geometry[0]
G = ox.graph_from_polygon(city_boundary_geometry, network_type="drive")

It seems that the edge is split into two (probably because it's a loop):

and it's saved to the .gpkg file:

So if the edge is saved to the .gpkg, then I guess the real question is:
Why does it disapear inside the function growbikenet?

I think it's somewhere in the import_network():

My (wild and unverified) guess

I think this specific edge of the Asti street network is a rare case where it is in fact two parallel links that even though represent two different LineStrings, they have the same osmid (405778748), so it is dropped when the graph is transformed into an undirected graph.