Adding metrics to strokes made by COINS #704

csebastiao · 2025-07-24T08:56:29Z

See #591 for explanation. Drafted without full tests and documentation. We ran the code with https://github.com/pysal/momepy/blob/main/ci/envs/312-latest.yaml.

Contribution

Adding a strokegraph.py file with 2 main functions:

strokes_to_graph taking momepy.COINS as an input and returning a dual graph where each node is a stroke in momepy.COINSand each edge is when two strokes are intersecting. Optional arguments include computing additional metrics on the strokes (degree, closeness, betweenness, connectivity, access, orthogonality, and spacing, see (El Gouj et al., 2022) for details), and returning the primal graph of the street network with the information about each stroke on every edge composing them.
graph_to_strokes taking the dual graph generated by strokes_to_graph as an input and returning a momepy.COINS object as an output.

Questions

Do we need graph_to_strokes, as it returns the output from momepy.COINS ?
Does the structure of strokes_to_graph where the return of the primal graph is optional make sense ?

TODOs

Finish the documentation.
Finish test coverage.

…/momepy into stroke-graph

codecov · 2025-07-24T09:59:04Z

Codecov Report

❌ Patch coverage is 49.07407% with 55 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.0%. Comparing base (4037c70) to head (6d170ee).
⚠️ Report is 129 commits behind head on main.

Files with missing lines	Patch %	Lines
momepy/strokegraph.py	32.1%	55 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main    #704     +/-   ##
=======================================
- Coverage   97.4%   97.0%   -0.3%     
=======================================
  Files         26      28      +2     
  Lines       4328    4303     -25     
=======================================
- Hits        4214    4176     -38     
- Misses       114     127     +13

Files with missing lines	Coverage Δ
momepy/__init__.py	`100.0% <100.0%> (ø)`
momepy/tests/test_strokegraph.py	`100.0% <100.0%> (ø)`
momepy/strokegraph.py	`32.1% <32.1%> (ø)`

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

anastassiavybornova · 2025-10-30T15:28:28Z

graph_to_strokes: pointless -- can be simply dropped @anastassiavybornova
for metrics that are vanilla networkx: do NOT compute separately. @anastassiavybornova
we do not need to return primal graph at all (anvy: make sure we can recreate -- mapping through geopandas) @anastassiavybornova
have a function that only transforms data structure coins>graph. instead of "strokes_to_graph", call it "coins_to_nx". @anastassiavybornova
for metrics that are derived: separate functions, if they rely on some vanilla stuff >> have it there inside the derived-metric function. in momepy, the logic is: every metric has its own function. steps: create graph > pass it to function > return metric. @anastassiavybornova
update documentation of each metric (in the separate functions, explain how metric is computed and what literature it is based on) @csebastiao
in examples/strokegraph.ipynb, showcase: visualize coins-to-nx process and metrics for Bubenec @csebastiao

once api is fixed, ping martin, get thumbs up, then we can work on tests.

csebastiao · 2025-12-08T15:37:35Z

Functions should be fixed, with added documentation and an example in a jupyter notebook. Tell me what you think @martinfleis (and @anastassiavybornova for the documentation) and then we can continue with the tests !

martinfleis

Cool! This looks much better and super close to what I think is mergeable. Some comments left in the code.

martinfleis · 2025-12-09T09:12:13Z

momepy/strokegraph.py

+    >>> gdf = momepy.remove_false_nodes(gdf)
+    >>> primal_graph = momepy.gdf_to_nx(gdf, preserve_index=True, approach="primal")
+    >>> lines = momepy.nx_to_gdf(primal_graph, points=False, lines=True)
+    >>> coins = momepy.COINS(lines)
+    >>> stroke_graph = momepy.coins_to_nx(coins)


Suggested change

>>> gdf = momepy.remove_false_nodes(gdf)

>>> primal_graph = momepy.gdf_to_nx(gdf, preserve_index=True, approach="primal")

>>> lines = momepy.nx_to_gdf(primal_graph, points=False, lines=True)

>>> coins = momepy.COINS(lines)

>>> stroke_graph = momepy.coins_to_nx(coins)

>>> coins = momepy.COINS(gdf)

>>> stroke_graph = momepy.coins_to_nx(coins)

This should work as well, no?

martinfleis · 2025-12-09T09:13:26Z

momepy/strokegraph.py

+    Creates the stroke graph of a street network. The stroke graph is similar to, but
+    not identical with, the dual graph. In the stroke graph, each stroke (see
+    ``momepy.COINS``) is a node; and each intersection between two strokes is an edge.


Suggested change

Creates the stroke graph of a street network. The stroke graph is similar to, but

not identical with, the dual graph. In the stroke graph, each stroke (see

``momepy.COINS``) is a node; and each intersection between two strokes is an edge.

Creates the continuity stroke graph of a street network. The continuity stroke graph is similar to, but

not identical with, the dual graph. In the continuity stroke graph, each stroke (see

``momepy.COINS``) is a node; and each intersection between two strokes is an edge.

martinfleis · 2025-12-09T09:13:37Z

momepy/strokegraph.py

+    Parameters
+    ----------
+    coins: momepy.COINS
+        Strokes computed from a street network.


Suggested change

Strokes computed from a street network.

Continuity strokes computed from a street network.

martinfleis · 2025-12-09T09:14:55Z

momepy/strokegraph.py

+    stroke_gdf["rep_point"] = stroke_gdf.geometry.apply(
+        lambda x: x.interpolate(0.5, normalized=True)
+    )


Suggested change

stroke_gdf["rep_point"] = stroke_gdf.geometry.apply(

lambda x: x.interpolate(0.5, normalized=True)

)

stroke_gdf["rep_point"] = stroke_gdf.geometry.interpolate(0.5, normalized=True)

Vectorized ops are always preferred. There's nearly no need for apply in geopandas.

martinfleis · 2025-12-09T09:21:36Z

momepy/strokegraph.py

+def stroke_connectivity(stroke_graph):
+    """
+    Computes the stroke's connectivity. Connectivity is defined as
+    the number of arcs a stroke intersects. Comparing to the degree,


What is an "arc" in this context?

martinfleis · 2025-12-09T09:26:19Z

docs/api.rst

+Computing dual graph
+---------------------
+
+Set of functions to create a dual graph and compute metrics on:


Suggested change

Computing dual graph

---------------------

Set of functions to create a dual graph and compute metrics on:

Computing continuity graph

---------------------------

Set of functions to create a continuity graph and compute metrics on:

martinfleis · 2025-12-09T09:27:41Z

docs/examples/strokegraph.ipynb

"It is a dual graph as defined in Porta, S., Crucitti, P. and Latora, V., 2006." - is it? That is not based on continuity strokes.

from matplotlib import pyplot as plt -> import matplotlib.pyplot as plt

Why do you go through the primal graph there? Seems wasteful.

I could go on but I might take a stab at revising this notebook myself later on. Generally it feels a bit too complicated. A user guide should narrowly focus whatever is being explained without other complications. there's a lot of additional code in here that is more a distraction than anything else. Like the conversion to nx graph, all those multilayer plots, stuff like stroke_betweenness_dict mapping and so on. Keep the notebook as is for now though. Might be easier for me to do that than trying to explain myself.

martinfleis · 2025-12-09T13:03:02Z

momepy/strokegraph.py

+    """
+    point = tuple(point)
+    coords = list(linestring.coords)
+    assert point in coords, "point not on linestring!"


Suggested change

assert point in coords, "point not on linestring!"

You got this covered by the more appropriate ValueError below.

martinfleis · 2025-12-09T13:08:06Z

momepy/strokegraph.py

+    stroke_gdf["edge_indeces"] = stroke_gdf.stroke_id.apply(
+        lambda x: list(stroke_attribute[stroke_attribute == x].index)
+    )


Suggested change

stroke_gdf["edge_indeces"] = stroke_gdf.stroke_id.apply(

lambda x: list(stroke_attribute[stroke_attribute == x].index)

)

stroke_attribute.groupby(stroke_attribute).apply(lambda group: group.index.tolist())

No for-loops over rows :). This is much faster.

martinfleis · 2025-12-09T13:11:08Z

momepy/strokegraph.py

+    # Add stroke ID to each edge on (primal) graph
+    nx.set_edge_attributes(
+        G=graph,
+        values={
+            e: int(stroke_attribute[graph.edges[e]["index_position"]])
+            for e in graph.edges
+        },
+        name="stroke_id",
+    )


I don't get the point of this. graph already has stroke_id on edges coming from gdf_to_nx. It is a wrong one or what?

anastassiavybornova and others added 29 commits April 14, 2025 11:43

add playground nb for strokes graph

06c0756

from input to primal graph with stroke attrs

512a18a

Add clse notebook

6ebaa17

update workflow input>stroke_graph

de362e4

Add angles, wrong angles for now

cc0442e

remove useless indexing and create graph w attribute, angle still wrong

0128adf

update strokegraph nb: finished workflow (angles tbd)

07b2e16

updated angle function

80e1754

update anvy nb (final angle comp)

b2b7bdc

rename nb

e21302c

add pickle-saving to clse nb

e84702a

run anvy nb

e516188

run clse nb

f87e14c

add comparison notebook

be863d3

solve angle issue in clse

d22ffe8

Add comparison function on osmnx graph

d8a7b09

add combination notebook

078ba96

add strokegraph.py functions

afac2f8

add draft for strokegraph tests

54c000a

add strokegraph to init

29c5ace

update tests for strokegraph

10b85e3

update strokegraph.py

7068131

update compare nb: clse vs x2

3d81d3a

add x2 nb and pickle

a3fa4cb

Merge branch 'pysal:main' into stroke-graph

7dea1fc

Add g2s and add metrics to primal graph from strokes

f63d47d

Merge branch 'stroke-graph' of https://github.com/anastassiavybornova…

e1f9e18

…/momepy into stroke-graph

Add doc g2s

4e65dfb

Add primal_graph doc, remove notebooks and pickles

7b9c02c

anastassiavybornova and others added 12 commits November 7, 2025 08:24

remove graph2stroke func

6158617

Merge branch 'main' into stroke-graph

c47ddc2

remove vanilla nx metrics

f2ed947

remove primal-graph return

f0f53d2

strokes_to_graph > coins_to_nx; remove metrics computation

6a375e1

derived stroke metrics as separate functions

4054462

add playground notebook

8201242

add draft of strokegraph example nb

d71a32a

update documentation of strokegraph, add example notebook

6748879

Update coins_to_nx example

3248e05

Update documentation and example notebook strokegraph

defba51

add doctest skip and format imports

6d170ee

martinfleis reviewed Dec 9, 2025

View reviewed changes

	Strokes computed from a street network.
	Continuity strokes computed from a street network.

Adding metrics to strokes made by COINS #704

Are you sure you want to change the base?

Adding metrics to strokes made by COINS #704

Conversation

csebastiao commented Jul 24, 2025

Contribution

Questions

TODOs

Uh oh!

codecov bot commented Jul 24, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

anastassiavybornova commented Oct 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

csebastiao commented Dec 8, 2025

Uh oh!

martinfleis left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov bot commented Jul 24, 2025 •

edited

Loading

anastassiavybornova commented Oct 30, 2025 •

edited

Loading