current loop logic breaks topology in certain scenarios

[jGaboardi]

Within:

• nodes.induce_nodes()
• nodes.remove_false_nodes()

see below

---

Putting this here from Discord:

may have found a bug in fix_topology() whereby if already fixed stuff is passed in, something happens inside that "un-fixes" it.

import geopandas, shapely, sgeop

p10 = shapely.Point(1,0)
p30 = shapely.Point(3,0)
p50 = shapely.Point(5,0)
p21 = shapely.Point(2,1)
p32 = shapely.Point(3,2)
p41 = shapely.Point(4,1)

line1030 = shapely.LineString((p10, p30))
line3050 = shapely.LineString((p30, p50))
line3041323130 = shapely.LineString((p30, p41, p32, p21, p30))

# input "roads" that already meet out simplified critera
frame = geopandas.GeoDataFrame(geometry=[line1030, line3050, line3041323130])
frame

                               geometry
0                 LINESTRING (1 0, 3 0)
1                 LINESTRING (3 0, 5 0)
2  LINESTRING (3 0, 4 1, 3 2, 2 1, 3 0)

fixed_topology = sgeop.nodes.fix_topology(frame)
fixed_topology

                                 geometry
0.0            LINESTRING (1 0, 3 0, 5 0)
3.0  LINESTRING (3 0, 4 1, 3 2, 2 1, 3 0)

---

xref #68

[jGaboardi]

• I have discovered 2 rough edges in fix_topology() (actual 1 in induce_nodes() and 1 in remove_false_nodes() within fix_topology()) that can lead to improper identification of loop endpoints that in turn can lead to incorrect topology. Both fixes are currently in the GH20_nodes_tests_fix_topology and tests are passing locally (macOS 14).
• CI suite failing on 5/6 FUA, seemingly with only very minor differences.
• Relavent chunks for the fixes here in induce_nodes() and here in remove_false_nodes().
• Still need to add tests
• I can either roll this stuff into #67 or make a fresh PR for it once #67 is merged.

[martinfleis]

I'd keep this separately from #67