<input type="checkbox" id=

Simpler code snippet: <div class="highlight highlight-source-python notranslate po

<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/us

The same behaviour as in <a class="issue-link js-issue-link" data-error-text="Failed t

<a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="/us

BUG: The dissolve feature generates a new geometry about geopandas HOT 4 OPEN

thomas-leduc commented on June 2, 2024

BUG: The dissolve feature generates a new geometry

from geopandas.

Comments (4)

thomas-leduc commented on June 2, 2024

Simpler code snippet:

from geopandas import GeoDataFrame
from shapely import LineString

gdf1 = GeoDataFrame([
	{"geometry": LineString([ (355041.15, 6688781.25, 0), (355040.9629213488, 6688781.437078651, 9.7) ])},
	{"geometry": LineString([ (355041.1500000001, 6688781.25, 0), (354841.1500000001, 6688781.25, 0) ])}
], crs="epsg:2154")

gdf2 = gdf1.dissolve()

assert (len(gdf2.loc[0, "geometry"].geoms) == len(gdf1)), "Bug"

from geopandas.

jdmcbr commented on June 2, 2024

@thomas-leduc I took a quick look, and will aim to return later. It seems to be a rounding error happening somewhere, resulting in a tiny small extra snippet.

If I truncate the first 5 digits from the x/y coords of the line string:

gdf1 = GeoDataFrame([
    {"geometry": LineString([ (1.15, 81.25, 0), (0.9629213488, 81.437078651, 9.7) ])},
    {"geometry": LineString([ (1.1500000001, 1.25, 0), (1.1500000001, 81.25, 0) ])}
], crs="epsg:2154")
gdf2 = gdf1.dissolve()
print(len(gdf2.loc[0, "geometry"].geoms) == len(gdf1))
# True

With the original geometry where the bug occurs:

gdf3.geometry.length
gdf1 = GeoDataFrame([
    {"geometry": LineString([ (355041.15, 6688781.25), (355040.9629213488, 6688781.437078651) ])},
    {"geometry": LineString([ (355041.1500000001, 6688781.25), (354841.1500000001, 6688781.25) ])}
], crs="epsg:2154")

gdf2 = gdf1.dissolve()
gdf3 = gpd.GeoDataFrame(geometry=[g for g in gdf2.loc[0, 'geometry'].geoms])
print(gdf3.geometry.length)

0    2.645692e-01
1    5.820766e-11
2    2.000000e+02
dtype: float64

from geopandas.

theroggy commented on June 2, 2024

The same behaviour as in shapely/shapely#2028 is happening here, because under the hood dissolve also uses shapely.union_all...

from geopandas import GeoDataFrame
import shapely
from shapely import LineString

gdf1 = GeoDataFrame([
	{"geometry": LineString([ (355041.15, 6688781.25, 0), (355040.9629213488, 6688781.437078651, 9.7) ])},
	{"geometry": LineString([ (355041.15, 6688781.25, 0), (354841.1500000001, 6688781.25, 0) ])}
], crs="epsg:2154")

gdf2 = gdf1.dissolve()
print(f"{gdf2.loc[0, 'geometry']=}")
# gdf2.loc[0, 'geometry']=<MULTILINESTRING Z ((355041.15 6688781.25 0, 355040.963 6688781.437 9.7), (3...>

merged = shapely.line_merge(gdf1.geometry)
print(f"{merged[0]=}")
# merged[0]=<LINESTRING Z (355041.15 6688781.25 0, 355040.963 6688781.437 9.7)>