How to convert string type to shapely object

Question

I used to have a Pandas data frame with geometry column (I guess you call it as GeoPandas data frame) where all the instances were multiline string with shapely.geometry data type.

When I saved it as a csv (with pandas.to_csv function) I guess python converts them into string type, because when I try to reload it (with pandas.read_csv function), read data frame has a string type data under geometry column.

How can I convert string type MultilineString rows under geometry column into shapely.geometry object.

active.geometry[0]

Gives:

'MULTILINESTRING ((4.6902978485015 50.8812913672182, 4.69041863376864 50.8813473567576, 
4.69048964428198 50.8814069648068, 4.69054925191282 50.8814797434961, 4.69058047643031 50.8815178675679, 4.69103203247511 50.8820691915014, 4.69112606776955 50.8821748368663, 4.69118560934351 50.8822527991125, 4.69122586159577 50.8823195293716, 4.69133401466373 50.8824824286198, 4.69153580362294 50.8827908199779, 4.69155260179444 50.8828164920095))'

BUT, this:

active.geometry[0][0]

Gives

'M'

instead of a shapely object of the first line string that the multiline string objects is composed of.

I also tried shapely.wkt function but it outputs only the first lineString, instead all the lineStrings that multiline string has

Answer 1

You need to convert your pandas DataFrame to a geopandas GeoDataFrame to access the geometry column as shapely object. This post describes how to achieve it.

one solution:

from shapely import wkt
df = pd.read_csv('myFile.csv')
df['geometry'] = df['geometry'].apply(wkt.loads)
gdf = gpd.GeoDataFrame(df, crs='epsg:4326')

Answer 2

In addition to the answer of CodeBard:

When Pandas read a csv file, the columns are strings or numbers, never shapely geometries. A Pandas DataFrame with a geometry column (string) is not a GeoPandas GeoDataFrame (geometry column, shapely geometry)

import pandas as pd
df = pd.read_csv("lat_long.csv")
df
          lat       lon           geometry
0  41.389474  2.156421  POINT (2.15642 41.38947)
1  41.383093  2.181116  POINT (2.18112 41.38309)
df.geometry[0]
'POINT (2.156421 41.389474)'
df.dtypes
lat         float64
lon         float64
geometry    object # not a shapely geometry, in Pandas, an object is a string and it performs strings operations
dtype: object

With the the solution of CodeBard

df['geometry'] = df['geometry'].apply(wkt.loads)
df.geometry[0]
<shapely.geometry.point.Point object at 0x1207ab828>
df.geometry[0].wkt
'POINT (2.156421 41.389474)'
# but
df.dtypes
lat         float64
lon         float64
geometry    object # 

Even if the rows of the column are shapely geometries, this is still not a shapely geometry field

This is why GeoPandas was created (add a geopspatial component to Pandas). With a GeoDataFrame,

gdf = gpd.GeoDataFrame(df)
gdf.geometry[0]
<shapely.geometry.point.Point object at 0x1207ab828>
gdf.dtypes
lat          float64
lon          float64
geometry     geometry #  shapely geometry field here

But when you print or export the GeoDataFrame to a csv file (text file), the geometry field must be a string, the wkt representation of the shapely geometry (= geometry.wkt, POINT (2.15642 41.38947)), and not the shapely geometry (Point(2.15642,41.38947))

 from shapely.geometry import Point
 pt = Point(2.15642,41.38947)
 pt
 <shapely.geometry.point.Point object at 0x10e12ab00>
 pt.wkt
 'POINT (2.15642 41.38947)'
 print(pt)
 POINT (2.15642 41.38947)


gdf
          lat       lon           geometry
0  41.389474  2.156421  POINT (2.15642 41.38947)
1  41.383093  2.181116  POINT (2.18112 41.38309)