Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Write GeoDataFrame into SQL Database

I hope that my question is not ridiculous since, surprisingly, this question has apparently not really been asked yet (to the best of my knowledge) on the popular websites.

The situation is that I have several csv files containing more than 1 Mio observations in total. Each observation contains, among others, a postal address. I am planning to read all files into a single GeoDataFrame, geocode the addresses, perform a spatial join given a shapefile and save some information from the polygon for each row. Quite standard, I suppose. This is a part of a one-time data cleaning process.

My goal is to set up a database with this final dataset. This is because it allows me to share and search the data quite easily as well as e.g. plot some observations on a website. Also, it makes it quite easy to select observations based on some criteria and then run some analyses.

My problem is that the feature of inserting a GeoDataFrame into a database seems not to be implemented yet - apparently because GeoPandas is supposed to be a subsitute for databases ("GeoPandas enables you to easily do operations in python that would otherwise require a spatial database such as PostGIS").

Of course, I could iterate through each line and insert each data point "manually", but I am looking for the best solution here. For any workaround I would also be afraid that the datatype may conflict with that of the database. Is there "a best way" to take here?

Thanks for your help.

like image 880
Jhonny Avatar asked Jul 13 '16 20:07

Jhonny


3 Answers

A version of Hamri Said's answer, but using a lambda, which in my opinion is a bit nicer because it is such a short function:

# Imports
from geoalchemy2 import Geometry, WKTElement
from sqlalchemy import *

geodataframe['geom'] = geodataframe['geom'].apply(lambda geom: WKTElement(geom.wkt, srid = <your_SRID>))

db_url = 'postgresql://username:password@host:socket/database'
engine = create_engine(db_url, echo=False)

# Use 'dtype' to specify column's type
# For the geom column, we will use GeoAlchemy's type 'Geometry'
your_geodataframe.to_sql(table_name, engine, if_exists='append', index=False, 
                         dtype={'geom': Geometry('POINT', srid= <your_srid>)})
like image 116
wfgeo Avatar answered Oct 25 '22 17:10

wfgeo


As mentioned before, @Kartik's answer works only for a single call, for appending data it raises a DataError since the geom column then expects the geometry to have an SRID. You can use GeoAlchemy to handle all the cases:

# Imports
from geoalchemy2 import Geometry, WKTElement
from sqlalchemy import *

# Use GeoAlchemy's WKTElement to create a geom with SRID
def create_wkt_element(geom):
    return WKTElement(geom.wkt, srid = <your_SRID>)

geodataframe['geom'] = geodataframe['geom'].apply(create_wkt_element)

db_url = 'postgresql://username:password@host:socket/database'
engine = create_engine(db_url, echo=False)

# Use 'dtype' to specify column's type
# For the geom column, we will use GeoAlchemy's type 'Geometry'
your_geodataframe.to_sql(table_name, engine, if_exists='append', index=False, 
                         dtype={'geom': Geometry('POINT', srid= <your_srid>)})
like image 35
Hamri Said Avatar answered Oct 25 '22 16:10

Hamri Said


So, I just implemented this for a PostGIS database, and I can paste my method here. For MySQL, you'll have to adapt the code.

First step was to convert the geocoded columns into WKB hex string, because I use SQLAlchemy, with an engine based on pyscopg, and both of those packages do not understand geo-types natively. Next step is to write that data into a SQL DB, as usual (note that all geometry columns should be converted to text columns holding the WKB hex string), and finally change the type of the columns to Geometry by executing a query. Refer to the following pseudocode:

# Imports
import sqlalchemy as sal
import geopandas as gpd

# Function to generate WKB hex
def wkb_hexer(line):
    return line.wkb_hex

# Convert `'geom'` column in GeoDataFrame `gdf` to hex
    # Note that following this step, the GeoDataFrame is just a regular DataFrame
    # because it does not have a geometry column anymore. Also note that
    # it is assumed the `'geom'` column is correctly datatyped.
gdf['geom'] = gdf['geom'].apply(wkb_hexer)

# Create SQL connection engine
engine = sal.create_engine('postgresql://username:password@host:socket/database')

# Connect to database using a context manager
with engine.connect() as conn, conn.begin():
    # Note use of regular Pandas `to_sql()` method.
    gdf.to_sql(table_name, con=conn, schema=schema_name,
               if_exists='append', index=False)
    # Convert the `'geom'` column back to Geometry datatype, from text
    sql = """ALTER TABLE schema_name.table_name
               ALTER COLUMN geom TYPE Geometry(LINESTRING, <SRID>)
                 USING ST_SetSRID(geom::Geometry, <SRID>)"""
    conn.execute(sql)
like image 45
Kartik Avatar answered Oct 25 '22 16:10

Kartik