csv-sparql2geojs/csv-sparql2geojs.py

# @version 0.2.0

from geojson import Point, LineString, Polygon, MultiPolygon, Feature, FeatureCollection
from geojson import dumps as geodumps
from csv import reader as csvreader
#import json
from pprint import pprint # For pretty-printing JSON
#import sys
#import urllib.request, urllib.parse, urllib.error


# Inputs
csvFileName = 'all-pois-italy.csv'

# Function to convert SPARQL query results to GeoJSON FeatureCollection
def sparql2geojs(csv):
    rows = 0
    header = False
    features = []
    for row in csv:
        if not header:
            header = True
            continue
        #pprint(result)
        featureId = row[0]
        feature = next((f for f in features if f and f['id'] == featureId), None)
        # SPARQL endpoint returns one POI more that once, if it has more than one property
        if feature is not None:
            updateGeoJSONFeature(feature, row)
        else:
            feature = createGeoJSONFeature(row)
            features.append(feature)
        rows += 1
        if rows % 1000 == 0:
            print(rows, ' lines processed')
    return FeatureCollection(features)

# Supplementary function to convert one result to a GeoJSON feature
def createGeoJSONFeature(row):
    # each result from the SPARQL query comes with just one property-value pair
    props = {row[1]:row[2]}
    id = row[0] # ID of the feature is its URI
    geom = parseWKT(row[3])
    #pprint(geom.get('type'))
    if geom.get('type').upper() == "POINT":
        #pprint(Feature(geometry = Point(coords), id = id, properties = props))
        return Feature(geometry = geom, id = id, properties = props)
    elif geom.get('type').upper() == "MULTIPOLYGON":
        #pprint(Feature(geometry = MultiPolygon(coords), id = id, properties = props))
        return Feature(geometry = MultiPolygon(coords), id = id, properties = props)

# Supplementary function to update one GeoJSON feature if it already exists
def updateGeoJSONFeature(feature, result):
    # Case 1: property does not exist in features's properties
    if result[1] not in feature['properties'].keys():
        feature['properties'][result[1]] = result[2]
    else: # Case 2: property already exists, hence it has more than 1 value
        pass

# Supplementary function to parse WKT Literal.
# It is further extended by specific functions for specific geometry types
def parseWKT(wktLiteral):
    type = wktLiteral.split("(")[0]
    if type.strip() == "POINT":
        wktCoords = wktLiteral[len(type):].replace("(", "").replace(")", "")
        return Point(parseWKTPoint(wktCoords))
    elif type.strip() == "LINESTRING":
        wktCoords = wktLiteral[len(type):]
        return LineString(parseWKTLineString(wktCoords))
    elif type.strip() == "POLYGON":
        wktCoords = wktLiteral[len(type)+1:-1]
        return Polygon(parseWKTPolygon(wktCoords))
    elif type.strip() == "MULTIPOLYGON":
        wktCoords = wktLiteral[len(type)+1:-1]
        return MultiPolygon(parseWKTMultiPolygon(wktCoords))
    #     # find polygons first ...
    #     parts = wktLiteral[15:].replace(')))', '').split(')),') #[15:] will intentionally left two ( in the beginning
    #     print(len(parts))
    #     coords = []
    #     for part in parts:
    #         # ... then rings of the polygons ...
    #         rings = part.split('),')
    #         print(len(rings))
    #         polygon = [] # this must be converted into tuple later on, but tuples are immutable
    #         for r in rings:
    #             # ... and finally points of the rings
    #             points = part[2:].split(',')
    #             pprint(points)
    #             ring = []
    #             for point in points:
    #                 if len(point) > 1:
    #                     pprint(point)
    #                     ring.append( (float(point.split(' ')[0]), float(point.split(' ')[1])) )
    #             polygon.append(ring)
    #         coords.append( tuple(polygon) )
    # TODO: other geometry types support
    else:
        print("Unsupported geometry type {}! Will produce empty point object without coordinates!".format(type))
    return Point([])

# Supplementary function to parse WKT Point geometry to its GeoJSON equivalent.
def parseWKTPoint(wktLiteral):
    lonLat = wktLiteral.split(" ")
    return (float(lonLat[0]), float(lonLat[1]))

# Supplementary function to parse WKT LineString geometry to its GeoJSON equivalent.
def parseWKTLineString(wktLiteral):
    lineString = []
    points = wktLiteral.replace("(", "").replace(")", "").split(",")
    for point in points:
        lineString.append(parseWKTPoint(point.strip()))
    return lineString

# Supplementary function to parse WKT Polygon geometry to its GeoJSON equivalent.
def parseWKTPolygon(wktLiteral):
    polygon = []
    rings = wktLiteral.split(",(")
    for ring in rings:
        polygon.append(parseWKTLineString(ring))
    return polygon

# Supplementary function to parse WKT MultiPolygon geometry to its GeoJSON equivalent.
def parseWKTMultiPolygon(wktLiteral):
    multipolygon = []
    polygons = wktLiteral.split(",((")
    for polygon in polygons:
        multipolygon.append(parseWKTPolygon(polygon))
    return multipolygon


# Executional part
with open(csvFileName, encoding='utf-8') as csvfile:
    filereader = csvreader(csvfile)
    print('File read, parsing ...')
    geojs = sparql2geojs(filereader)
#pprint(js)
#pprint(geojs)
print(geojs.is_valid)
print(geojs.errors())

# uncomment following lines if you want to save the output into a file
outFileName = csvFileName.split(".")[0]
with open(outFileName + ".geojson", 'w') as out:
    out.write(geodumps(geojs))