closest neighbour in python postgis

closest.py

import psycopg2
import pandas
from collections import defaultdict

data = pandas.read_csv("learning_dataset.csv", sep=";", names="gid,borough,block,lot,zipcode,firecomp,policeprct,ltdheight,bldgclass,landuse,easements,ownertype,lotarea,bldgarea,comarea,resarea,officearea,retailarea,garagearea,strgearea,factryarea,otherarea,areasource,numbldgs,numfloors,unitsres,unitstotal,lotfront,lotdepth,bldgfront,bldgdepth,ext,proxcode,irrlotcode,lottype,bsmtcode,assessland,assesstot,exemptland,exempttot,yearbuilt,builtcode,yearalter1,yearalter2,histdist,landmark,builtfar,residfar,commfar,facilfar".split(","))
# Connect to an existing database
conn = psycopg2.connect("dbname=tahar user=vallette")

# Open a cursor to perform database operations
cur = conn.cursor()
cur2 = conn.cursor()

# for each fire find the closest building
fire_to_building = {}
cur.execute("SELECT unique_key, st_x(st_centroid(st_transform(geom,4326))), st_y(st_centroid(st_transform(geom,4326))) FROM nyc_fireaccidents;")
for fire in cur.fetchall():
	cur2.execute(
		"SELECT gid FROM mappluto_manhattan \
		ORDER BY st_transform(geom,4326) <-> st_setsrid(st_makepoint(%s,%s),4326) \
		LIMIT 1;" % (str(fire[1]), str(fire[2]))
	)
	fire_to_building[int(fire[0])] = cur2.fetchone()[0]
	print fire, fire_to_building[int(fire[0])]

building_to_fires = defaultdict(list)
for key, value in sorted(fire_to_building.iteritems()):
    building_to_fires[value].append(key)

data["number_of_fires"] = data["gid"].apply(lambda x: len(building_to_fires[x]))

# Close communication with the database
cur.close()
conn.close()

data.to_csv("to_learn.csv",sep=";")