generate-buildings-stations.py

'''
When everything is finished, use `gzip -k -f buildingObs.n3` to compress files.
'''

import random
import time
prefix = '''
prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns>
prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>
prefix sosa: <http://www.w3.org/ns/sosa/> 
prefix xsd: <http://www.w3.org/2001/XMLSchema#> 
prefix ex: <http://www.example3.com/> 
'''

'''
Weather station observations
'''
n = time.time()

# 10 years
# 300 stations * 15 obs/day * 10 years * 365 days/year
size = 300*15*10*365

df_weather_obs = pd.DataFrame({
  # 300 stations. 0-299, then repeats.
  'station': np.arange(size) % 300,
  'result': np.random.randint(0, 10, size, dtype=np.uint32),
  # Every 300 rows, increment by 1 until 14. Then repeat.
  'property': (np.arange(size) / 300).astype(int) % 15, #% 300
  # Each day contains 300 stations * 15 parameters per station per day.
  # Convert to seconds per day, then subtract from today.
  'time': n - ((np.arange(size) / (300*15)).astype(int) * 24*60*60)
})
df_weather_obs['time'] = df_weather_obs['time'].astype('datetime64[s]')
df_weather_obs['year'] = df_weather_obs['time'].dt.year
df_weather_obs['month'] = df_weather_obs['time'].dt.month
df_weather_obs['day'] = df_weather_obs['time'].dt.day
df_weather_obs['idx'] = df_weather_obs.index

# Assume that the stations occasionally don't report one or more parameters on some days.
df_weather_obs = df_weather_obs.sample(frac=9999/10000)

# Write to n3 file
fn = 'weatherObs.n3'
# Reorder cols
df_weather_obs = df_weather_obs[['idx','station','year','month','day','property','result']] 
with open(fn, 'w') as f:
  f.write(prefix)
  np.savetxt(f, df_weather_obs.values, fmt='''
ex:weatherObservation-%s a sosa:Observation ;
  sosa:hasFeatureOfInterest ex:weatherStation-%s ;
  sosa:resultTime "%s-%s-%sT00:00:00"^^xsd:dateTime ;
  sosa:hasProperty "%s"^^xsd:string ;
  sosa:hasSimpleResult "%s"^^xsd:double ;
  ex:observationType "weather"^^xsd:string .
''')

'''
Building observations
'''
n = time.time()
# 600 buildings * 1 obs/day * 365 days/year * 10 years
size = 600*1*365*10
df_building_obs = pd.DataFrame({
  'building': np.arange(size) % 600,
  'time': n - ((np.arange(size) / 600).astype(int) * 24*60*60),
  'result': np.random.randint(0, 10, size, dtype=np.uint32)
})
df_building_obs['time'] = df_building_obs['time'].astype('datetime64[s]')
df_building_obs['year'] = df_building_obs['time'].dt.year
df_building_obs['month'] = df_building_obs['time'].dt.month
df_building_obs['day'] = df_building_obs['time'].dt.day
df_building_obs['idx'] = df_building_obs.index

# 2190000 rows means every building has 1 obs per day but we only assume 
# 50 obs per year per building.
df_building_obs = df_building_obs.sample(frac=50/365)

# Write to n3 file
fn = 'buildingObs.n3'
# Reorder cols
df_building_obs = df_building_obs[['idx','building','year','month','day','result']] 
with open(fn, 'w') as f:
  f.write(prefix)
  np.savetxt(f, df_building_obs.values, fmt='''
ex:buildingObservation-%s a sosa:Observation ;
  sosa:hasFeatureOfInterest ex:building-%s ;
  sosa:resultTime "%s-%s-%sT00:00:00"^^xsd:dateTime ;
  sosa:hasSimpleResult "%s"^^xsd:double ;
  ex:observationType "building"^^xsd:string .
''')
  
'''
Distance + FOI
'''
# fois
size = 300
df1 = pd.DataFrame({
  'station': np.arange(size).astype(int),
  'lat': np.random.uniform(-90, 90, size),
  'lon': np.random.uniform(-180, 180, size),
})
size = 600
df2 = pd.DataFrame({
  'building': np.arange(size).astype(int),
  'lat': np.random.uniform(-90, 90, size),
  'lon': np.random.uniform(-180, 180, size),
})
# dist
dist = pd.DataFrame({
  'building': (np.arange(600*300) % 600).astype(int), 
  'station': (np.arange(600*300) / 600).astype(int),
})
from geopy import distance
dist['km'] = dist.apply(lambda x: distance.distance(
  (df1.iloc[x.station]['lat'], df1.iloc[x.station]['lon']),
  (df2.iloc[x.building]['lat'], df2.iloc[x.building]['lon'])).km, axis=1)
dist['km'] = dist['km'].astype(int)
dist = dist.iloc[:, [1, 2, 0, 0, 2, 1]] # reorder cols

# Write FOIs to file
fn = 'foi.n3'
with open(fn, 'w') as f:
  f.write(prefix)
  np.savetxt(f, df1.to_numpy(dtype=str), fmt='''
ex:weatherStation-%s a sosa:Sensor ;
  ex:geoType "weather"^^xsd:string ;
  geo:asWKT "<http://www.opengis.net/def/crs/EPSG/0/4326> POINT(%s %s)"^^geo:wktLiteral .
''')
  np.savetxt(f, df2.to_numpy(dtype=str), fmt='''
ex:building-%s a sosa:Sensor ;
  ex:geoType "building"^^xsd:string ;
  geo:asWKT "<http://www.opengis.net/def/crs/EPSG/0/4326> POINT(%s %s)"^^geo:wktLiteral .
''')
  
# Write to distances to file
fn = 'dist.n3'
with open(fn, 'w') as f:
  f.write(prefix)
  np.savetxt(f, dist.to_numpy(dtype=str), fmt='''
ex:weatherStation-%s ex:hasDistanceToBuilding [
  ex:distance "%s"^^xsd:decimal ;
  ex:building ex:building-%s ] .

ex:building-%s ex:hasDistanceToStation [
  ex:distance "%s"^^xsd:decimal ;
  ex:weatherStation ex:weatherStation-%s ] .
''')