def test_smokes_friends_cancer_gamma():
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
cur.execute(open("test/test_data/smokesfriendscancer-symmetric-20.sql", "r").read())
domain = 20
P1 = 0.667129
P2 = 0.776870
F = 0.009952
C = 0.091123
dnf = safesample.query_parser.parse("P1(x,y),Rel*!(x),Friends(x,y),~Rel*!(y) v P2(x),Rel*!(x),~Cancer(x)")
plan = algorithm.getSafeQueryPlan(dnf)
sql = plan.generateSQL_DNF()
for r in range(5):
cur.execute("drop table if exists Rel;")
cur.execute("create table Rel as (select id, v0, CASE WHEN random() < p THEN 1 ELSE 0 END as p from smokes);")
cur.execute("select count(*) from Rel where p = 1")
cardinality = cur.fetchone()[0]
cur.execute(sql)
prob = cur.fetchone()[0]
#print prob
#print (1 - (1-(1 - (1-P2*(1-C))**cardinality))*(1-(1-(1-P1*F)**(cardinality*(domain-cardinality)))))
assert abs(prob - (1 - (1-(1 - (1-P2*(1-C))**cardinality))*(1-(1-(1-P1*F)**(cardinality*(domain-cardinality)))))) < 0.000001
def test_smokes_friends_cancer_query1_generic():
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
cur.execute(open("test/test_data/smokesfriendscancer-symmetric-20.sql", "r").read())
domain = 20
P1 = 0.667129
P2 = 0.776870
F = 0.009952
C = 0.091123
for r in range(1):
cur.execute("drop table if exists Rel;")
cur.execute("create table Rel as (select id, v0, CASE WHEN random() < p THEN 1 ELSE 0 END as p from smokes);")
cur.execute("select count(*) from Rel where p = 1")
cardinality = cur.fetchone()[0]
probGamma = (1 - (1-(1 - (1-P2*(1-C))**cardinality))*(1-(1-(1-P1*F)**(cardinality*(domain-cardinality)))))
probConstants = [0]*(domain+1)
dnf = safesample.query_parser.parse("Rel*![c]() v P1(x,y),Rel*!(x),Friends(x,y),~Rel*(y) v P2(x),Rel*!(x),~Cancer(x)" )
plan = algorithm.getSafeQueryPlan(dnf)
sql = plan.generateSQL_DNF()
cur.execute(sql)
for (const, prob) in cur:
probConstants[const] = prob
for c in range(1, domain+1):
cur.execute("select count(*) from Rel where p = 1 and v0 = %d" % (c))
inRelSample = (cur.fetchone()[0] > 0)
if inRelSample:
assert probConstants[c] == 1
else:
assert abs(probConstants[c] - probGamma) < 0.000001
def exactProb(queryDNF):
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
plan = algorithm.getSafeQueryPlan(queryDNF)
querySQL = plan.generateSQL_DNF()
cur.execute(querySQL)
return cur.fetchone()[0]
def test_q3_generic():
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
cur.execute(open("test/test_data/non_boolean_sampling.sql", "r").read())
dnf = safesample.query_parser.parse("Smokes*![c](),Friends[-c](x)")
plan = algorithm.getSafeQueryPlan(dnf)
sql = plan.generateSQL_DNF()
cur.execute(sql)
constants = set()
for (c, p) in cur:
constants.add(c)
assert p == 0.59349810824172
assert constants == set((3,4,6))
def test_q2_generic():
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
cur.execute(open("test/test_data/non_boolean_sampling.sql", "r").read())
dnf = safesample.query_parser.parse("Smokes*,Friends[-c,c](x)")
plan = algorithm.getSafeQueryPlan(dnf)
sql = plan.generateSQL_DNF()
cur.execute(sql)
constants = set((3,4,6))
count = 0
for (c, p) in cur:
count += 1
if c in constants:
assert p == 0.0198049576959999
else:
assert p == 0.0295598587570093
assert count == 10
def test_smokes_friends_cancer_query2():
conn = psycopg2.connect(dbname="sampling_test")
conn.autocommit = True
cur = conn.cursor()
cur.execute(open("test/test_data/smokesfriendscancer-symmetric-20.sql", "r").read())
domain = 20
P1 = 0.667129
P2 = 0.776870
F = 0.009952
C = 0.091123
for r in range(1):
cur.execute("drop table if exists Rel;")
cur.execute("create table Rel as (select id, v0, CASE WHEN random() < p THEN 1 ELSE 0 END as p from smokes);")
cur.execute("select count(*) from Rel where p = 1")
cardinality = cur.fetchone()[0]
for c in range(1, domain+1):
dnf = safesample.query_parser.parse("Cancer[%d]() v P1(x,y),Rel*(x),Friends(x,y),~Rel*(y) v P2[-%d](x),Rel*[-%d](x),~Cancer[-%d](x) v P2[%d](),Rel*[%d]()" % (c,c,c,c,c,c))
plan = algorithm.getSafeQueryPlan(dnf)
sql = plan.generateSQL_DNF()
cur.execute(sql)
probQ1 = cur.fetchone()[0]
cur.execute("select count(*) from Rel where p = 1 and v0 = %d" % (c))
cardinalityJustC = cur.fetchone()[0] # because we shattered, need to see if c is in Rel* or not
q1 = C
q2 = 1-(1-P1*F)**(cardinality*(domain-cardinality))
q3 = 1-(1-P2*(1-C))**(cardinality - cardinalityJustC)
q4 = P2*cardinalityJustC
probQuery = 1 - (1-q1)*(1-q2)*(1-q3)*(1-q4)
print "Cancer[%d]() v P1(x,y),Rel*(x),Friends(x,y),~Rel*(y) v P2[-%d](x),Rel*[-%d](x),~Cancer[-%d](x) v P2[%d](),Rel*[%d]()" % (c,c,c,c,c,c)
print probQ1
print probQuery
assert abs(probQ1 - probQuery) < 0.000001
def exactProb(queryDNF):
plan = algorithm.getSafeQueryPlan(queryDNF)
querySQL = plan.generateSQL_DNF()
exactProb = safesample.query_parser.executeSQL(querySQL)
return exactProb
def exactProb(queryDNF):
plan = algorithm.getSafeQueryPlan(queryDNF)
querySQL = plan.generateSQL_DNF()
exactProb = safesample.query_parser.executeSQL(querySQL)
return exactProb