def test_save_load_feature_table():
#mg.init_jvm()
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
feature_table0 = mg.get_features_for_blocking(A, B)
mg.save_object(feature_table0, filename)
feature_table1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(feature_table0), type(feature_table1))
assert_equal(len(feature_table0), len(feature_table1))
assert sorted(feature_table0.columns) == sorted(feature_table1.columns)
ft0_functions = list(feature_table0['function'])
ft1_functions = list(feature_table1['function'])
for f0, f1 in zip(ft0_functions, ft1_functions):
a = f0(A.ix[1], B.ix[2])
b = f1(A.ix[1], B.ix[2])
if math.isnan(a) == False and math.isnan(b) == False:
assert_equal(a, b)
if math.isnan(a) == True:
assert_equal(math.isnan(b), True)
def test_rb_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['name_name_mel(ltuple, rtuple) < 0.4',
'birth_year_birth_year_lev(ltuple, rtuple) < 0.5'],
feature_table)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) != 1'],
feature_table)
C = rb.block_tables(A, B, 'zipcode', 'zipcode')
s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
A['dummy'] = 1
B['dummy'] = 1
ab = mg.AttrEquivalenceBlocker()
D = ab.block_tables(A, B, 'dummy','dummy')
fv = mg.extract_feature_vecs(D, feature_table=feature_table)
expected_ids = fv.ix[((fv.name_name_mel >= 0.4) | (fv.birth_year_birth_year_lev >= 0.5)) &
fv.zipcode_zipcode_exm == 1 ,
['ltable.ID', 'rtable.ID',
]]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_ab_block_tables_skd():
start_time = time.time()
#A = mg.load_dataset('bikedekho_clean', 'ID')
A = mg.load_dataset('bowker', 'ID')
a_load_time = time.time()
print("Loading table A --- %s seconds ---" % (a_load_time - start_time))
#B = mg.load_dataset('bikewale_clean', 'ID')
B = mg.load_dataset('walmart', 'ID')
b_load_time = time.time()
print("Loading table B --- %s seconds ---" % (b_load_time - a_load_time))
ab = mg.AttrEquivalenceBlocker()
ab_time = time.time()
print("Created an AE blocker --- %s seconds ---" % (ab_time - b_load_time))
#C = ab.block_tables_skd(A, B, 'city_posted', 'city_posted', 'city_posted', 'city_posted')
C = ab.block_tables(A, B, 'pubYear', 'pubYear', 'pubYear', 'pubYear')
#C = ab.block_tables_skd(A, B, 'isbn', 'isbn', 'isbn', 'isbn')
print("Size of candset C: %d" % (len(C)))
c_time = time.time()
print("Block tables --- %s seconds ---" % (c_time - ab_time))
#s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.city_posted', 'rtable.city_posted'])
s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.pubYear', 'rtable.pubYear'])
#s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.isbn', 'rtable.isbn'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
#k1 = np.array(C[['ltable.city_posted']])
k1 = np.array(C[['ltable.pubYear']])
#k1 = np.array(C[['ltable.isbn']])
#k2 = np.array(C[['rtable.city_posted']])
k2 = np.array(C[['rtable.pubYear']])
#k2 = np.array(C[['rtable.isbn']])
assert_equal(all(k1 == k2), True)
def test_bb_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
C = bb.block_tables(A, B, 'zipcode', 'zipcode')
s1 = sorted(
['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
feature_table = mg.get_features_for_blocking(A, B)
A['dummy'] = 1
B['dummy'] = 1
ab = mg.AttrEquivalenceBlocker()
D = ab.block_tables(A, B, 'dummy', 'dummy')
fv = mg.extract_feature_vecs(D, feature_table=feature_table)
expected_ids = fv.ix[(fv.name_name_mel >= 0.4), ['ltable.ID', 'rtable.ID']]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(
expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(
actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_ab_block_candset_skd():
#A = mg.load_dataset('table_A')
A = mg.load_dataset('bikedekho_clean', 'ID')
#B = mg.load_dataset('table_B')
B = mg.load_dataset('bikewale_clean', 'ID')
ab = mg.AttrEquivalenceBlocker()
#C = ab.block_tables(A, B, 'zipcode', 'zipcode', ['zipcode', 'birth_year'], ['zipcode', 'birth_year'])
C = ab.block_tables_opt(A, B, 'city_posted', 'city_posted',
['bike_name', 'city_posted', 'km_driven', 'price', 'color', 'model_year'],
['bike_name', 'city_posted', 'km_driven', 'price', 'color', 'model_year'])
print "Size of C: ", len(C)
#D = ab.block_candset_skd(C, 'birth_year', 'birth_year')
D = ab.block_candset_joblib(C, 'model_year', 'model_year')
print "Size of D: ", len(D)
#s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'ltable.birth_year', 'rtable.zipcode',
# 'rtable.birth_year'])
s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.bike_name', 'ltable.city_posted',
'ltable.km_driven', 'ltable.price', 'ltable.color', 'ltable.model_year',
'rtable.bike_name', 'rtable.city_posted', 'rtable.km_driven', 'rtable.price',
'rtable.color', 'rtable.model_year'])
assert_equal(s1, sorted(D.columns))
assert_equal(D.get_key(), '_id')
assert_equal(D.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(D.get_property('foreign_key_rtable'), 'rtable.ID')
#k1 = np.array(D[['ltable.birth_year']])
k1 = np.array(D[['ltable.model_year']])
#k2 = np.array(D[['rtable.birth_year']])
k2 = np.array(D[['rtable.model_year']])
assert_equal(all(k1 == k2), True)
def test_save_load_rb_blocker():
#mg.init_jvm()
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
feature_table = mg.get_features_for_blocking(A, B)
rb0 = mg.RuleBasedBlocker()
rb0.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) == 1'], feature_table)
rb0.add_rule(['birth_year_birth_year_anm(ltuple, rtuple) > 0.95',
'name_name_mel(ltuple, rtuple)> 0.4'],
feature_table)
C0 = rb0.block_tables(A, B)
mg.save_object(rb0, filename)
rb1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(rb0), type(rb1))
assert_equal(len(rb0.rules), len(rb1.rules))
assert_equal(len(rb0.rule_source), len(rb1.rule_source))
assert_equal(rb0.rule_cnt, rb1.rule_cnt)
C1 = rb1.block_tables(A, B)
assert_equal(len(C0), len(C1))
assert_equal(sorted(C0.columns), sorted(C0.columns))
def test_save_load_rb_blocker():
#mg.init_jvm()
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
feature_table = mg.get_features_for_blocking(A, B)
rb0 = mg.RuleBasedBlocker()
rb0.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) == 1'], feature_table)
rb0.add_rule([
'birth_year_birth_year_anm(ltuple, rtuple) > 0.95',
'name_name_mel(ltuple, rtuple)> 0.4'
], feature_table)
C0 = rb0.block_tables(A, B)
mg.save_object(rb0, filename)
rb1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(rb0), type(rb1))
assert_equal(len(rb0.rules), len(rb1.rules))
assert_equal(len(rb0.rule_source), len(rb1.rule_source))
assert_equal(rb0.rule_cnt, rb1.rule_cnt)
C1 = rb1.block_tables(A, B)
assert_equal(len(C0), len(C1))
assert_equal(sorted(C0.columns), sorted(C0.columns))
def test_save_load_bb_blocker():
#mg.init_jvm('/Library/Java/JavaVirtualMachines/jdk1.8.0_45.jdk/Contents/Home/jre/lib/server/libjvm.dylib')
from magellan.feature.simfunctions import jaccard
from magellan.feature.tokenizers import tok_qgram
def block_fn_1(ltuple, rtuple):
val = jaccard(tok_qgram(ltuple['address'], 3),
tok_qgram(rtuple['address'], 3))
if val < 0.4:
return True
else:
return False
bb0 = mg.BlackBoxBlocker()
bb0.set_black_box_function(block_fn_1)
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
C0 = bb0.block_tables(A, B)
mg.save_object(bb0, filename)
bb1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(bb0), type(bb1))
C1 = bb1.block_tables(A, B)
assert_equal(len(C0), len(C1))
assert_equal(sorted(C0.columns), sorted(C0.columns))
def test_ab_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
assert_equal(ab.block_tuples(A.ix[1], B.ix[2], 'zipcode', 'zipcode'),
False)
assert_equal(ab.block_tuples(A.ix[2], B.ix[2], 'zipcode', 'zipcode'), True)
def test_rb_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule([
'name_name_mel(ltuple, rtuple) < 0.4',
'birth_year_birth_year_lev(ltuple, rtuple) < 0.5'
], feature_table)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) != 1'], feature_table)
C = rb.block_tables(A, B, 'zipcode', 'zipcode')
s1 = sorted(
['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
A['dummy'] = 1
B['dummy'] = 1
ab = mg.AttrEquivalenceBlocker()
D = ab.block_tables(A, B, 'dummy', 'dummy')
fv = mg.extract_feature_vecs(D, feature_table=feature_table)
expected_ids = fv.ix[((fv.name_name_mel >= 0.4) |
(fv.birth_year_birth_year_lev >= 0.5))
& fv.zipcode_zipcode_exm == 1, [
'ltable.ID',
'rtable.ID',
]]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(
expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(
actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_save_load_feature_table():
#mg.init_jvm()
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
feature_table0 = mg.get_features_for_blocking(A, B)
mg.save_object(feature_table0, filename)
feature_table1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(feature_table0), type(feature_table1))
assert_equal(len(feature_table0), len(feature_table1))
assert sorted(feature_table0.columns) == sorted(feature_table1.columns)
ft0_functions = list(feature_table0['function'])
ft1_functions = list(feature_table1['function'])
for f0, f1 in zip(ft0_functions, ft1_functions):
a = f0(A.ix[1], B.ix[2])
b = f1(A.ix[1], B.ix[2])
if math.isnan(a) == False and math.isnan(b) == False:
assert_equal(a, b)
if math.isnan(a) == True:
assert_equal(math.isnan(b), True)
def test_bb_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
assert_equal(bb.block_tuples(A.ix[0], B.ix[0]), True)
assert_equal(bb.block_tuples(A.ix[2], B.ix[1]), False)
def test_save_load_bb_blocker():
#mg.init_jvm('/Library/Java/JavaVirtualMachines/jdk1.8.0_45.jdk/Contents/Home/jre/lib/server/libjvm.dylib')
from magellan.feature.simfunctions import jaccard
from magellan.feature.tokenizers import tok_qgram
def block_fn_1(ltuple, rtuple):
val = jaccard(tok_qgram(ltuple['address'], 3), tok_qgram(rtuple['address'], 3))
if val < 0.4:
return True
else:
return False
bb0 = mg.BlackBoxBlocker()
bb0.set_black_box_function(block_fn_1)
filename = '__mg_obj__.pkl'
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
C0 = bb0.block_tables(A, B)
mg.save_object(bb0, filename)
bb1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(bb0), type(bb1))
C1 = bb1.block_tables(A, B)
assert_equal(len(C0), len(C1))
assert_equal(sorted(C0.columns), sorted(C0.columns))
def test_bb_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
assert_equal(bb.block_tuples(A.ix[0], B.ix[0]), True)
assert_equal(bb.block_tuples(A.ix[2], B.ix[1]), False)
def test_ab_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'name', 'name')
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_ab_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'name', 'name')
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_bb_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(evil_block_fn)
C = bb.block_tables(A, B)
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_bb_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(evil_block_fn)
C = bb.block_tables(A, B)
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_rb_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) >= 0'], feature_table)
C = rb.block_tables(A, B)
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_rb_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['name_name_mel(ltuple, rtuple) < 0.4',
'birth_year_birth_year_lev(ltuple, rtuple) < 0.5'],
feature_table)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) != 1'],
feature_table)
assert_equal(rb.block_tuples(A.ix[0], B.ix[0]), False)
assert_equal(rb.block_tuples(A.ix[1], B.ix[1]), True)
def test_bb_block_candset_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'birth_year', 'birth_year')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(evil_block_fn)
D = bb.block_candset(C)
assert_equal(len(D), 0)
assert_equal(sorted(D.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(D.get_key(), '_id')
assert_equal(D.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(D.get_property('foreign_key_rtable'), 'rtable.ID')
def test_bb_block_candset_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'birth_year', 'birth_year')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(evil_block_fn)
D = bb.block_candset(C)
assert_equal(len(D), 0)
assert_equal(sorted(D.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(D.get_key(), '_id')
assert_equal(D.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(D.get_property('foreign_key_rtable'), 'rtable.ID')
def test_ab_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'zipcode', 'zipcode', 'zipcode', 'zipcode')
s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
k1 = np.array(C[['ltable.zipcode']])
k2 = np.array(C[['rtable.zipcode']])
assert_equal(all(k1 == k2), True)
def test_rb_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule([
'name_name_mel(ltuple, rtuple) < 0.4',
'birth_year_birth_year_lev(ltuple, rtuple) < 0.5'
], feature_table)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) != 1'], feature_table)
assert_equal(rb.block_tuples(A.ix[0], B.ix[0]), False)
assert_equal(rb.block_tuples(A.ix[1], B.ix[1]), True)
def test_rb_block_tables_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) >= 0'],
feature_table)
C = rb.block_tables(A, B)
assert_equal(len(C), 0)
assert_equal(sorted(C.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
def test_ab_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'zipcode', 'zipcode', 'zipcode', 'zipcode')
s1 = sorted(
['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
k1 = np.array(C[['ltable.zipcode']])
k2 = np.array(C[['rtable.zipcode']])
assert_equal(all(k1 == k2), True)
def test_rb_block_candset_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'birth_year', 'birth_year')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) >= 0'], feature_table)
D = rb.block_candset(C)
assert_equal(len(D), 0)
assert_equal(sorted(D.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(D.get_key(), '_id')
assert_equal(D.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(D.get_property('foreign_key_rtable'), 'rtable.ID')
def test_rb_block_candset_wi_no_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
C = ab.block_tables(A, B, 'birth_year', 'birth_year')
rb = mg.RuleBasedBlocker()
feature_table = mg.get_features_for_blocking(A, B)
rb.add_rule(['zipcode_zipcode_exm(ltuple, rtuple) >= 0'],
feature_table)
D = rb.block_candset(C)
assert_equal(len(D), 0)
assert_equal(sorted(D.columns), sorted(['_id', 'ltable.ID', 'rtable.ID']))
assert_equal(D.get_key(), '_id')
assert_equal(D.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(D.get_property('foreign_key_rtable'), 'rtable.ID')
def test_bb_block_candset():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
E = ab.block_tables(A, B, 'zipcode', 'zipcode')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
C = bb.block_candset(E)
feature_table = mg.get_features_for_blocking(A, B)
fv = mg.extract_feature_vecs(C, feature_table=feature_table)
expected_ids = fv.ix[(fv.name_name_mel >= 0.4) ,
['ltable.ID', 'rtable.ID']]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_bb_block_candset():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
E = ab.block_tables(A, B, 'zipcode', 'zipcode')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
C = bb.block_candset(E)
feature_table = mg.get_features_for_blocking(A, B)
fv = mg.extract_feature_vecs(C, feature_table=feature_table)
expected_ids = fv.ix[(fv.name_name_mel >= 0.4), ['ltable.ID', 'rtable.ID']]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(
expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(
actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_bb_block_tables():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
bb = mg.BlackBoxBlocker()
bb.set_black_box_function(block_fn)
C = bb.block_tables(A, B, 'zipcode', 'zipcode')
s1 = sorted(['_id', 'ltable.ID', 'rtable.ID', 'ltable.zipcode', 'rtable.zipcode'])
assert_equal(s1, sorted(C.columns))
assert_equal(C.get_key(), '_id')
assert_equal(C.get_property('foreign_key_ltable'), 'ltable.ID')
assert_equal(C.get_property('foreign_key_rtable'), 'rtable.ID')
feature_table = mg.get_features_for_blocking(A, B)
A['dummy'] = 1
B['dummy'] = 1
ab = mg.AttrEquivalenceBlocker()
D = ab.block_tables(A, B, 'dummy','dummy')
fv = mg.extract_feature_vecs(D, feature_table=feature_table)
expected_ids = fv.ix[(fv.name_name_mel >= 0.4) ,
['ltable.ID', 'rtable.ID']]
actual_ids = C[['ltable.ID', 'rtable.ID']]
ids_exp = list(expected_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
ids_act = list(actual_ids.set_index(['ltable.ID', 'rtable.ID']).index.values)
assert_equal(cmp(ids_exp, ids_act), 0)
def test_ab_block_tuples():
A = mg.load_dataset('table_A')
B = mg.load_dataset('table_B')
ab = mg.AttrEquivalenceBlocker()
assert_equal(ab.block_tuples(A.ix[1], B.ix[2], 'zipcode', 'zipcode'), False)
assert_equal(ab.block_tuples(A.ix[2], B.ix[2], 'zipcode', 'zipcode'), True)
