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_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_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_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_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_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_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_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)
