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_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)
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_save_load_ab_blocker():
filename = '__mg_obj__.pkl'
ab0 = mg.AttrEquivalenceBlocker()
mg.save_object(ab0, filename)
ab1 = mg.load_object(filename)
try:
os.remove(filename)
except OSError:
pass
assert_equal(type(ab0), type(ab1))
def test_bl_combiner_wi_no_tuples():
A = mg.read_csv(path_for_A, key='ID')
B = mg.read_csv(path_for_B, key='ID')
ab = mg.AttrEquivalenceBlocker()
C1 = ab.block_tables(A, B, 'name', 'name')
C = mg.combine_block_outputs_via_union([C1, C1, C1])
s = ['_id', 'ltable.ID', 'rtable.ID']
assert_equal(sorted(s) == sorted(C.columns), True)
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_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_bl_combiner_wi_no_fields():
A = mg.read_csv(path_for_A, key='ID')
B = mg.read_csv(path_for_B, key='ID')
ab = mg.AttrEquivalenceBlocker()
C1 = ab.block_tables(A, B, 'zipcode', 'zipcode')
C2 = ab.block_tables(A, B, 'hourly_wage', 'hourly_wage')
C3 = ab.block_tables(A, B, 'birth_year', 'birth_year')
C = mg.combine_block_outputs_via_union([C1, C2, C3])
s = ['_id', 'ltable.ID', 'rtable.ID']
assert_equal(sorted(s) == sorted(C.columns), True)
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_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_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_tocsv_vtable_wi_props():
filename = '__tbl__.csv'
A = mg.read_csv(path_for_A, key='ID')
B = mg.read_csv(path_for_B, key='ID')
ab = mg.AttrEquivalenceBlocker()
C1 = ab.block_tables(A, B, 'zipcode', 'zipcode')
C1.to_csv(filename, suppress_properties=False)
C2 = mg.read_csv(filename, ltable=A, rtable=B)
try:
os.remove(filename)
except OSError:
pass
assert_equal(C2.get_key(), C1.get_key())
assert_equal(C2.get_property('foreign_key_ltable'), C1.get_property('foreign_key_ltable'))
assert_equal(C2.get_property('foreign_key_rtable'), C1.get_property('foreign_key_rtable'))
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)