# Import recordlinkage
import recordlinkage
# Create indexing object
indexer = recordlinkage.Index()
# Generate pairs blocked on state
indexer.block('state')

# Generate the pairs
pairs = indexer.index(census_A, census_B)

# Create a Compare object
compare_cl = recordlinkage.Compare()

# Find exact matches for pairs of date_of_birth and state
compare_cl.exact('date_of_birth', 'date_of_birth', label='date_of_birth')compare_cl.exact('state', 'state', label='state')

# Find similar matches for pairs of surname and address_1 using string similarity
compare_cl.string('surname', 'surname', threshold=0.85, label='surname')compare_cl.string('address_1', 'address_1', threshold=0.85, label='address_1')

# Find matches
potential_matches = compare_cl.compute(pairs, census_A, census_B)

matches = potential_matches[potential_matches.sum(axis = 1) >= 3]


# Finding duplicates in census_B
census_B_duplicates = census_B[census_B.index.isin(duplicate_rows)]

# Finding new rows in census_B
census_B_new = census_B[~census_B.index.isin(duplicate_rows)]
# Link the DataFrames!
full_census = census_A.append(census_B_new)