def test_groups_correctness(requests):
# We assume that anonlink computes the right results.
with open(DATA_PATH, 'rb') as f:
# Here's some filters I prepared earlier.
filters = pickle.load(f)
candidate_pairs = anonlink.candidate_generation.find_candidate_pairs(
filters, anonlink.similarities.dice_coefficient_accelerated, THRESHOLD)
true_groups = anonlink.solving.greedy_solve(candidate_pairs)
filter_size = len(filters[0][0])
assert all(
len(filter_) == filter_size for dataset in filters
for filter_ in dataset)
packed_filters = [
b''.join(binary_pack_filters(f, filter_size)) for f in filters
]
project_data, _ = create_project_upload_data(requests,
packed_filters,
result_type='groups',
binary=True,
hash_size=DATA_HASH_SIZE)
try:
run = post_run(requests, project_data, threshold=THRESHOLD)
result_groups = get_run_result(requests, project_data, run)['groups']
finally:
delete_project(requests, project_data)
# Compare ES result with anonlink.
result_group_set = {frozenset(map(tuple, g)) for g in result_groups}
true_group_set = set(map(frozenset, true_groups))
assert result_group_set == true_group_set
def project(request, requests, result_type_number_parties):
result_type, number_parties = result_type_number_parties
project = create_project_no_data(requests,
result_type=result_type,
number_parties=number_parties)
yield project
# Release project resource
delete_project(requests, project)
def test_delete_project_after_creating_run_with_clks(
requests, result_type_number_parties):
result_type, number_parties = result_type_number_parties
project, _ = create_project_upload_fake_data(requests,
[100] * number_parties,
overlap=0.5,
result_type=result_type)
post_run(requests, project, 0.9)
delete_project(requests, project)
def test_similarity_scores(requests, the_truth):
project_data, _, _ = create_project_upload_data(requests, the_truth['clks_a'], the_truth['clks_b'],
result_type='similarity_scores')
run = post_run(requests, project_data, threshold=the_truth['threshold'])
result = get_run_result(requests, project_data, run, timeout=60)
# compare the result with the truth
ss = result['similarity_scores']
ts = the_truth['similarity_scores']
assert len(ss) == len(ts)
for es_score, true_score in zip(ss, ts):
assert es_score[0] == true_score[0] and es_score[1] == true_score[2]
assert es_score[2] == pytest.approx(true_score[1], 1e-10), 'similarity scores are different'
delete_project(requests, project_data)
def test_similarity_scores(requests, the_truth):
project_data, _ = create_project_upload_data(
requests, (the_truth['clks_a'], the_truth['clks_b']),
result_type='similarity_scores')
run = post_run(requests, project_data, threshold=the_truth['threshold'])
result = get_run_result(requests, project_data, run, timeout=60)
true_scores = the_truth['similarity_scores']
result_scores = {(a, b): sim for a, b, sim in result['similarity_scores']}
# Anonlink is more strict on enforcing the k parameter. Hence the
# subset.
assert true_scores.keys() <= result_scores.keys()
for pair in true_scores:
assert true_scores[pair] == result_scores[pair]
delete_project(requests, project_data)
def mapping_project(request, requests):
size, overlap, encoding_size = request.param
prj = create_project_response(requests, size, overlap, 'mapping',
encoding_size)
yield prj
delete_project(requests, prj)
def permutations_project(request, requests):
size, overlap, encoding_size = request.param
prj = create_project_response(requests, size, overlap, 'permutations',
encoding_size)
yield prj
delete_project(requests, prj)
def project(request, requests):
result_type = request.param
project = create_project_no_data(requests, result_type)
yield project
# Release project resource
delete_project(requests, project)
def similarity_scores_project(request, requests):
size, overlap, encoding_size = request.param
prj = create_project_response(requests, size, overlap, 'similarity_scores',
encoding_size)
yield prj
delete_project(requests, prj)
def test_delete_project_types(requests, project):
delete_project(requests, project)
def test_delete_project_after_creating_run_with_clks(requests, result_type):
project, dp1, dp2 = create_project_upload_fake_data(
requests, [100, 100], overlap=0.5, result_type=result_type)
post_run(requests, project, 0.9)
delete_project(requests, project)