def test_load_cache(self) -> None: cache_element = DataMemoryElement() i1 = LinearHashIndex(cache_element) # noinspection PyTypeChecker i1.build_index([[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]) # load called on initialization. i2 = LinearHashIndex(cache_element) self.assertEqual(i1.cache_element, i2.cache_element) self.assertEqual(i1.index, i2.index)
def test_build_index_with_cache(self) -> None: cache_element = DataMemoryElement() i = LinearHashIndex(cache_element) # noinspection PyTypeChecker i.build_index([[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]) self.assertEqual(i.index, {1, 2, 3, 4}) self.assertFalse(cache_element.is_empty())
def test_update_index_with_hash_index(self) -> None: # Similar test to `test_update_index_add_new_descriptors` but with a # linear hash index. descr_set = MemoryDescriptorSet() hash_kvs = MemoryKeyValueStore() linear_hi = LinearHashIndex() # simplest hash index, heap-sorts. index = LSHNearestNeighborIndex(DummyHashFunctor(), descr_set, hash_kvs, linear_hi) descriptors1 = [ DescriptorMemoryElement('t', 0), DescriptorMemoryElement('t', 1), DescriptorMemoryElement('t', 2), DescriptorMemoryElement('t', 3), DescriptorMemoryElement('t', 4), ] descriptors2 = [ DescriptorMemoryElement('t', 5), DescriptorMemoryElement('t', 6), ] # Vectors of length 1 for easy dummy hashing prediction. for d in descriptors1 + descriptors2: d.set_vector(np.ones(1, float) * d.uuid()) # Build initial index. index.build_index(descriptors1) # Initial hash index should only encode hashes for first batch of # descriptors. self.assertSetEqual(linear_hi.index, {0, 1, 2, 3, 4}) # Update index and check that components have new data. index.update_index(descriptors2) # Now the hash index should include all descriptor hashes. self.assertSetEqual(linear_hi.index, {0, 1, 2, 3, 4, 5, 6})
def test_remove_from_index(self) -> None: # Test that actual removal occurs. i = LinearHashIndex() i.index = {0, 1, 2} # noinspection PyTypeChecker i.remove_from_index([[0, 0], [1, 0]]) self.assertSetEqual(i.index, {1})
def test_save_cache_remove_from_index(self) -> None: # Test that the cache is updated appropriately on a removal. cache_element = DataMemoryElement() self.assertTrue(cache_element.is_empty()) i = LinearHashIndex(cache_element) # noinspection PyTypeChecker i.build_index([ [0, 1, 0], # 2 [0, 1, 1], # 3 [1, 0, 0], # 4 [1, 1, 0] ]) # 6 self.assertFalse(cache_element.is_empty()) self.assertSetEqual( set(numpy.load(BytesIO(cache_element.get_bytes()))), {2, 3, 4, 6}) # noinspection PyTypeChecker i.remove_from_index([ [0, 1, 1], # 3 [1, 0, 0] ]) # 4 self.assertFalse(cache_element.is_empty()) self.assertSetEqual( set(numpy.load(BytesIO(cache_element.get_bytes()))), {2, 6})
def test_update_index_no_index(self) -> None: # Test calling update index with no existing index. Should result the # same as calling build_index with no index. i = LinearHashIndex() # noinspection PyTypeChecker i.update_index([[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]) self.assertEqual(i.index, {1, 2, 3, 4}) self.assertIsNone(i.cache_element)
def test_remove_from_index_single_not_in_index(self) -> None: # Test attempting to remove single hash not in the index. i = LinearHashIndex() i.index = {0, 1, 2} self.assertRaises( KeyError, i.remove_from_index, [[1, 0, 0]] # 4 ) self.assertSetEqual(i.index, {0, 1, 2})
def test_update_index_add_hashes(self) -> None: i = LinearHashIndex() # Build index with some initial hashes # noinspection PyTypeChecker i.build_index([[0, 0], [0, 1]]) self.assertSetEqual(i.index, {0, 1}) # Update index with new stuff # noinspection PyTypeChecker i.update_index([[1, 0], [1, 1]]) self.assertSetEqual(i.index, {0, 1, 2, 3})
def test_get_config(self) -> None: i = LinearHashIndex() # Without cache element expected_c = LinearHashIndex.get_default_config() self.assertEqual(i.get_config(), expected_c) # With cache element i.cache_element = DataMemoryElement() expected_c['cache_element'][ 'type'] = 'smqtk_dataprovider.impls.data_element.memory.DataMemoryElement' self.assertEqual(i.get_config(), expected_c)
def test_save_cache_build_index(self) -> None: cache_element = DataMemoryElement() self.assertTrue(cache_element.is_empty()) i = LinearHashIndex(cache_element) # noinspection PyTypeChecker i.build_index([[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]) self.assertFalse(cache_element.is_empty()) # Check byte content expected_cache = {1, 2, 3, 4} actual_cache = set(numpy.load(BytesIO(cache_element.get_bytes()))) self.assertSetEqual(expected_cache, actual_cache)
def test_nn(self) -> None: i = LinearHashIndex() # noinspection PyTypeChecker i.build_index([[0, 1, 0], [1, 1, 0], [0, 1, 1], [0, 0, 1]]) # noinspection PyTypeChecker near_codes, near_dists = i.nn([0, 0, 0], 4) self.assertEqual(set(map(tuple, near_codes[:2])), {(0, 1, 0), (0, 0, 1)}) self.assertEqual(set(map(tuple, near_codes[2:])), {(1, 1, 0), (0, 1, 1)}) numpy.testing.assert_array_almost_equal( near_dists, (1 / 3., 1 / 3., 2 / 3., 2 / 3.))
def test_configuration(self) -> None: i = LSHNearestNeighborIndex( lsh_functor=ItqFunctor(), descriptor_set=MemoryDescriptorSet(), hash2uuids_kvstore=MemoryKeyValueStore(), hash_index=LinearHashIndex(), distance_method='euclidean', read_only=True ) for inst in configuration_test_helper(i): # type: LSHNearestNeighborIndex assert isinstance(inst.lsh_functor, LshFunctor) assert isinstance(inst.descriptor_set, MemoryDescriptorSet) assert isinstance(inst.hash_index, LinearHashIndex) assert isinstance(inst.hash2uuids_kvstore, MemoryKeyValueStore) assert inst.distance_method == 'euclidean' assert inst.read_only is True
def test_remove_from_index_one_of_many_not_in_index(self) -> None: # Test attempting to remove hashes where one of them is not in the # index. i = LinearHashIndex() i.index = {0, 1, 2} self.assertRaises( KeyError, i.remove_from_index, [ [0, 0], # 0 [0, 1], # 1 [1, 1] ] # 3 ) # Check that the index has not been modified. self.assertSetEqual(i.index, {0, 1, 2})
def test_build_index_fresh_build_with_hash_index(self) -> None: descr_set = MemoryDescriptorSet() hash_kvs = MemoryKeyValueStore() linear_hi = LinearHashIndex() # simplest hash index, heap-sorts. index = LSHNearestNeighborIndex(DummyHashFunctor(), descr_set, hash_kvs, linear_hi) descriptors = [ DescriptorMemoryElement('t', 0), DescriptorMemoryElement('t', 1), DescriptorMemoryElement('t', 2), DescriptorMemoryElement('t', 3), DescriptorMemoryElement('t', 4), ] # Vectors of length 1 for easy dummy hashing prediction. for i, d in enumerate(descriptors): d.set_vector(np.ones(1, float) * i) index.build_index(descriptors) # Hash index should have been built with hash vectors, and linearHI # converts those to integers for storage. self.assertEqual(linear_hi.index, {0, 1, 2, 3, 4})
def _nn( self, d: DescriptorElement, n: int = 1 ) -> Tuple[Tuple[DescriptorElement, ...], Tuple[float, ...]]: """ Internal method to be implemented by sub-classes to return the nearest `N` neighbors to the given descriptor element. When this internal method is called, we have already checked that there is a vector in ``d`` and our index is not empty. :param d: Descriptor element to compute the neighbors of. :param n: Number of nearest neighbors to find. :return: Tuple of nearest N DescriptorElement instances, and a tuple of the distance values to those neighbors. """ LOG.debug("generating hash for descriptor") d_v = d.vector() d_h = self.lsh_functor.get_hash(d_v) def comp_descr_dist(d2_v: numpy.ndarray) -> float: return self._distance_function(d_v, d2_v) with self._model_lock: LOG.debug("getting near hashes") hi = self.hash_index if hi is None: # Make on-the-fly linear index hi = LinearHashIndex() # not calling ``build_index`` because we already have the int # hashes. hi.index = set(cast(Iterator[int], self.hash2uuids_kvstore.keys())) near_hashes, _ = hi.nn(d_h, n) LOG.debug("getting UUIDs of descriptors for nearby hashes") neighbor_uuids: List[Hashable] = [] for h_int in map(bit_vector_to_int_large, near_hashes): # If descriptor hash not in our map, we effectively skip it. # Get set of descriptor UUIDs for a hash code. near_uuids: Set[Hashable] = self.hash2uuids_kvstore.get(h_int, set()) # Accumulate matching descriptor UUIDs to a list. neighbor_uuids.extend(near_uuids) LOG.debug("-- matched %d UUIDs", len(neighbor_uuids)) LOG.debug("getting descriptors for neighbor_uuids") neighbors = \ list(self.descriptor_set.get_many_descriptors(neighbor_uuids)) # Done with model parts at this point, so releasing lock. LOG.debug(f"ordering descriptors via distance method {self.distance_method}") LOG.debug('-- getting element vectors') neighbor_vectors = numpy.asarray(list( parallel_map(lambda d_: d_.vector(), neighbors) )) LOG.debug('-- calculating distances') distances = list(map(comp_descr_dist, neighbor_vectors)) LOG.debug('-- ordering') ordered = sorted(zip(neighbors, distances), key=lambda p: p[1]) LOG.debug(f'-- slicing top n={n}') r_descrs: Tuple[DescriptorElement, ...] r_dists: Tuple[float, ...] r_descrs, r_dists = zip(*(ordered[:n])) return r_descrs, r_dists
def test_update_index_no_input(self) -> None: i = LinearHashIndex() self.assertRaises(ValueError, i.update_index, [])
def _make_hi_linear(self) -> LinearHashIndex: return LinearHashIndex()
def test_build_index_no_cache(self) -> None: i = LinearHashIndex() # noinspection PyTypeChecker i.build_index([[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]]) self.assertEqual(i.index, {1, 2, 3, 4}) self.assertIsNone(i.cache_element)
def test_save_cache_readonly_update_index(self) -> None: ro_cache = DataMemoryElement(readonly=True) i = LinearHashIndex(ro_cache) self.assertRaisesRegex(ValueError, "is read-only", i.update_index, [[0, 1, 0], [1, 0, 0], [0, 1, 1], [0, 0, 1]])