def test_init(self):
self.assertRaises(TypeError, _tsinfer.AncestorBuilder)
for bad_value in [None, "serf", [[], []], ["asdf"], {}]:
with self.assertRaises(TypeError):
_tsinfer.AncestorBuilder(num_samples=2, max_sites=bad_value)
_tsinfer.AncestorBuilder(num_samples=bad_value, max_sites=2)
for bad_num_samples in [0, 1]:
with self.assertRaises(_tsinfer.LibraryError):
_tsinfer.AncestorBuilder(num_samples=bad_num_samples,
max_sites=0)
def test_add_too_many_sites(self):
for max_sites in range(10):
ab = _tsinfer.AncestorBuilder(num_samples=2, max_sites=max_sites)
for j in range(max_sites):
ab.add_site(time=1, genotypes=[0, 1])
for j in range(2 * max_sites):
with pytest.raises(_tsinfer.LibraryError):
ab.add_site(time=1, genotypes=[0, 1])
def test_add_site(self):
ab = _tsinfer.AncestorBuilder(num_samples=2, max_sites=10)
for bad_type in ["sdf", {}, None]:
with pytest.raises(TypeError):
ab.add_site(time=bad_type, genotypes=[0, 0])
for bad_genotypes in ["asdf", [[], []], [0, 1, 2]]:
with pytest.raises(ValueError):
ab.add_site(time=0, genotypes=bad_genotypes)
def test_add_too_many_sites(self):
for max_sites in range(10):
ab = _tsinfer.AncestorBuilder(num_samples=2, max_sites=max_sites)
for _ in range(max_sites):
ab.add_site(time=1, genotypes=[0, 1])
for _ in range(2 * max_sites):
with pytest.raises(_tsinfer.LibraryError) as record:
ab.add_site(time=1, genotypes=[0, 1])
msg = "Cannot add more sites than the specified maximum."
assert str(record.value) == msg
def build_ancestors(input_data, ancestor_data, progress=False, method="C"):
num_sites = input_data.num_variant_sites
num_samples = input_data.num_samples
if method == "C":
logger.debug("Using C AncestorBuilder implementation")
ancestor_builder = _tsinfer.AncestorBuilder(num_samples, num_sites)
else:
logger.debug("Using Python AncestorBuilder implementation")
ancestor_builder = algorithm.AncestorBuilder(num_samples, num_sites)
progress_monitor = tqdm.tqdm(total=num_sites, disable=not progress)
frequency = input_data.frequency[:]
logger.info("Starting site addition")
for j, (site_id, genotypes) in enumerate(input_data.variants()):
ancestor_builder.add_site(j, int(frequency[site_id]), genotypes)
progress_monitor.update()
progress_monitor.close()
logger.info("Finished adding sites")
descriptors = ancestor_builder.ancestor_descriptors()
if len(descriptors) > 0:
num_ancestors = len(descriptors)
logger.info("Starting build for {} ancestors".format(num_ancestors))
a = np.zeros(num_sites, dtype=np.uint8)
root_time = descriptors[0][0] + 1
ultimate_ancestor_time = root_time + 1
# Add the ultimate ancestor. This is an awkward hack really; we don't
# ever insert this ancestor. The only reason to add it here is that
# it makes sure that the ancestor IDs we have in the ancestor file are
# the same as in the ancestor tree sequence. This seems worthwhile.
ancestor_data.add_ancestor(start=0,
end=num_sites,
time=ultimate_ancestor_time,
focal_sites=[],
haplotype=a)
# Hack to ensure we always have a root with zeros at every position.
ancestor_data.add_ancestor(start=0,
end=num_sites,
time=root_time,
focal_sites=np.array([], dtype=np.int32),
haplotype=a)
progress_monitor = tqdm.tqdm(total=len(descriptors),
disable=not progress)
for freq, focal_sites in descriptors:
before = time.perf_counter()
# TODO: This is a read-only process so we can multithread it.
s, e = ancestor_builder.make_ancestor(focal_sites, a)
assert np.all(a[s:e] != UNKNOWN_ALLELE)
assert np.all(a[:s] == UNKNOWN_ALLELE)
assert np.all(a[e:] == UNKNOWN_ALLELE)
duration = time.perf_counter() - before
logger.debug(
"Made ancestor with {} focal sites and length={} in {:.2f}s.".
format(focal_sites.shape[0], e - s, duration))
ancestor_data.add_ancestor(start=s,
end=e,
time=freq,
focal_sites=focal_sites,
haplotype=a)
progress_monitor.update()
progress_monitor.close()
logger.info("Finished building ancestors")