def test_aggregate2(self):
schema = hl.tstruct(status=hl.tint32, GT=hl.tcall, qPheno=hl.tint32)
rows = [{'status': 0, 'GT': hl.Call([0, 0]), 'qPheno': 3},
{'status': 0, 'GT': hl.Call([0, 1]), 'qPheno': 13}]
kt = hl.Table.parallelize(rows, schema)
result = convert_struct_to_dict(
kt.group_by(status=kt.status)
.aggregate(
x1=agg.collect(kt.qPheno * 2),
x2=agg.explode(lambda elt: agg.collect(elt), [kt.qPheno, kt.qPheno + 1]),
x3=agg.min(kt.qPheno),
x4=agg.max(kt.qPheno),
x5=agg.sum(kt.qPheno),
x6=agg.product(hl.int64(kt.qPheno)),
x7=agg.count(),
x8=agg.count_where(kt.qPheno == 3),
x9=agg.fraction(kt.qPheno == 1),
x10=agg.stats(hl.float64(kt.qPheno)),
x11=agg.hardy_weinberg_test(kt.GT),
x13=agg.inbreeding(kt.GT, 0.1),
x14=agg.call_stats(kt.GT, ["A", "T"]),
x15=agg.collect(hl.Struct(a=5, b="foo", c=hl.Struct(banana='apple')))[0],
x16=agg.collect(hl.Struct(a=5, b="foo", c=hl.Struct(banana='apple')).c.banana)[0],
x17=agg.explode(lambda elt: agg.collect(elt), hl.null(hl.tarray(hl.tint32))),
x18=agg.explode(lambda elt: agg.collect(elt), hl.null(hl.tset(hl.tint32))),
x19=agg.take(kt.GT, 1, ordering=-kt.qPheno)
).take(1)[0])
expected = {u'status': 0,
u'x13': {u'n_called': 2, u'expected_homs': 1.64, u'f_stat': -1.777777777777777,
u'observed_homs': 1},
u'x14': {u'AC': [3, 1], u'AF': [0.75, 0.25], u'AN': 4, u'homozygote_count': [1, 0]},
u'x15': {u'a': 5, u'c': {u'banana': u'apple'}, u'b': u'foo'},
u'x10': {u'min': 3.0, u'max': 13.0, u'sum': 16.0, u'stdev': 5.0, u'n': 2, u'mean': 8.0},
u'x8': 1, u'x9': 0.0, u'x16': u'apple',
u'x11': {u'het_freq_hwe': 0.5, u'p_value': 0.5},
u'x2': [3, 4, 13, 14], u'x3': 3, u'x1': [6, 26], u'x6': 39, u'x7': 2, u'x4': 13, u'x5': 16,
u'x17': [],
u'x18': [],
u'x19': [hl.Call([0, 1])]}
self.maxDiff = None
self.assertDictEqual(result, expected)
def test_aggregate2(self):
schema = hl.tstruct(status=hl.tint32, GT=hl.tcall, qPheno=hl.tint32)
rows = [{'status': 0, 'GT': hl.Call([0, 0]), 'qPheno': 3},
{'status': 0, 'GT': hl.Call([0, 1]), 'qPheno': 13}]
kt = hl.Table.parallelize(rows, schema)
result = convert_struct_to_dict(
kt.group_by(status=kt.status)
.aggregate(
x1=agg.collect(kt.qPheno * 2),
x2=agg.explode(lambda elt: agg.collect(elt), [kt.qPheno, kt.qPheno + 1]),
x3=agg.min(kt.qPheno),
x4=agg.max(kt.qPheno),
x5=agg.sum(kt.qPheno),
x6=agg.product(hl.int64(kt.qPheno)),
x7=agg.count(),
x8=agg.count_where(kt.qPheno == 3),
x9=agg.fraction(kt.qPheno == 1),
x10=agg.stats(hl.float64(kt.qPheno)),
x11=agg.hardy_weinberg_test(kt.GT),
x13=agg.inbreeding(kt.GT, 0.1),
x14=agg.call_stats(kt.GT, ["A", "T"]),
x15=agg.collect(hl.Struct(a=5, b="foo", c=hl.Struct(banana='apple')))[0],
x16=agg.collect(hl.Struct(a=5, b="foo", c=hl.Struct(banana='apple')).c.banana)[0],
x17=agg.explode(lambda elt: agg.collect(elt), hl.null(hl.tarray(hl.tint32))),
x18=agg.explode(lambda elt: agg.collect(elt), hl.null(hl.tset(hl.tint32))),
x19=agg.take(kt.GT, 1, ordering=-kt.qPheno)
).take(1)[0])
expected = {u'status': 0,
u'x13': {u'n_called': 2, u'expected_homs': 1.64, u'f_stat': -1.777777777777777,
u'observed_homs': 1},
u'x14': {u'AC': [3, 1], u'AF': [0.75, 0.25], u'AN': 4, u'homozygote_count': [1, 0]},
u'x15': {u'a': 5, u'c': {u'banana': u'apple'}, u'b': u'foo'},
u'x10': {u'min': 3.0, u'max': 13.0, u'sum': 16.0, u'stdev': 5.0, u'n': 2, u'mean': 8.0},
u'x8': 1, u'x9': 0.0, u'x16': u'apple',
u'x11': {u'het_freq_hwe': 0.5, u'p_value': 0.5},
u'x2': [3, 4, 13, 14], u'x3': 3, u'x1': [6, 26], u'x6': 39, u'x7': 2, u'x4': 13, u'x5': 16,
u'x17': [],
u'x18': [],
u'x19': [hl.Call([0, 1])]}
self.maxDiff = None
self.assertDictEqual(result, expected)
# COMMAND ---------- # MAGIC %md `counter` is an aggregation function that counts the number of occurrences of each unique element. We can use this to pull out the population distribution by passing in a Hail Expression for the field that we want to count by. # COMMAND ---------- pprint(table.aggregate(agg.counter(table.SuperPopulation))) # COMMAND ---------- # MAGIC %md `stats` is an aggregation function that produces some useful statistics about numeric collections. We can use this to see the distribution of the CaffeineConsumption phenotype. # COMMAND ---------- pprint(table.aggregate(agg.stats(table.CaffeineConsumption))) # COMMAND ---------- # MAGIC %md However, these metrics aren't perfectly representative of the samples in our dataset. Here's why: # COMMAND ---------- table.count() # COMMAND ---------- mt.count_cols() # COMMAND ----------