def test_grouped_slices(self):
a = nd.asarray([[1, 2, 3], [1, 4, 5]])
gb = nd.groupby(a[:, 1:], a[:, 0])
self.assertEqual(nd.as_py(gb.groups), [1])
self.assertEqual(nd.as_py(gb), [[[2, 3], [4, 5]]])
a = nd.asarray([[1, 2, 3], [3, 1, 7], [1, 4, 5], [2, 6, 7], [3, 2, 5]])
gb = nd.groupby(a[:, 1:], a[:, 0])
self.assertEqual(nd.as_py(gb.groups), [1, 2, 3])
self.assertEqual(nd.as_py(gb), [[[2, 3], [4, 5]],
[[6, 7]],
[[1, 7], [2, 5]]])
def groupby(self, json_cmd):
print('GroupBy operation')
cmd = json.loads(json_cmd)
array_url = cmd.get('input', self.base_url + self.array_name)
if not array_url.startswith(self.base_url):
raise RuntimeError('Input array must start with the base url')
array_name = array_url[len(self.base_url):]
fields = cmd['fields']
arr = self.get_session_array(array_name)[...].ddesc.dynd_arr()
# Do the groupby, get its groups, then
# evaluate it because deferred operations
# through the groupby won't work well yet.
res = nd.groupby(arr, nd.fields(arr, *fields))
groups = res.groups
res = res.eval()
# Write out the groupby result
defarr_gb = self.array_provider.create_deferred_array_filename(
self.session_name, 'groupby_', array(res))
dshape_gb = nd.dshape_of(res)
defarr_gb[0].write(
json.dumps({
'dshape': dshape_gb,
'command': 'groupby',
'params': {
'fields': fields
}
}))
defarr_gb[0].close()
# Write out the groups
defarr_groups = self.array_provider.create_deferred_array_filename(
self.session_name, 'groups_', groups)
dshape_groups = nd.dshape_of(groups)
defarr_groups[0].write(
json.dumps({
'dshape': dshape_groups,
'command': 'groupby.groups',
'params': {
'fields': fields
}
}))
defarr_groups[0].close()
content_type = 'application/json; charset=utf-8'
body = json.dumps({
'session': self.base_url + self.session_name,
'output_gb': self.base_url + defarr_gb[1],
'dshape_gb': dshape_gb,
'output_groups': self.base_url + defarr_groups[1],
'dshape_groups': dshape_groups
})
return (content_type, body)
def groupby(self, json_cmd):
print('GroupBy operation')
cmd = json.loads(json_cmd)
array_url = cmd.get('input', self.base_url + self.array_name)
if not array_url.startswith(self.base_url):
raise RuntimeError('Input array must start with the base url')
array_name = array_url[len(self.base_url):]
fields = cmd['fields']
arr = self.get_session_array(array_name)[...].ddesc.dynd_arr()
# Do the groupby, get its groups, then
# evaluate it because deferred operations
# through the groupby won't work well yet.
res = nd.groupby(arr, nd.fields(arr, *fields))
groups = res.groups
res = res.eval()
# Write out the groupby result
defarr_gb = self.array_provider.create_deferred_array_filename(
self.session_name, 'groupby_', array(res))
dshape_gb = nd.dshape_of(res)
defarr_gb[0].write(json.dumps({
'dshape': dshape_gb,
'command': 'groupby',
'params': {
'fields': fields
}
}))
defarr_gb[0].close()
# Write out the groups
defarr_groups = self.array_provider.create_deferred_array_filename(
self.session_name, 'groups_', groups)
dshape_groups = nd.dshape_of(groups)
defarr_groups[0].write(json.dumps({
'dshape': dshape_groups,
'command': 'groupby.groups',
'params': {
'fields': fields
}
}))
defarr_groups[0].close()
content_type = 'application/json; charset=utf-8'
body = json.dumps({
'session': self.base_url + self.session_name,
'output_gb': self.base_url + defarr_gb[1],
'dshape_gb': dshape_gb,
'output_groups': self.base_url + defarr_groups[1],
'dshape_groups': dshape_groups
})
return (content_type, body)
def test_immutable(self):
a = nd.array([
('x', 0),
('y', 1),
('x', 2),
('x', 3),
('y', 4)],
dtype='{A: string, B: int32}').eval_immutable()
gb = nd.groupby(a, nd.fields(a, 'A'))
self.assertEqual(nd.as_py(gb.groups), [{'A': 'x'}, {'A': 'y'}])
self.assertEqual(nd.as_py(gb), [
[{'A': 'x', 'B': 0},
{'A': 'x', 'B': 2},
{'A': 'x', 'B': 3}],
[{'A': 'y', 'B': 1},
{'A': 'y', 'B': 4}]])
def test_aggregate(self):
a = nd.array([
('A', 1, 2),
('A', 3, 4),
('B', 1.5, 2.5),
('A', 0.5, 9),
('C', 1, 5),
('B', 2, 2)],
dtype='c{cat: string, x: float32, y: float32}')
gb = nd.groupby(a, nd.fields(a, 'cat')).eval()
b = nd.make_computed_fields(gb, 1,
fields=[('sum_x', ndt.float32, 'sum(x)'),
('mean_y', ndt.float32, 'mean(y)'),
('max_x', ndt.float32, 'max(x)'),
('max_y', ndt.float32, 'max(y)')])
self.assertEqual(nd.as_py(b.sum_x), [4.5, 3.5, 1])
self.assertEqual(nd.as_py(b.mean_y), [5, 2.25, 5])
self.assertEqual(nd.as_py(b.max_x), [3, 2, 1])
self.assertEqual(nd.as_py(b.max_y), [9, 2.5, 5])
def test_type_id(self):
# Numeric type id
self.assertEqual(self.type_id_of(ndt.bool),
_lowlevel.type_id.BOOL)
self.assertEqual(self.type_id_of(ndt.int8),
_lowlevel.type_id.INT8)
self.assertEqual(self.type_id_of(ndt.int16),
_lowlevel.type_id.INT16)
self.assertEqual(self.type_id_of(ndt.int32),
_lowlevel.type_id.INT32)
self.assertEqual(self.type_id_of(ndt.int64),
_lowlevel.type_id.INT64)
self.assertEqual(self.type_id_of(ndt.uint8),
_lowlevel.type_id.UINT8)
self.assertEqual(self.type_id_of(ndt.uint16),
_lowlevel.type_id.UINT16)
self.assertEqual(self.type_id_of(ndt.uint32),
_lowlevel.type_id.UINT32)
self.assertEqual(self.type_id_of(ndt.uint64),
_lowlevel.type_id.UINT64)
self.assertEqual(self.type_id_of(ndt.float32),
_lowlevel.type_id.FLOAT32)
self.assertEqual(self.type_id_of(ndt.float64),
_lowlevel.type_id.FLOAT64)
self.assertEqual(self.type_id_of(ndt.complex_float32),
_lowlevel.type_id.COMPLEX_FLOAT32)
self.assertEqual(self.type_id_of(ndt.complex_float64),
_lowlevel.type_id.COMPLEX_FLOAT64)
# String/bytes
self.assertEqual(self.type_id_of(ndt.string),
_lowlevel.type_id.STRING)
self.assertEqual(self.type_id_of(ndt.make_fixedstring(16)),
_lowlevel.type_id.FIXEDSTRING)
self.assertEqual(self.type_id_of(ndt.bytes),
_lowlevel.type_id.BYTES)
self.assertEqual(self.type_id_of(ndt.make_fixedbytes(16)),
_lowlevel.type_id.FIXEDBYTES)
self.assertEqual(self.type_id_of(ndt.json),
_lowlevel.type_id.JSON)
# Date
self.assertEqual(self.type_id_of(ndt.date),
_lowlevel.type_id.DATE)
# Property
self.assertEqual(self.type_id_of(nd.type_of(ndt.date(2000, 1, 1).year)),
_lowlevel.type_id.PROPERTY)
# Categorical
self.assertEqual(self.type_id_of(ndt.make_categorical([1, 2, 3])),
_lowlevel.type_id.CATEGORICAL)
# Struct
self.assertEqual(self.type_id_of(ndt.make_struct(
[ndt.int32, ndt.int32], ['x', 'y'])),
_lowlevel.type_id.STRUCT)
self.assertEqual(self.type_id_of(ndt.type('{x : int32, y : int32}')),
_lowlevel.type_id.FIXEDSTRUCT)
# Convert/byteswap/view
self.assertEqual(self.type_id_of(ndt.make_convert(
ndt.int32, ndt.int8)),
_lowlevel.type_id.CONVERT)
self.assertEqual(self.type_id_of(ndt.make_byteswap(ndt.int32)),
_lowlevel.type_id.BYTESWAP)
self.assertEqual(self.type_id_of(ndt.make_view(
ndt.int32, ndt.uint32)),
_lowlevel.type_id.VIEW)
# CUDA types
if ndt.cuda_support:
self.assertEqual(self.type_id_of(ndt.type('cuda_device[int32]')),
_lowlevel.type_id.CUDA_DEVICE)
self.assertEqual(self.type_id_of(ndt.type('cuda_host[int32]')),
_lowlevel.type_id.CUDA_HOST)
# Uniform arrays
self.assertEqual(self.type_id_of(ndt.type('3 * int32')),
_lowlevel.type_id.FIXED_DIM)
self.assertEqual(self.type_id_of(ndt.type('strided * int32')),
_lowlevel.type_id.STRIDED_DIM)
self.assertEqual(self.type_id_of(ndt.type('var * int32')),
_lowlevel.type_id.VAR_DIM)
# GroupBy
self.assertEqual(self.type_id_of(nd.type_of(nd.groupby([1, 2],
['a', 'a']))),
_lowlevel.type_id.GROUPBY)
# Type
self.assertEqual(self.type_id_of(ndt.type('type')),
_lowlevel.type_id.TYPE)
def groupby(sreader, key, val, dtype, path=None, lines_per_chunk=LPC):
"""Group the `val` field in `sreader` stream of lines by `key` index.
Parameters
----------
sreader : iterator
Iterator over a stream of CSV lines.
key : string
The name of the field to be grouped by.
val : string
The field name with the values that have to be grouped.
dtype : dynd dtype
The DyND data type with all the fields of the CSV lines,
including the `key` and `val` names.
path : string
The path of the file where the BLZ array with the final
grouping will be stored. If None (default), the BLZ will be
stored in-memory (and hence non-persistent).
lines_per_chunk : int
The number of chunks that have to be read to be grouped by
in-memory. For optimal perfomance, some experimentation
should be needed. The default value should work reasonably
well, though.
Returns
-------
output : BLZ table
Returns a BLZ table with column names that are the groups
resulting from the groupby operation. The columns are filled
with the `val` field of the lines delivered by `sreader`.
"""
try:
nptype = get_nptype(dtype, val)
except ValueError:
raise ValueError("`val` should be a valid field")
# Start reading chunks
prev_keys = set()
while True:
ndbuf = nd.array(islice(sreader, lines_per_chunk), dtype)
if len(ndbuf) == 0: break # CSV data exhausted
# Do the groupby for this chunk
keys = getattr(ndbuf, key)
if val is None:
vals = ndbuf
else:
vals = getattr(ndbuf, val)
sby = nd.groupby(vals, keys)
lkeys = nd.as_py(sby.groups)
skeys = set(lkeys)
# BLZ does not understand dynd objects (yet)
sby = nd.as_py(sby.eval())
if len(prev_keys) == 0:
# Add the initial keys to a BLZ table
columns = [np.array(sby[i], nptype) for i in range(len(lkeys))]
ssby = blz.btable(columns=columns, names=lkeys, rootdir=path,
mode='w')
else:
# Have we new keys?
new_keys = skeys.difference(prev_keys)
for new_key in new_keys:
# Get the index of the new key
idx = lkeys.index(new_key)
# and add the values as a new columns
ssby.addcol(sby[idx], new_key, dtype=nptype)
# Now fill the pre-existing keys
existing_keys = skeys.intersection(prev_keys)
for existing_key in existing_keys:
# Get the index of the existing key
idx = lkeys.index(existing_key)
# and append the values here
ssby[existing_key].append(sby[idx])
# Add the new keys to the existing ones
prev_keys |= skeys
# Before returning, flush all data into disk
if path is not None:
ssby.flush()
return ssby
def groupby(sreader, key, val, dtype, path=None, lines_per_chunk=LPC):
"""Group the `val` field in `sreader` stream of lines by `key` index.
Parameters
----------
sreader : iterator
Iterator over a stream of CSV lines.
key : string
The name of the field to be grouped by.
val : string
The field name with the values that have to be grouped.
dtype : dynd dtype
The DyND data type with all the fields of the CSV lines,
including the `key` and `val` names.
path : string
The path of the file where the BLZ array with the final
grouping will be stored. If None (default), the BLZ will be
stored in-memory (and hence non-persistent).
lines_per_chunk : int
The number of chunks that have to be read to be grouped by
in-memory. For optimal perfomance, some experimentation
should be needed. The default value should work reasonably
well, though.
Returns
-------
output : BLZ table
Returns a BLZ table with column names that are the groups
resulting from the groupby operation. The columns are filled
with the `val` field of the lines delivered by `sreader`.
"""
try:
nptype = get_nptype(dtype, val)
except ValueError:
raise ValueError("`val` should be a valid field")
# Start reading chunks
prev_keys = set()
while True:
ndbuf = nd.array(islice(sreader, lines_per_chunk), dtype)
if len(ndbuf) == 0: break # CSV data exhausted
# Do the groupby for this chunk
keys = getattr(ndbuf, key)
if val is None:
vals = ndbuf
else:
vals = getattr(ndbuf, val)
sby = nd.groupby(vals, keys)
lkeys = nd.as_py(sby.groups)
skeys = set(lkeys)
# BLZ does not understand dynd objects (yet)
sby = nd.as_py(sby.eval())
if len(prev_keys) == 0:
# Add the initial keys to a BLZ table
columns = [np.array(sby[i], nptype) for i in range(len(lkeys))]
ssby = blz.btable(columns=columns,
names=lkeys,
rootdir=path,
mode='w')
else:
# Have we new keys?
new_keys = skeys.difference(prev_keys)
for new_key in new_keys:
# Get the index of the new key
idx = lkeys.index(new_key)
# and add the values as a new columns
ssby.addcol(sby[idx], new_key, dtype=nptype)
# Now fill the pre-existing keys
existing_keys = skeys.intersection(prev_keys)
for existing_key in existing_keys:
# Get the index of the existing key
idx = lkeys.index(existing_key)
# and append the values here
ssby[existing_key].append(sby[idx])
# Add the new keys to the existing ones
prev_keys |= skeys
# Before returning, flush all data into disk
if path is not None:
ssby.flush()
return ssby
def groupby(sreader, key, val, dtype, path=None, lines_per_chunk=LPC):
"""Group the `val` field in `sreader` stream of lines by `key` index.
Parameters
----------
sreader : iterator
Iterator over a stream of CSV lines.
key : string
The name of the field to be grouped by.
val : string
The field name with the values that have to be grouped.
dtype : dynd dtype
The DyND data type with all the fields of the CSV lines,
including the `key` and `val` names.
path : string
The path of the file where the BLZ array with the final
grouping will be stored. If None (default), the BLZ will be
stored in-memory (and hence non-persistent).
lines_per_chunk : int
The number of chunks that have to be read to be grouped by
in-memory. For optimal perfomance, some experimentation
should be needed. The default value should work reasonably
well, though.
Returns
-------
output : BLZ table
Returns a BLZ table with column names that are the groups
resulting from the groupby operation. The columns are filled
with the `val` field of the lines delivered by `sreader`.
"""
def get_nptype(dtype, val):
# Convert the `val` field into a numpy dtype
dytype = dtype[nd.as_py(dtype.field_names).index(val)]
# strings and bytes cannot be natively represented in numpy
if dytype == ndt.string:
nptype = np.dtype("U%d" % MAXCHARS)
elif dytype == ndt.bytes:
nptype = np.dtype("S%d" % MAXCHARS)
else:
# There should be no problems with the rest
nptype = dytype.as_numpy()
return nptype
if val is None:
types = [(bytes(name), get_nptype(dtype, name))
for name in nd.as_py(dtype.field_names)]
nptype = np.dtype(types)
else:
nptype = get_nptype(dtype, val)
# Start reading chunks
prev_keys = set()
while True:
ndbuf = nd.array(islice(sreader, lines_per_chunk), dtype)
if len(ndbuf) == 0: break # CSV data exhausted
# Do the groupby for this chunk
keys = getattr(ndbuf, key)
if val is None:
vals = ndbuf
else:
vals = getattr(ndbuf, val)
sby = nd.groupby(vals, keys)
skeys = set(nd.as_py(sby.groups))
lkeys = list(skeys)
# BLZ does not understand dynd objects (yet)
sby = nd.as_py(sby.eval())
if len(prev_keys) == 0:
# Check path and if it exists, remove it and every
# directory below it
if os.path.exists(path): rmtree(path)
# Add the initial keys to a BLZ table
columns = [np.array(sby[i], nptype) for i in range(len(lkeys))]
ssby = blz.btable(columns=columns, names=lkeys, rootdir=path)
else:
# Have we new keys?
new_keys = skeys.difference(prev_keys)
for new_key in new_keys:
# Get the index of the new key
idx = lkeys.index(new_key)
# and add the values as a new columns
ssby.addcol(sby[idx], new_key, dtype=nptype)
# Now fill the pre-existing keys
existing_keys = skeys.intersection(prev_keys)
for existing_key in existing_keys:
# Get the index of the existing key
idx = lkeys.index(existing_key)
# and append the values here
ssby[existing_key].append(sby[idx])
assert skeys == existing_keys | new_keys
# Add the new keys to the existing ones
prev_keys |= skeys
# Before returning, flush all data into disk
ssby.flush()
return ssby
