class Series(_Frame, dd.core.Series):
_partition_type = cudf.Series
def count(self, split_every=False):
return reduction(
self, chunk=M.count, aggregate=np.sum, split_every=split_every, meta="i8"
)
def mean(self, split_every=False):
sum = self.sum(split_every=split_every)
n = self.count(split_every=split_every)
return sum / n
def unique_k(self, k, split_every=None):
return reduction(
self,
chunk=M.unique_k,
aggregate=unique_k_agg,
meta=self._meta,
token="unique-k",
split_every=split_every,
k=k,
)
# ----------------------------------------------------------------------
# Accessor Methods
# ----------------------------------------------------------------------
dt = CachedAccessor("dt", DatetimeAccessor)
cat = CachedAccessor("cat", CategoricalAccessor)
class Series(_Frame, dd.core.Series):
_partition_type = cudf.Series
def count(self, split_every=False):
return reduction(
self, chunk=M.count, aggregate=np.sum, split_every=split_every, meta="i8"
)
def mean(self, split_every=False):
sum = self.sum(split_every=split_every)
n = self.count(split_every=split_every)
return sum / n
def unique_k(self, k, split_every=None):
return reduction(
self,
chunk=M.unique_k,
aggregate=unique_k_agg,
meta=self._meta,
token="unique-k",
split_every=split_every,
k=k,
)
@derived_from(pd.DataFrame)
def var(self, axis=None, skipna=True, ddof=1, split_every=False, dtype=None, out=None):
axis = self._validate_axis(axis)
meta = self._meta_nonempty.var(axis=axis, skipna=skipna)
if axis == 1:
result = map_partitions(M.var, self, meta=meta,
token=self._token_prefix + 'var',
axis=axis, skipna=skipna, ddof=ddof)
return handle_out(out, result)
else:
num = self._get_numeric_data()
x = 1.0 * num.sum(skipna=skipna, split_every=split_every)
x2 = 1.0 * (num ** 2).sum(skipna=skipna, split_every=split_every)
n = num.count(split_every=split_every)
name = self._token_prefix + 'var'
result = map_partitions(var_aggregate, x2, x, n,
token=name, meta=meta, ddof=ddof)
if isinstance(self, DataFrame):
result.divisions = (min(self.columns), max(self.columns))
return handle_out(out, result)
# ----------------------------------------------------------------------
# Accessor Methods
# ----------------------------------------------------------------------
dt = CachedAccessor("dt", DatetimeAccessor)
cat = CachedAccessor("cat", CategoricalAccessor)
class Series(_Frame):
_partition_type = gd.Series
@property
def dtype(self):
return self._meta.dtype
def astype(self, dtype):
if dtype == self.dtype:
return self
return self.map_partitions(M.astype, dtype=dtype)
def sum(self, split_every=False):
return reduction(self, chunk=M.sum, aggregate=np.sum,
split_every=split_every, meta=self.dtype)
def count(self, split_every=False):
return reduction(self, chunk=M.count, aggregate=np.sum,
split_every=split_every, meta='i8')
def mean(self, split_every=False):
sum = self.sum(split_every=split_every)
n = self.count(split_every=split_every)
return sum / n
def var(self, ddof=1, split_every=False):
sum2 = reduction(self, chunk=sum_of_squares, aggregate=np.sum,
split_every=split_every, meta='f8')
sum = self.sum(split_every=split_every)
n = self.count(split_every=split_every)
return map_partitions(var_aggregate, sum2, sum, n, ddof=ddof,
meta='f8')
def std(self, ddof=1, split_every=False):
var = self.var(ddof=ddof, split_every=split_every)
return map_partitions(np.sqrt, var, dtype=np.float64)
def min(self, split_every=False):
return reduction(self, chunk=M.min, aggregate=np.min,
split_every=split_every, meta=self.dtype)
def max(self, split_every=False):
return reduction(self, chunk=M.max, aggregate=np.max,
split_every=split_every, meta=self.dtype)
def ceil(self):
return self.map_partitions(M.ceil)
def floor(self):
return self.map_partitions(M.floor)
def fillna(self, value):
if not np.can_cast(value, self.dtype):
raise TypeError("fill value must match dtype of series")
return self.map_partitions(M.fillna, value, meta=self)
def nlargest(self, n=5, split_every=None):
return reduction(self, chunk=M.nlargest, aggregate=nlargest_agg,
meta=self._meta, token='series-nlargest',
split_every=split_every, n=n)
def nsmallest(self, n=5, split_every=None):
return reduction(self, chunk=M.nsmallest, aggregate=nsmallest_agg,
meta=self._meta, token='series-nsmallest',
split_every=split_every, n=n)
def unique_k(self, k, split_every=None):
return reduction(self, chunk=M.unique_k, aggregate=unique_k_agg,
meta=self._meta, token='unique-k',
split_every=split_every, k=k)
# ----------------------------------------------------------------------
# Accessor Methods
# ----------------------------------------------------------------------
dt = CachedAccessor("dt", DatetimeAccessor)
cat = CachedAccessor("cat", CategoricalAccessor)