与 numpy 函数的兼容性

与 numpy 函数的兼容性¶

下表描述了 numpy 和 dask.array 函数之间的兼容性。请注意，这两个版本之间可能存在一些不一致。

此表是手动编译的，可能无法反映当前的 Dask 状态。欢迎更新贡献。

空白条目表示该函数在 Dask 中未实现。
直接实现是对 numpy 函数的直接调用。
逐元素实现源自 numpy，但应用于逐元素：参数应为 dask 数组。
Dask 等价物是 Dask 实现，它们可能在参数上相对于 numpy 函数有所缺失或增加。

在 Array API 框架下有更深入的比较，可通过 Array API Comparison repository 获取。

NumPy	Dask	实现
`numpy.absolute`	`dask.array.absolute` or `dask.array.abs`	直接 (ufunc)
`numpy.add`	`dask.array.add`	直接 (ufunc)
`numpy.all`	`dask.array.all` 21	dask 等效
`numpy.allclose`	`dask.array.allclose`	dask 等效
`numpy.amax`	`dask.array.max` 21 22	dask 等效
`numpy.amin`	`dask.array.min` 21 22	dask 等效
`numpy.angle`	`dask.array.angle` 23	dask 等效
`numpy.any`	`dask.array.any` 21	dask 等效
`numpy.append`	`dask.array.append`	dask 等效
`numpy.apply_along_axis`	`dask.array.apply_along_axis`	dask 等效
`numpy.apply_over_axes`	`dask.array.apply_over_axes`	dask 等效
`numpy.arange`	`dask.array.arange`	dask 等效
`numpy.arccos`	`dask.array.arccos`	直接 (ufunc)
`numpy.arccosh`	`dask.array.arccosh`	直接 (ufunc)
`numpy.arcsin`	`dask.array.arcsin`	直接 (ufunc)
`numpy.arcsinh`	`dask.array.arcsinh`	直接 (ufunc)
`numpy.arctan`	`dask.array.arctan`	直接 (ufunc)
`numpy.arctan2`	`dask.array.arctan2`	直接 (ufunc)
`numpy.arctanh`	`dask.array.arctanh`	直接 (ufunc)
`numpy.argmax`	`dask.array.argmax`	dask 等效
`numpy.argmin`	`dask.array.argmin`	dask 等效
`numpy.argpartition`	-
`numpy.argsort`	`dask.array.argtopk` 25
`numpy.argwhere`	`dask.array.argwhere`	dask 等效
`numpy.around`	`dask.array.around` 23 26 or `dask.array.round`	dask 等效
`numpy.array`	`dask.array.array`	dask 等效
`numpy.array2string`	-
`numpy.array_equal`	-
`numpy.array_equiv`	-
`numpy.array_repr`	-
`numpy.array_split`	-
`numpy.array_str`	-
`numpy.asanyarray`	`dask.array.asanyarray`	dask 等效
`numpy.asarray`	`dask.array.asarray`	dask 等效
`numpy.asarray_chkfinite`	-
`numpy.ascontiguousarray`	-
`numpy.asfarray`	-
`numpy.asfortranarray`	-
`numpy.asmatrix`	- 27
`numpy.atleast_1d`	`dask.array.atleast_1d`	dask 等效
`numpy.atleast_2d`	`dask.array.atleast_2d`	dask 等效
`numpy.atleast_3d`	`dask.array.atleast_3d`	dask 等效
`numpy.average`	`dask.array.average`	dask 等效
`numpy.bartlett`	-
`numpy.bincount`	`dask.array.bincount`	dask 等效
`numpy.bitwise_and`	`dask.array.bitwise_and`	直接 (ufunc)
`numpy.bitwise_or`	`dask.array.bitwise_or`	直接 (ufunc)
`numpy.bitwise_xor`	`dask.array.bitwise_xor`	直接 (ufunc)
`numpy.blackman`	-
`numpy.block`	`dask.array.block`	dask 等效
`numpy.bmat`	- 27
`numpy.broadcast`	-
`numpy.broadcast_arrays`	`dask.array.broadcast_arrays`	dask 等效
`numpy.broadcast_shapes`	-
`numpy.broadcast_to`	`dask.array.broadcast_to`	dask 等效
`numpy.byte_bounds`	-
`numpy.c_`	-
`numpy.can_cast`	-
`numpy.cbrt`	`dask.array.cbrt`	直接 (ufunc)
`numpy.ceil`	`dask.array.ceil`	直接 (ufunc)
`numpy.choose`	`dask.array.choose` 28	dask 等效
`numpy.clip`	`dask.array.clip` 23 26	直接（非ufunc元素级）
`numpy.column_stack`	-
`numpy.common_type`	-
`numpy.compress`	`dask.array.compress` 26	dask 等效
`numpy.concatenate`	`dask.array.concatenate`	dask 等效
`numpy.conj`	`dask.array.conj`	直接 (ufunc)
`numpy.conjugate`	`dask.array.conj`	直接 (ufunc)
`numpy.convolve`	-
`numpy.copy`	-
`numpy.copysign`	`dask.array.copysign`	直接 (ufunc)
`numpy.copyto`	-
`numpy.corrcoef`	`dask.array.corrcoef`	dask 等效
`numpy.correlate`	-
`numpy.cos`	`dask.array.cos`	直接 (ufunc)
`numpy.cosh`	`dask.array.cosh`	直接 (ufunc)
`numpy.count_nonzero`	`dask.array.count_nonzero` 29	dask 等效
`numpy.cov`	`dask.array.cov` 30	dask 等效
`numpy.cross`	-
`numpy.cumprod`	`dask.array.cumprod` 23 38	dask 等效
`numpy.cumsum`	`dask.array.cumsum` 23 38	dask 等效
`numpy.datetime_as_string`	-
`numpy.deg2rad`	`dask.array.deg2rad`	直接 (ufunc)
`numpy.degrees`	`dask.array.degrees`	直接 (ufunc)
`numpy.delete`	`dask.array.delete`	dask 等效
`numpy.diag`	`dask.array.diag`	dask 等效
`numpy.diag_indices`	-
`numpy.diag_indices_from`	-
`numpy.diagflat`	-
`numpy.diagonal`	`dask.array.diagonal`	dask 等效
`numpy.diff`	`dask.array.diff`	dask 等效
`numpy.digitize`	`dask.array.digitize` 23	dask 等效
`numpy.divide`	`dask.array.divide`	直接 (ufunc)
`numpy.divmod`	`dask.array.divmod`	dask 等效
`numpy.dot`	`dask.array.dot` 26	dask 等效
`numpy.dsplit`	-
`numpy.dstack`	`dask.array.dstack`	dask 等效
`numpy.ediff1d`	`dask.array.ediff1d`	dask 等效
`numpy.einsum`	`dask.array.einsum` 26	dask 等效
`numpy.einsum_path`	-
`numpy.empty`	`dask.array.empty`	dask 等效
`numpy.empty_like`	`dask.array.empty_like`	dask 等效
`numpy.equal`	`dask.array.equal`	直接 (ufunc)
`numpy.exp`	`dask.array.exp`	直接 (ufunc)
`numpy.exp2`	`dask.array.exp2`	直接 (ufunc)
`numpy.expand_dims`	`dask.array.expand_dims`	dask 等效
`numpy.expm1`	`dask.array.expm1`	直接 (ufunc)
`numpy.extract`	`dask.array.extract`	dask 等效
`numpy.eye`	`dask.array.eye`	dask 等效
`numpy.fabs`	`dask.array.fabs`	直接 (ufunc)
`numpy.fill_diagonal`	-
`numpy.fix`	`dask.array.fix`	直接（非ufunc元素级）
`numpy.flatnonzero`	`dask.array.flatnonzero`	dask 等效
`numpy.flip`	`dask.array.flip`	dask 等效
`numpy.fliplr`	`dask.array.fliplr`	dask 等效
`numpy.flipud`	`dask.array.flipud`	dask 等效
`numpy.float_power`	`dask.array.float_power`	直接 (ufunc)
`numpy.floor`	`dask.array.floor`	直接 (ufunc)
`numpy.floor_divide`	`dask.array.floor_divide`	直接 (ufunc)
`numpy.fmax`	`dask.array.fmax`	直接 (ufunc)
`numpy.fmin`	`dask.array.fmin`	直接 (ufunc)
`numpy.fmod`	`dask.array.fmod`	直接 (ufunc)
`numpy.frexp`	`dask.array.frexp`	dask 等效
`numpy.from_dlpack`	-
`numpy.frombuffer`	-
`numpy.fromfile`	-
`numpy.fromfunction`	`dask.array.fromfunction` 31	dask 等效
`numpy.fromiter`	-
`numpy.frompyfunc`	`dask.array.frompyfunc` 32	dask 等效
`numpy.fromregex`	-
`numpy.fromstring`	-
`numpy.full`	`dask.array.full`	dask 等效
`numpy.full_like`	`dask.array.full_like`	dask 等效
`numpy.gcd`	-
`numpy.genfromtxt`	-
`numpy.geomspace`	-
`numpy.gradient`	`dask.array.gradient` 33	dask 等效
`numpy.greater`	`dask.array.greater`	直接 (ufunc)
`numpy.greater_equal`	`dask.array.greater_equal`	直接 (ufunc)
`numpy.hamming`	-
`numpy.hanning`	-
`numpy.heaviside`	-
`numpy.histogram`	`dask.array.histogram`	dask 等效
`numpy.histogram2d`	`dask.array.histogram2d`	dask 等效
`numpy.histogram_bin_edges`	-
`numpy.histogramdd`	`dask.array.histogramdd` 34	dask 等效
`numpy.hsplit`	-
`numpy.hstack`	`dask.array.hstack`	dask 等效
`numpy.hypot`	`dask.array.hypot`	直接 (ufunc)
`numpy.i0`	`dask.array.i0`	直接（非ufunc元素级）
`numpy.identity`	-
`numpy.imag`	`dask.array.imag`	直接（非ufunc元素级）
`numpy.in1d`	-
`numpy.indices`	`dask.array.indices`	dask 等效
`numpy.inner`	-
`numpy.insert`	`dask.array.insert` 35	dask 等效
`numpy.interp`	-
`numpy.intersect1d`	-
`numpy.invert`	`dask.array.invert` or `dask.array.bitwise_not`	直接 (ufunc)
`numpy.is_busday`	-
`numpy.isclose`	`dask.array.isclose`	dask 等效
`numpy.iscomplex`	`dask.array.iscomplex`	直接（非ufunc元素级）
`numpy.iscomplexobj`	-
`numpy.isfinite`	`dask.array.isfinite`	直接 (ufunc)
`numpy.isfortran`	-
`numpy.isin`	`dask.array.isin`	dask 等效
`numpy.isinf`	`dask.array.isinf`	直接 (ufunc)
`numpy.isnan`	`dask.array.isnan`	直接 (ufunc)
`numpy.isnat`	-
`numpy.isneginf`	`dask.array.isneginf`	直接 (ufunc)
`numpy.isposinf`	`dask.array.isposinf`	直接 (ufunc)
`numpy.isreal`	`dask.array.isreal`	直接（非ufunc元素级）
`numpy.ix_`	-
`numpy.kaiser`	-
`numpy.kron`	-
`numpy.lcm`	-
`numpy.ldexp`	`dask.array.ldexp`	直接 (ufunc)
`numpy.left_shift`	`dask.array.left_shift`	直接 (ufunc)
`numpy.less`	`dask.array.less`	直接 (ufunc)
`numpy.less_equal`	`dask.array.less_equal`	直接 (ufunc)
`numpy.lexsort`	-
`numpy.linspace`	`dask.array.linspace`	dask 等效
`numpy.load`	-
`numpy.loadtxt`	-
`numpy.log`	`dask.array.log`	直接 (ufunc)
`numpy.log10`	`dask.array.log10`	直接 (ufunc)
`numpy.log1p`	`dask.array.log1p`	直接 (ufunc)
`numpy.log2`	`dask.array.log2`	直接 (ufunc)
`numpy.logaddexp`	`dask.array.logaddexp`	直接 (ufunc)
`numpy.logaddexp2`	`dask.array.logaddexp2`	直接 (ufunc)
`numpy.logical_and`	`dask.array.logical_and`	直接 (ufunc)
`numpy.logical_not`	`dask.array.logical_not`	直接 (ufunc)
`numpy.logical_or`	`dask.array.logical_or`	直接 (ufunc)
`numpy.logical_xor`	`dask.array.logical_xor`	直接 (ufunc)
`numpy.logspace`	-
`numpy.mask_indices`	-
`numpy.mat`	- 27
`numpy.matmul`	`dask.array.matmul`	dask 等效
`numpy.matrix`	- 27
`numpy.maximum`	`dask.array.maximum`	直接 (ufunc)
`numpy.may_share_memory`	-
`numpy.mean`	`dask.array.mean` 21	dask 等效
`numpy.median`	`dask.array.median` 36	dask 等效
`numpy.memmap`	-
`numpy.meshgrid`	`dask.array.meshgrid` 37	dask 等效
`numpy.mgrid`	-
`numpy.minimum`	`dask.array.minimum`	直接 (ufunc)
`numpy.mintypecode`	-
`numpy.mod`	`dask.array.mod`	直接 (ufunc)
`numpy.modf`	`dask.array.modf`	dask 等效
`numpy.moveaxis`	`dask.array.moveaxis`	dask 等效
`numpy.multiply`	`dask.array.multiply`	直接 (ufunc)
`numpy.nan_to_num`	`dask.array.nan_to_num`	直接（非ufunc元素级）
`numpy.nanargmax`	`dask.array.nanargmax`	dask 等效
`numpy.nanargmin`	`dask.array.nanargmin`	dask 等效
`numpy.nancumprod`	`dask.array.nancumprod` 23 38	dask 等效
`numpy.nancumsum`	`dask.array.nancumsum` 23 38	dask 等效
`numpy.nanmax`	`dask.array.nanmax` 21 22	dask 等效
`numpy.nanmean`	`dask.array.nanmean` 21	dask 等效
`numpy.nanmedian`	`dask.array.nanmedian` 36	dask 等效
`numpy.nanmin`	`dask.array.nanmin` 21 22	dask 等效
`numpy.nanpercentile`	-
`numpy.nanprod`	`dask.array.nanprod` 21 22	dask 等效
`numpy.nanquantile`	-
`numpy.nanstd`	`dask.array.nanstd` 21	dask 等效
`numpy.nansum`	`dask.array.nansum` 21 22	dask 等效
`numpy.nanvar`	`dask.array.nanvar` 21	dask 等效
`numpy.ndenumerate`	-
`numpy.ndindex`	-
`numpy.nditer`	-
`numpy.negative`	`dask.array.negative`	直接 (ufunc)
`numpy.nested_iters`	-
`numpy.nextafter`	`dask.array.nextafter`	直接 (ufunc)
`numpy.nonzero`	`dask.array.nonzero`	dask 等效
`numpy.not_equal`	`dask.array.not_equal`	直接 (ufunc)
`numpy.ogrid`	-
`numpy.ones`	`dask.array.ones`	dask 等效
`numpy.ones_like`	`dask.array.ones_like`	dask 等效
`numpy.outer`	`dask.array.outer`	dask 等效
`numpy.packbits`	-
`numpy.pad`	`dask.array.pad`	dask 等效
`numpy.partition`	-
`numpy.percentile`	`dask.array.percentile`	dask 等效
`numpy.piecewise`	`dask.array.piecewise`	dask 等效
`numpy.place`	-
`numpy.poly`	-
`numpy.poly1d`	-
`numpy.polyadd`	-
`numpy.polyder`	-
`numpy.polydiv`	-
`numpy.polyfit`	-
`numpy.polyint`	-
`numpy.polymul`	-
`numpy.polysub`	-
`numpy.polyval`	-
`numpy.positive`	`dask.array.positive`	直接 (ufunc)
`numpy.power`	`dask.array.power`	直接 (ufunc)
`numpy.prod`	`dask.array.prod`	dask 等效
`numpy.ptp`	`dask.array.ptp`	dask 等效
`numpy.put`	-
`numpy.put_along_axis`	-
`numpy.putmask`	-
`numpy.quantile`	-
`numpy.r_`	-
`numpy.rad2deg`	`dask.array.rad2deg`	直接 (ufunc)
`numpy.radians`	`dask.array.radians`	直接 (ufunc)
`numpy.ravel`	`dask.array.ravel` 23 24	dask 等效
`numpy.ravel_multi_index`	`dask.array.ravel_multi_index`	dask 等效
`numpy.real`	`dask.array.real`	直接（非ufunc元素级）
`numpy.real_if_close`	-
`numpy.reciprocal`	`dask.array.reciprocal`	直接 (ufunc)
`numpy.remainder`	`dask.array.remainder`	直接 (ufunc)
`numpy.repeat`	`dask.array.repeat`	dask 等效
`numpy.require`	-
`numpy.reshape`	`dask.array.reshape`	dask 等效
`numpy.resize`	-
`numpy.result_type`	`dask.array.result_type`	dask 等效
`numpy.right_shift`	`dask.array.right_shift`	直接 (ufunc)
`numpy.rint`	`dask.array.rint`	直接 (ufunc)
`numpy.roll`	`dask.array.roll`	dask 等效
`numpy.rollaxis`	`dask.array.rollaxis`	dask 等效
`numpy.roots`	-
`numpy.rot90`	`dask.array.rot90`	dask 等效
`numpy.row_stack`	-
`numpy.save`	-
`numpy.savetxt`	-
`numpy.savez`	-
`numpy.savez_compressed`	-
`numpy.searchsorted`	`dask.array.searchsorted`	dask 等效
`numpy.select`	`dask.array.select`	dask 等效
`numpy.setdiff1d`	-
`numpy.setxor1d`	-
`numpy.shape`	`dask.array.shape` 23	dask 等效
`numpy.shares_memory`	-
`numpy.sign`	`dask.array.sign`	直接 (ufunc)
`numpy.signbit`	`dask.array.signbit`	直接 (ufunc)
`numpy.sin`	`dask.array.sin`	直接 (ufunc)
`numpy.sinc`	`dask.array.sinc`	直接（非ufunc元素级）
`numpy.sinh`	`dask.array.sinh`	直接 (ufunc)
`numpy.sort`	`dask.array.topk` 25
`numpy.sort_complex`	-
`numpy.source`	-
`numpy.spacing`	`dask.array.spacing`	直接 (ufunc)
`numpy.split`	-
`numpy.sqrt`	`dask.array.sqrt`	直接 (ufunc)
`numpy.square`	`dask.array.square`	直接 (ufunc)
`numpy.squeeze`	`dask.array.squeeze`	dask 等效
`numpy.stack`	`dask.array.stack`	dask 等效
`numpy.std`	`dask.array.std` 21	dask 等效
`numpy.subtract`	`dask.array.subtract`	直接 (ufunc)
`numpy.sum`	`dask.array.sum` 21 22	dask 等效
`numpy.swapaxes`	`dask.array.swapaxes`	dask 等效
`numpy.take`	`dask.array.take` 28	dask 等效
`numpy.take_along_axis`	-
`numpy.tan`	`dask.array.tan`	直接 (ufunc)
`numpy.tanh`	`dask.array.tanh`	直接 (ufunc)
`numpy.tensordot`	`dask.array.tensordot`	dask 等效
`numpy.tile`	`dask.array.tile`	dask 等效
`numpy.trace`	`dask.array.trace` 26	dask 等效
`numpy.transpose`	`dask.array.transpose`	dask 等效
`numpy.trapz`	-
`numpy.tri`	`dask.array.tri`	dask 等效
`numpy.tril`	`dask.array.tril`	dask 等效
`numpy.tril_indices`	`dask.array.tril_indices`	dask 等效
`numpy.tril_indices_from`	`dask.array.tril_indices_from`	dask 等效
`numpy.trim_zeros`	-
`numpy.triu`	`dask.array.triu`	dask 等效
`numpy.triu_indices`	`dask.array.triu_indices`	dask 等效
`numpy.triu_indices_from`	`dask.array.triu_indices_from`	dask 等效
`numpy.true_divide`	`dask.array.true_divide`	直接 (ufunc)
`numpy.trunc`	`dask.array.trunc`	直接 (ufunc)
`numpy.union1d`	`dask.array.union1d`	dask 等效
`numpy.unique`	`dask.array.unique` 39	dask 等效
`numpy.unpackbits`	-
`numpy.unravel_index`	`dask.array.unravel_index`	dask 等效
`numpy.unwrap`	-
`numpy.vander`	-
`numpy.var`	`dask.array.var` 21	dask 等效
`numpy.vdot`	`dask.array.vdot`	dask 等效
`numpy.vsplit`	-
`numpy.vstack`	`dask.array.vstack` 40	dask 等效
`numpy.where`	`dask.array.where`	dask 等效
`numpy.zeros`	`dask.array.zeros`	dask 等效
`numpy.zeros_like`	`dask.array.zeros_like`	dask 等效

脚注

21(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15): where 参数不受支持。
22(1,2,3,4,5,6,7): initial 参数不受支持。
23(1,2,3,4,5,6,7,8,9,10): 输入必须是一个 dask 数组。
24: 不支持 order 参数。
25(1,2): 排序操作以难以并行化而闻名。对并行友好的替代方案是对k个最大元素进行排序。
26(1,2,3,4,5,6): 不支持 out 参数。
27(1,2,3,4): 不鼓励在 NumPy 中使用 numpy.matrix，因此没有必要添加它。
28(1,2): mode 参数不受支持。
29: keepdims 参数不受支持。
30: fweights、aweights、dtype 参数不受支持。
31: like 参数不受支持。可调用函数不受支持。
32: 不支持多于一个输出。
33: edge_order 参数不受支持。
34: 输入数据（样本）的分块仅允许沿第0轴（行）进行。
35: 仅对单调的 obj 参数实现。
36(1,2): overwrite_input 参数不支持。
37: copy 参数不受支持。
38(1,2,3,4): Dask 实现引入了一个额外的参数 method。
39: axis 参数不受支持。
40: casting 参数不受支持。

dask.array.utils.meta_from_array

包