Compatibility with numpy functions

Compatibility with numpy functions¶

The following table describes the compatibilities between numpy and dask.array functions. Please be aware that some inconsistencies with the two versions may exist.

This table has been compiled manually and may not reflect the current Dask state. Update contributions are welcome.

A blank entry indicates that the function is not implemented in Dask.
Direct implementation are direct calls to numpy functions.
Element-wise implementations are derived from numpy but applied element-wise: the argument should be a dask array.
Dask equivalent are Dask implementations, which may lack or add parameters with respect to the numpy function.

A more in-depth comparison in the framework of the Array API is available via the Array API Comparison repository.

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

Footnotes

21(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15): where parameter not supported.
22(1,2,3,4,5,6,7): initial parameter not supported.
23(1,2,3,4,5,6,7,8,9,10): Input must be a dask array.
24: order parameter not supported.
25(1,2): Sort operations are notoriously difficult to do in parallel. Parallel-friendly alternatives sort the k largest elements.
26(1,2,3,4,5,6): out parameter not supported.
27(1,2,3,4): Use of numpy.matrix is discouraged in NumPy and thus there is no need to add it.
28(1,2): mode parameter not supported.
29: keepdims parameter not supported.
30: fweights, aweights, dtype parameters not supported.
31: like parameter not supported. Callable functions not supported.
32: Not implemented with more than one output.
33: edge_order parameter not supported.
34: Chunking of the input data (sample) is only allowed along the 0th (row) axis.
35: Only implemented for monotonic obj arguments.
36(1,2): overwrite_input parameter not supported.
37: copy parameter not supported.
38(1,2,3,4): Dask implementation introduces an additional parameter method.
39: axis parameter not supported.
40: casting parameter not supported.

dask.array.utils.meta_from_array

Bag