You can run this notebook in a live session or view it on Github.

# Applying unvectorized functions with apply_ufunc¶

This example will illustrate how to conveniently apply an unvectorized function func to xarray objects using apply_ufunc. func expects 1D numpy arrays and returns a 1D numpy array. Our goal is to coveniently apply this function along a dimension of xarray objects that may or may not wrap dask arrays with a signature.

We will illustrate this using np.interp:

Signature: np.interp(x, xp, fp, left=None, right=None, period=None)
Docstring:
One-dimensional linear interpolation.

Returns the one-dimensional piecewise linear interpolant to a function
with given discrete data points (xp, fp), evaluated at x.


and write an xr_interp function with signature

xr_interp(xarray_object, dimension_name, new_coordinate_to_interpolate_to)


First lets load an example dataset

[1]:

import xarray as xr
import numpy as np

xr.set_options(display_style="html")  # fancy HTML repr

air = (
.air.sortby("lat")  # np.interp needs coordinate in ascending order
.isel(time=slice(4), lon=slice(3))
)  # choose a small subset for convenience
air

[1]:

<xarray.DataArray 'air' (time: 4, lat: 25, lon: 3)>
array([[[296.29   , 296.79   , 297.1    ],
[295.9    , 296.19998, 296.79   ],
[296.6    , 296.19998, 296.4    ],
[297.     , 296.69998, 296.1    ],
[295.4    , 295.69998, 295.79   ],
[293.79   , 294.1    , 294.6    ],
[293.1    , 293.29   , 293.29   ],
[290.19998, 290.79   , 291.4    ],
[287.9    , 288.     , 288.29   ],
[286.5    , 286.5    , 285.69998],
[284.6    , 284.9    , 284.19998],
[282.79   , 283.19998, 282.6    ],
[280.     , 280.69998, 280.19998],
[278.4    , 279.     , 279.     ],
[277.29   , 277.4    , 277.79   ],
[276.69998, 277.4    , 277.69998],
[275.9    , 276.9    , 276.9    ],
[274.79   , 275.19998, 275.6    ],
[273.69998, 273.6    , 273.79   ],
[272.1    , 270.9    , 270.     ],
...
[293.     , 293.5    , 294.29   ],
[291.9    , 291.9    , 292.19998],
[289.19998, 289.4    , 289.9    ],
[286.6    , 287.1    , 287.9    ],
[284.79   , 284.79   , 285.4    ],
[282.79   , 282.     , 282.69998],
[281.19998, 280.19998, 280.6    ],
[279.5    , 278.69998, 278.6    ],
[278.     , 277.69998, 277.6    ],
[276.4    , 275.9    , 276.4    ],
[275.6    , 275.69998, 276.1    ],
[274.5    , 275.6    , 276.29   ],
[273.4    , 274.5    , 275.5    ],
[274.1    , 274.     , 273.5    ],
[273.29   , 272.6    , 271.5    ],
[272.79   , 272.4    , 271.9    ],
[267.69998, 266.29   , 264.4    ],
[256.6    , 254.7    , 252.09999],
[246.29999, 245.29999, 244.2    ],
[241.89   , 241.79999, 241.79999]]], dtype=float32)
Coordinates:
* lat      (lat) float32 15.0 17.5 20.0 22.5 25.0 ... 65.0 67.5 70.0 72.5 75.0
* lon      (lon) float32 200.0 202.5 205.0
* time     (time) datetime64[ns] 2013-01-01 ... 2013-01-01T18:00:00
Attributes:
long_name:     4xDaily Air temperature at sigma level 995
units:         degK
precision:     2
GRIB_id:       11
GRIB_name:     TMP
var_desc:      Air temperature
dataset:       NMC Reanalysis
level_desc:    Surface
statistic:     Individual Obs
parent_stat:   Other
actual_range:  [185.16 322.1 ]

The function we will apply is np.interp which expects 1D numpy arrays. This functionality is already implemented in xarray so we use that capability to make sure we are not making mistakes.

[2]:

newlat = np.linspace(15, 75, 100)
air.interp(lat=newlat)

[2]:

<xarray.DataArray 'air' (time: 4, lat: 100, lon: 3)>
array([[[296.29000854, 296.79000854, 297.1000061 ],
[296.19545954, 296.64697173, 297.02485518],
[296.10091053, 296.50393491, 296.94970426],
...,
[242.46059851, 243.46969695, 244.08181672],
[241.83029767, 242.98484846, 243.79090837],
[241.19999683, 242.49999997, 243.50000003]],

[[296.29000854, 297.19998169, 297.3999939 ],
[296.26818385, 297.07876957, 297.25211866],
[296.24635916, 296.95755744, 297.10424342],
...,
[242.82726354, 243.37878187, 243.63332714],
[242.46362716, 243.03938941, 243.36665899],
[242.09999079, 242.69999695, 243.09999084]],

[[296.3999939 , 296.29000854, 296.3999939 ],
[296.35150609, 296.34091556, 296.37333078],
[296.30301828, 296.39182258, 296.34666767],
...,
[243.4151408 , 243.26181628, 243.12423612],
[242.85756431, 242.73090659, 242.71211194],
[242.29998782, 242.19999689, 242.29998776]],

[[297.5       , 297.69998169, 297.5       ],
[297.37878788, 297.65150128, 297.40303179],
[297.25757575, 297.60302087, 297.30606357],
...,
[244.02817552, 243.49695752, 242.96362858],
[242.9590874 , 242.64847269, 242.38180817],
[241.88999927, 241.79998785, 241.79998776]]])
Coordinates:
* lon      (lon) float32 200.0 202.5 205.0
* time     (time) datetime64[ns] 2013-01-01 ... 2013-01-01T18:00:00
* lat      (lat) float64 15.0 15.61 16.21 16.82 ... 73.18 73.79 74.39 75.0
Attributes:
long_name:     4xDaily Air temperature at sigma level 995
units:         degK
precision:     2
GRIB_id:       11
GRIB_name:     TMP
var_desc:      Air temperature
dataset:       NMC Reanalysis
level_desc:    Surface
statistic:     Individual Obs
parent_stat:   Other
actual_range:  [185.16 322.1 ]

Let’s define a function that works with one vector of data along lat at a time.

[3]:

def interp1d_np(data, x, xi):
return np.interp(xi, x, data)

interped = interp1d_np(air.isel(time=0, lon=0), air.lat, newlat)
expected = air.interp(lat=newlat)

# no errors are raised if values are equal to within floating point precision
np.testing.assert_allclose(expected.isel(time=0, lon=0).values, interped)


## No errors are raised so our interpolation is working.¶

This function consumes and returns numpy arrays, which means we need to do a lot of work to convert the result back to an xarray object with meaningful metadata. This is where apply_ufunc is very useful.

## apply_ufunc¶

Apply a vectorized function for unlabeled arrays on xarray objects.

The function will be mapped over the data variable(s) of the input arguments using
xarray’s standard rules for labeled computation, including alignment, broadcasting,
looping over GroupBy/Dataset variables, and merging of coordinates.


apply_ufunc has many capabilities but for simplicity this example will focus on the common task of vectorizing 1D functions over nD xarray objects. We will iteratively build up the right set of arguments to apply_ufunc and read through many error messages in doing so.

[4]:

xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_485/2982343242.py in <module>
----> 1 xr.apply_ufunc(
2     interp1d_np,  # first the function
3     air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
4     air.lat,
5     newlat,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_ufunc(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)
1163     # feed DataArray apply_variable_ufunc through apply_dataarray_vfunc
1164     elif any(isinstance(a, DataArray) for a in args):
-> 1165         return apply_dataarray_vfunc(
1166             variables_vfunc,
1167             *args,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_dataarray_vfunc(func, signature, join, exclude_dims, keep_attrs, *args)
288
289     data_vars = [getattr(a, "variable", a) for a in args]
--> 290     result_var = func(*data_vars)
291
292     if signature.num_outputs > 1:

765         for dim, new_size in var.sizes.items():
766             if dim in dim_sizes and new_size != dim_sizes[dim]:
--> 767                 raise ValueError(
768                     "size of dimension {!r} on inputs was unexpectedly "
769                     "changed by applied function from {} to {}. Only "

ValueError: size of dimension 'lat' on inputs was unexpectedly changed by applied function from 25 to 100. Only dimensions specified in exclude_dims with xarray.apply_ufunc are allowed to change size.


apply_ufunc needs to know a lot of information about what our function does so that it can reconstruct the outputs. In this case, the size of dimension lat has changed and we need to explicitly specify that this will happen. xarray helpfully tells us that we need to specify the kwarg exclude_dims.

## exclude_dims¶

exclude_dims : set, optional
Core dimensions on the inputs to exclude from alignment and
broadcasting entirely. Any input coordinates along these dimensions
will be dropped. Each excluded dimension must also appear in
input_core_dims for at least one argument. Only dimensions listed
here are allowed to change size between input and output objects.

[5]:

xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_485/2413120057.py in <module>
----> 1 xr.apply_ufunc(
2     interp1d_np,  # first the function
3     air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
4     air.lat,
5     newlat,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_ufunc(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)
1079             )
1080         if not exclude_dims <= signature.all_core_dims:
-> 1081             raise ValueError(
1082                 f"each dimension in exclude_dims must also be a "
1083                 f"core dimension in the function signature. "

ValueError: each dimension in exclude_dims must also be a core dimension in the function signature. Please make {'lat'} a core dimension


## Core dimensions¶

Core dimensions are central to using apply_ufunc. In our case, our function expects to receive a 1D vector along lat — this is the dimension that is “core” to the function’s functionality. Multiple core dimensions are possible. apply_ufunc needs to know which dimensions of each variable are core dimensions.

input_core_dims : Sequence[Sequence], optional
List of the same length as args giving the list of core dimensions
on each input argument that should not be broadcast. By default, we
assume there are no core dimensions on any input arguments.

For example, input_core_dims=[[], ['time']] indicates that all
dimensions on the first argument and all dimensions other than 'time'
on the second argument should be broadcast.

Core dimensions are automatically moved to the last axes of input
variables before applying func, which facilitates using NumPy style
generalized ufuncs [2]_.

output_core_dims : List[tuple], optional
List of the same length as the number of output arguments from
func, giving the list of core dimensions on each output that were
not broadcast on the inputs. By default, we assume that func
outputs exactly one array, with axes corresponding to each broadcast
dimension.

Core dimensions are assumed to appear as the last dimensions of each
output in the provided order.


Next we specify "lat" as input_core_dims on both air and air.lat

[6]:

xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
input_core_dims=[["lat"], ["lat"], []],
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_485/2853147781.py in <module>
----> 1 xr.apply_ufunc(
2     interp1d_np,  # first the function
3     air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
4     air.lat,
5     newlat,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_ufunc(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)
1163     # feed DataArray apply_variable_ufunc through apply_dataarray_vfunc
1164     elif any(isinstance(a, DataArray) for a in args):
-> 1165         return apply_dataarray_vfunc(
1166             variables_vfunc,
1167             *args,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_dataarray_vfunc(func, signature, join, exclude_dims, keep_attrs, *args)
288
289     data_vars = [getattr(a, "variable", a) for a in args]
--> 290     result_var = func(*data_vars)
291
292     if signature.num_outputs > 1:

756         data = as_compatible_data(data)
757         if data.ndim != len(dims):
--> 758             raise ValueError(
759                 "applied function returned data with unexpected "
760                 f"number of dimensions. Received {data.ndim} dimension(s) but "

ValueError: applied function returned data with unexpected number of dimensions. Received 1 dimension(s) but expected 0 dimensions with names: ()


xarray is telling us that it expected to receive back a numpy array with 0 dimensions but instead received an array with 1 dimension corresponding to newlat. We can fix this by specifying output_core_dims

[7]:

xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
input_core_dims=[["lat"], ["lat"], []],  # list with one entry per arg
output_core_dims=[["lat"]],
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
)

[7]:

<xarray.DataArray (lat: 100)>
array([296.29000854, 296.19545954, 296.10091053, 296.00636153,
295.91181252, 296.04848133, 296.21818126, 296.38788119,
296.55758112, 296.67273227, 296.76970048, 296.8666687 ,
296.96363692, 296.75757483, 296.36969457, 295.9818143 ,
295.59393403, 295.20484416, 294.81454468, 294.4242452 ,
294.03394572, 293.72728105, 293.56000773, 293.39273441,
293.22546109, 292.92424705, 292.22121083, 291.5181746 ,
290.81513838, 290.13028509, 289.57271229, 289.01513949,
288.45756669, 287.8999939 , 287.56060144, 287.22120898,
286.88181652, 286.54242406, 286.09697099, 285.63636641,
285.17576183, 284.71515725, 284.27091564, 283.83212835,
283.39334106, 282.95455378, 282.36727998, 281.69091427,
281.01454856, 280.33818285, 279.80605987, 279.4181796 ,
279.03029933, 278.64241906, 278.29908614, 278.02999878,
277.76091142, 277.49182406, 277.25424934, 277.11121253,
276.96817571, 276.8251389 , 276.67573964, 276.4818032 ,
276.28786677, 276.09393033, 275.8999939 , 275.63090654,
275.36181918, 275.09273182, 274.82364446, 274.55879073,
274.29454179, 274.03029286, 273.76604392, 273.40907704,
273.02120417, 272.6333313 , 272.24545843, 272.46364154,
273.04545824, 273.62727495, 274.20909165, 273.53030303,
271.59090909, 269.65151515, 267.71212121, 265.        ,
261.        , 257.        , 253.        , 249.62424168,
248.12120842, 246.61817516, 245.1151419 , 243.72120019,
243.09089938, 242.46059857, 241.83029776, 241.19999695])
Coordinates:
lon      float32 200.0
time     datetime64[ns] 2013-01-01
Dimensions without coordinates: lat

Finally we get some output! Let’s check that this is right

[8]:

interped = xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(time=0, lon=0),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
input_core_dims=[["lat"], ["lat"], []],  # list with one entry per arg
output_core_dims=[["lat"]],
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
)
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(expected.isel(time=0, lon=0), interped)


No errors are raised so it is right!

## Vectorization with np.vectorize¶

Now our function currently only works on one vector of data which is not so useful given our 3D dataset. Let’s try passing the whole dataset. We add a print statement so we can see what our function receives.

[9]:

def interp1d_np(data, x, xi):
print(f"data: {data.shape} | x: {x.shape} | xi: {xi.shape}")
return np.interp(xi, x, data)

interped = xr.apply_ufunc(
interp1d_np,  # first the function
air.isel(
lon=slice(3), time=slice(4)
),  # now arguments in the order expected by 'interp1_np'
air.lat,
newlat,
input_core_dims=[["lat"], ["lat"], []],  # list with one entry per arg
output_core_dims=[["lat"]],
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
)
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(expected.isel(time=0, lon=0), interped)

data: (4, 3, 25) | x: (25,) | xi: (100,)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_485/4068599170.py in <module>
4
5
----> 6 interped = xr.apply_ufunc(
7     interp1d_np,  # first the function
8     air.isel(

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_ufunc(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)
1163     # feed DataArray apply_variable_ufunc through apply_dataarray_vfunc
1164     elif any(isinstance(a, DataArray) for a in args):
-> 1165         return apply_dataarray_vfunc(
1166             variables_vfunc,
1167             *args,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_dataarray_vfunc(func, signature, join, exclude_dims, keep_attrs, *args)
288
289     data_vars = [getattr(a, "variable", a) for a in args]
--> 290     result_var = func(*data_vars)
291
292     if signature.num_outputs > 1:

731             )
732
--> 733     result_data = func(*input_data)
734
735     if signature.num_outputs == 1:

/tmp/ipykernel_485/4068599170.py in interp1d_np(data, x, xi)
1 def interp1d_np(data, x, xi):
2     print(f"data: {data.shape} | x: {x.shape} | xi: {xi.shape}")
----> 3     return np.interp(xi, x, data)
4
5

<__array_function__ internals> in interp(*args, **kwargs)

~/checkouts/readthedocs.org/user_builds/xray/conda/stable/lib/python3.8/site-packages/numpy/lib/function_base.py in interp(x, xp, fp, left, right, period)
1426         fp = np.concatenate((fp[-1:], fp, fp[0:1]))
1427
-> 1428     return interp_func(x, xp, fp, left, right)
1429
1430

ValueError: object too deep for desired array


That’s a hard-to-interpret error but our print call helpfully printed the shapes of the input data:

data: (10, 53, 25) | x: (25,) | xi: (100,)


We see that air has been passed as a 3D numpy array which is not what np.interp expects. Instead we want loop over all combinations of lon and time; and apply our function to each corresponding vector of data along lat. apply_ufunc makes this easy by specifying vectorize=True:

vectorize : bool, optional
If True, then assume func only takes arrays defined over core
dimensions as input and vectorize it automatically with
:py:func:numpy.vectorize. This option exists for convenience, but is
almost always slower than supplying a pre-vectorized function.
Using this option requires NumPy version 1.12 or newer.


Also see the documentation for np.vectorize: https://docs.scipy.org/doc/numpy/reference/generated/numpy.vectorize.html. Most importantly

The vectorize function is provided primarily for convenience, not for performance.
The implementation is essentially a for loop.

[10]:

def interp1d_np(data, x, xi):
print(f"data: {data.shape} | x: {x.shape} | xi: {xi.shape}")
return np.interp(xi, x, data)

interped = xr.apply_ufunc(
interp1d_np,  # first the function
air,  # now arguments in the order expected by 'interp1_np'
air.lat,  # as above
newlat,  # as above
input_core_dims=[["lat"], ["lat"], []],  # list with one entry per arg
output_core_dims=[["lat"]],  # returned data has one dimension
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be set!
vectorize=True,  # loop over non-core dims
)
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(expected, interped)

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_485/1853810661.py in <module>
4
5
----> 6 interped = xr.apply_ufunc(
7     interp1d_np,  # first the function
8     air,  # now arguments in the order expected by 'interp1_np'

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_ufunc(func, input_core_dims, output_core_dims, exclude_dims, vectorize, join, dataset_join, dataset_fill_value, keep_attrs, kwargs, dask, output_dtypes, output_sizes, meta, dask_gufunc_kwargs, *args)
1163     # feed DataArray apply_variable_ufunc through apply_dataarray_vfunc
1164     elif any(isinstance(a, DataArray) for a in args):
-> 1165         return apply_dataarray_vfunc(
1166             variables_vfunc,
1167             *args,

~/checkouts/readthedocs.org/user_builds/xray/checkouts/stable/xarray/core/computation.py in apply_dataarray_vfunc(func, signature, join, exclude_dims, keep_attrs, *args)
288
289     data_vars = [getattr(a, "variable", a) for a in args]
--> 290     result_var = func(*data_vars)
291
292     if signature.num_outputs > 1:

731             )
732
--> 733     result_data = func(*input_data)
734
735     if signature.num_outputs == 1:

2111             vargs.extend([kwargs[_n] for _n in names])
2112
-> 2113         return self._vectorize_call(func=func, args=vargs)
2114
2115     def _get_ufunc_and_otypes(self, func, args):

2185         """Vectorized call to func over positional args."""
2186         if self.signature is not None:
-> 2187             res = self._vectorize_call_with_signature(func, args)
2188         elif not args:
2189             res = func()

2214         args = tuple(asanyarray(arg) for arg in args)
2215
-> 2216         broadcast_shape, dim_sizes = _parse_input_dimensions(
2217             args, input_core_dims)

1881         dummy_array = np.lib.stride_tricks.as_strided(0, arg.shape[:ndim])
1885

418     # use the old-iterator because np.nditer does not handle size 0 arrays
419     # consistently
421     # unfortunately, it cannot handle 32 or more arguments directly
422     for pos in range(32, len(args), 31):

ValueError: shape mismatch: objects cannot be broadcast to a single shape


This unfortunately is another cryptic error from numpy.

Notice that newlat is not an xarray object. Let’s add a dimension name new_lat and modify the call. Note this cannot be lat because xarray expects dimensions to be the same size (or broadcastable) among all inputs. output_core_dims needs to be modified appropriately. We’ll manually rename new_lat back to lat for easy checking.

[11]:

def interp1d_np(data, x, xi):
print(f"data: {data.shape} | x: {x.shape} | xi: {xi.shape}")
return np.interp(xi, x, data)

interped = xr.apply_ufunc(
interp1d_np,  # first the function
air,  # now arguments in the order expected by 'interp1_np'
air.lat,  # as above
newlat,  # as above
input_core_dims=[["lat"], ["lat"], ["new_lat"]],  # list with one entry per arg
output_core_dims=[["new_lat"]],  # returned data has one dimension
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be a set!
vectorize=True,  # loop over non-core dims
)
interped = interped.rename({"new_lat": "lat"})
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(
expected.transpose(*interped.dims), interped  # order of dims is different
)
interped

data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)

[11]:

<xarray.DataArray (time: 4, lon: 3, lat: 100)>
array([[[296.29000854, 296.19545954, 296.10091053, ..., 242.46059857,
241.83029776, 241.19999695],
[296.79000854, 296.64697173, 296.50393492, ..., 243.46969697,
242.98484848, 242.5       ],
[297.1000061 , 297.02485518, 296.94970426, ..., 244.0818167 ,
243.79090835, 243.5       ]],

[[296.29000854, 296.26818385, 296.24635916, ..., 242.82726357,
242.46362721, 242.09999084],
[297.19998169, 297.07876957, 296.95755745, ..., 243.37878187,
243.03938941, 242.69999695],
[297.3999939 , 297.25211866, 297.10424342, ..., 243.63332714,
243.36665899, 243.09999084]],

[[296.3999939 , 296.35150609, 296.30301828, ..., 243.41514079,
242.85756429, 242.29998779],
[296.29000854, 296.34091556, 296.39182258, ..., 243.26181631,
242.73090663, 242.19999695],
[296.3999939 , 296.37333078, 296.34666767, ..., 243.12423614,
242.71211196, 242.29998779]],

[[297.5       , 297.37878788, 297.25757576, ..., 244.02817559,
242.95908749, 241.88999939],
[297.69998169, 297.65150128, 297.60302087, ..., 243.49695749,
242.64847264, 241.79998779],
[297.5       , 297.40303178, 297.30606357, ..., 242.9636286 ,
242.38180819, 241.79998779]]])
Coordinates:
* lon      (lon) float32 200.0 202.5 205.0
* time     (time) datetime64[ns] 2013-01-01 ... 2013-01-01T18:00:00
* lat      (lat) float64 15.0 15.61 16.21 16.82 ... 73.18 73.79 74.39 75.0

Notice that the printed input shapes are all 1D and correspond to one vector along the lat dimension.

The result is now an xarray object with coordinate values copied over from data. This is why apply_ufunc is so convenient; it takes care of a lot of boilerplate necessary to apply functions that consume and produce numpy arrays to xarray objects.

One final point: lat is now the last dimension in interped. This is a “property” of core dimensions: they are moved to the end before being sent to interp1d_np as was noted in the docstring for input_core_dims

Core dimensions are automatically moved to the last axes of input
variables before applying func, which facilitates using NumPy style
generalized ufuncs [2]_.


So far our function can only handle numpy arrays. A real benefit of apply_ufunc is the ability to easily parallelize over dask chunks when needed.

We want to apply this function in a vectorized fashion over each chunk of the dask array. This is possible using dask’s blockwise, map_blocks, or apply_gufunc. Xarray’s apply_ufunc wraps dask’s apply_gufunc and asking it to map the function over chunks using apply_gufunc is as simple as specifying dask="parallelized". With this level of flexibility we need to provide dask with some extra information: 1. output_dtypes: dtypes of all returned objects, and 2. output_sizes: lengths of any new dimensions.

Here we need to specify output_dtypes since apply_ufunc can infer the size of the new dimension new_lat from the argument corresponding to the third element in input_core_dims. Here I choose the chunk sizes to illustrate that np.vectorize is still applied so that our function receives 1D vectors even though the blocks are 3D.

[12]:

def interp1d_np(data, x, xi):
print(f"data: {data.shape} | x: {x.shape} | xi: {xi.shape}")
return np.interp(xi, x, data)

interped = xr.apply_ufunc(
interp1d_np,  # first the function
air.chunk(
{"time": 2, "lon": 2}
),  # now arguments in the order expected by 'interp1_np'
air.lat,  # as above
newlat,  # as above
input_core_dims=[["lat"], ["lat"], ["new_lat"]],  # list with one entry per arg
output_core_dims=[["new_lat"]],  # returned data has one dimension
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be a set!
vectorize=True,  # loop over non-core dims
output_dtypes=[air.dtype],  # one per output
).rename({"new_lat": "lat"})
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(expected.transpose(*interped.dims), interped)

data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)
data: (25,) | x: (25,) | xi: (100,)


Yay! our function is receiving 1D vectors, so we’ve successfully parallelized applying a 1D function over a block. If you have a distributed dashboard up, you should see computes happening as equality is checked.

## High performance vectorization: gufuncs, numba & guvectorize¶

np.vectorize is a very convenient function but is unfortunately slow. It is only marginally faster than writing a for loop in Python and looping. A common way to get around this is to write a base interpolation function that can handle nD arrays in a compiled language like Fortran and then pass that to apply_ufunc.

Another option is to use the numba package which provides a very convenient guvectorize decorator: https://numba.pydata.org/numba-doc/latest/user/vectorize.html#the-guvectorize-decorator

Any decorated function gets compiled and will loop over any non-core dimension in parallel when necessary. We need to specify some extra information:

1. Our function cannot return a variable any more. Instead it must receive a variable (the last argument) whose contents the function will modify. So we change from def interp1d_np(data, x, xi) to def interp1d_np_gufunc(data, x, xi, out). Our computed results must be assigned to out. All values of out must be assigned explicitly.

2. guvectorize needs to know the dtypes of the input and output. This is specified in string form as the first argument. Each element of the tuple corresponds to each argument of the function. In this case, we specify float64 for all inputs and outputs: "(float64[:], float64[:], float64[:], float64[:])" corresponding to data, x, xi, out

3. Now we need to tell numba the size of the dimensions the function takes as inputs and returns as output i.e. core dimensions. This is done in symbolic form i.e. data and x are vectors of the same length, say n; xi and the output out have a different length, say m. So the second argument is (again as a string) "(n), (n), (m) -> (m)." corresponding again to data, x, xi, out

[13]:

from numba import float64, guvectorize

@guvectorize("(float64[:], float64[:], float64[:], float64[:])", "(n), (n), (m) -> (m)")
def interp1d_np_gufunc(data, x, xi, out):
# numba doesn't really like this.
# seem to support fstrings so do it the old way
print(
"data: " + str(data.shape) + " | x:" + str(x.shape) + " | xi: " + str(xi.shape)
)
out[:] = np.interp(xi, x, data)
# gufuncs don't return data
# instead you assign to a the last arg
# return np.interp(xi, x, data)

/tmp/ipykernel_485/959665420.py:4: NumbaWarning:
Compilation is falling back to object mode WITHOUT looplifting enabled because Function "interp1d_np_gufunc" failed type inference due to: No implementation of function Function(<class 'str'>) found for signature:

>>> str(UniTuple(int64 x 1))

There are 6 candidate implementations:
- Of which 6 did not match due to:
Overload of function 'str': File: <numerous>: Line N/A.
With argument(s): '(UniTuple(int64 x 1))':
No match.

During: resolving callee type: Function(<class 'str'>)
During: typing of call at /tmp/ipykernel_485/959665420.py (9)

File "../../../../../../../../../../tmp/ipykernel_485/959665420.py", line 9:
<source missing, REPL/exec in use?>

@guvectorize("(float64[:], float64[:], float64[:], float64[:])", "(n), (n), (m) -> (m)")
/home/docs/checkouts/readthedocs.org/user_builds/xray/conda/stable/lib/python3.8/site-packages/numba/core/object_mode_passes.py:151: NumbaWarning: Function "interp1d_np_gufunc" was compiled in object mode without forceobj=True.

File "../../../../../../../../../../tmp/ipykernel_485/959665420.py", line 4:
<source missing, REPL/exec in use?>

warnings.warn(errors.NumbaWarning(warn_msg,
Fall-back from the nopython compilation path to the object mode compilation path has been detected, this is deprecated behaviour.

File "../../../../../../../../../../tmp/ipykernel_485/959665420.py", line 4:
<source missing, REPL/exec in use?>



The warnings are about object-mode compilation relating to the print statement. This means we don’t get much speed up: https://numba.pydata.org/numba-doc/latest/user/performance-tips.html#no-python-mode-vs-object-mode. We’ll keep the print statement temporarily to make sure that guvectorize acts like we want it to.

[14]:

interped = xr.apply_ufunc(
interp1d_np_gufunc,  # first the function
air.chunk(
{"time": 2, "lon": 2}
),  # now arguments in the order expected by 'interp1_np'
air.lat,  # as above
newlat,  # as above
input_core_dims=[["lat"], ["lat"], ["new_lat"]],  # list with one entry per arg
output_core_dims=[["new_lat"]],  # returned data has one dimension
exclude_dims=set(("lat",)),  # dimensions allowed to change size. Must be a set!
# vectorize=True,  # not needed since numba takes care of vectorizing
output_dtypes=[air.dtype],  # one per output
).rename({"new_lat": "lat"})
interped["lat"] = newlat  # need to add this manually
xr.testing.assert_allclose(expected.transpose(*interped.dims), interped)

data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)
data: (25,) | x:(25,) | xi: (100,)


Yay! Our function is receiving 1D vectors and is working automatically with dask arrays. Finally let’s comment out the print line and wrap everything up in a nice reusable function

[15]:

from numba import float64, guvectorize

@guvectorize(
"(float64[:], float64[:], float64[:], float64[:])",
"(n), (n), (m) -> (m)",
nopython=True,
)
def interp1d_np_gufunc(data, x, xi, out):
out[:] = np.interp(xi, x, data)

def xr_interp(data, dim, newdim):

interped = xr.apply_ufunc(
interp1d_np_gufunc,  # first the function
data,  # now arguments in the order expected by 'interp1_np'
data[dim],  # as above
newdim,  # as above
input_core_dims=[[dim], [dim], ["__newdim__"]],  # list with one entry per arg
output_core_dims=[["__newdim__"]],  # returned data has one dimension
exclude_dims=set((dim,)),  # dimensions allowed to change size. Must be a set!
# vectorize=True,  # not needed since numba takes care of vectorizing
output_dtypes=[
data.dtype
],  # one per output; could also be float or np.dtype("float64")
).rename({"__newdim__": dim})
interped[dim] = newdim  # need to add this manually

return interped

xr.testing.assert_allclose(
expected.transpose(*interped.dims),
xr_interp(air.chunk({"time": 2, "lon": 2}), "lat", newlat),
)


This technique is generalizable to any 1D function.