issues

Code Sample

Hello,

I performed a quick comparison of the newly introduced method ìnterp()` with an adapter (draft) to the sdf (scientific data format) library: https://gist.github.com/gwin-zegal/b955c3ef63f5ad51eec6329dd2e620be#file-array_sdf_interp-py

Code for a micro comparison (2D array) in python (include the above gist first):

```python

arr = xr.DataArray(np.sort(np.sort(np.random.RandomState(123).rand(30,4), axis=0), axis=1), coords=[('tension', np.arange(10, 40)), ('resistance', np.linspace(100, 500, 4))])

res = {'xarray': [], 'c_sdf' : []} x = np.logspace(1, 4, num=10, dtype=np.int16) for size in x: new_tension = arr.tension[0].data + np.random.random_sample(size=size) * (arr.tension[-1].data - arr.tension[0].data) new_resistance = arr.resistance[0].data + np.random.random_sample(size=size) * (arr.resistance[-1].data - arr.resistance[0].data)

interp_xr = %timeit -qo arr.interp({'tension': new_tension, 'resistance': new_resistance})
res['xarray'].append(interp_xr)

interp_c_sdf = %timeit -qo arr(new_tension, new_resistance)
res['c_sdf'].append(interp_c_sdf)

```

Problem description

The time spent for array.interp()is growing exponentially... over two 2min (xarray internal interp) on my old machine compared to 9ms (C-SDF wrapper) for 10_000 interpolations.

The C-SDF code is slow (a copy of the array is performed and algorithms not so optimized), but xarray implementation is not usable in daily life on my machine!

{'xarray': [<TimeitResult : 6.46 ms ± 223 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 6.46 ms ± 88.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 6.99 ms ± 141 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 8.91 ms ± 52 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 19.8 ms ± 183 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 112 ms ± 638 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)>, <TimeitResult : 584 ms ± 19.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)>, <TimeitResult : 2.63 s ± 43.4 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)>, <TimeitResult : 15.5 s ± 147 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)>, <TimeitResult : 2min 23s ± 18.7 s per loop (mean ± std. dev. of 7 runs, 1 loop each)>], 'c_sdf': [<TimeitResult : 1.08 ms ± 7.96 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.09 ms ± 7.91 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.1 ms ± 12.9 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.13 ms ± 9.95 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.19 ms ± 7.76 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.32 ms ± 9.47 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 1.59 ms ± 10.5 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)>, <TimeitResult : 2.19 ms ± 31.8 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 3.51 ms ± 34.9 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>, <TimeitResult : 9.27 ms ± 307 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)>]}

Performance issue on my machine or is it confirmed by others?

Output of xr.show_versions()