Coverage increased (+0.03%) to 95.283% when pulling 7be79c0 on cartesian_nearest_index into fcde8c9 on master.

Thanks @oyamad !! This is clever.

In the following example I map (0.1, 0.1, 0.1) to the nearest grid point in R^3, which should be (0.0, 0.0, 0.0). I get a different result. Am I misunderstanding how to use the function?

[ins] In [14]: nodes = (np.linspace(0, 1, 5), np.linspace(0, 2, 3), np.linspace(0, 1, 4
          ...: ))

[ins] In [15]: nodes
Out[15]: 
(array([0.  , 0.25, 0.5 , 0.75, 1.  ]),
 array([0., 1., 2.]),
 array([0.        , 0.33333333, 0.66666667, 1.        ]))

[ins] In [16]: prod = qe.cartesian(nodes)

[ins] In [17]: qe.cartesian_nearest_index((0.1, 0.1, 0.1), prod)
Out[17]: 22

[ins] In [18]: prod[22]
Out[18]: array([0.25      , 2.        , 0.66666667])

@jstac Thanks for your review!

You are supposed to pass nodes, not prod.

In [4]: qe.cartesian_nearest_index((0.1, 0.1, 0.1), nodes)
Out[4]: 0

In your call qe.cartesian_nearest_index((0.1, 0.1, 0.1), prod), it searches over the 60-dimensional Cartesian product that would be generated by the 60 arrays in prod, but only the first 3 dimensions (where len(x) = 3) are looked up. I will add lines to raise an error when the dimensions do not match as in this case.

Many thanks @oyamad , this looks good to me, and is greatly appreciated.

@oyamad or @mmcky please merge when ready.