问题描述
我已阅读此帖子并怀疑我需要更好地了解索引.我想做类似np.ndenumerate的操作，但是(在这种情况下)仅在前三个维度上执行，并返回一系列向量:
I have read this post and suspect I need to understand indexing better. I would like to do something like np.ndenumerate, but (in this particular case) only over the first three dimensions, returning a series of vectors:
x = np.linspace(0., 1., 4)
y = np.linspace(0., 1., 5)
z = np.linspace(0., 1., 2)
X, Y, Z = np.meshgrid(x, y, z) # define a grid
F = np.zeros(X.shape + (3,)) # a 3D vector field
F = np.random.rand(5*4*2*3).reshape(5,4,2,3) # added this later just to give non-zero for testing
thetas = np.linspace(0, 2.*np.pi, 21)[:-1] # drop the duplicate
v = np.array([(np.cos(theta), np.sin(theta), 0.0) for theta in thetas])
for tup, vec in magical_enumerate(F, axis=(0,1,2)): # it's magic! (need correct expression)
F(tup) = np.cross(v, vec).sum(axis=0) # sum over 20 vectors in v
有没有一种方法可以在没有大量循环或列表解释的情况下进行操作?网格将很大，因此请欣赏numpythony和速度.非顺序尺寸(例如，轴=(0,2))怎么办?谢谢！
Is there a way to do this without lots of loops or list interpretations? The grids will be large, so numpythony and speed are appreciated. What about non-sequential dimensions (e.g. axis=(0,2))? Thanks!
推荐答案
np.ndindex可能会解决问题.它会在一组维度上生成迭代器.
np.ndindex might do the trick. It generates an iterator over a set of dimensions.
In [231]: F=np.zeros((2,3,3,3))
In [232]: for tup in np.ndindex(F.shape[:3]):
# vec = data[tup] etc
F[tup]=tup
.....:
In [233]: F
Out[233]:
array([[[[ 0., 0., 0.],
[ 0., 0., 1.],
[ 0., 0., 2.]],
...
[ 1., 2., 1.],
[ 1., 2., 2.]]]])
我建议同时查看ndenumerate和ndindex的代码. ndindex在multi_index模式下使用更新的nditer. ndenumerate使用flat迭代所有值.
I'd suggest looking at the code for both ndenumerate and ndindex. ndindex is using the more recent nditer in a multi_index mode. ndenumerate uses flat to iterate over all values.
我在其他SO中建议了如何构建以ndindex为模型的自己的multi_index迭代器.在nditer上进行搜索可能会产生这些结果.
I've suggested in other SO how you could construct your own multi_index iterator modeled on ndindex. A search on nditer might produce those.
这不会在多个循环上带来速度优势，因为您仍在处理相同数量的内部大多数计算.
This is not going to give a speed advantage over multiple loops, because you still are dealing with the same number of inner most calculations.
对于非顺序维，同样的ndindex也可以使用，但是必须先操作tup才能将其用作索引:
As for non-sequential dimensions, this same ndindex will work, but you have to manipulate tup before using it as an index:
In [243]: for tup in np.ndindex((F.shape[0],F.shape[2])):
tup1=(tup[0],slice(None),tup[1])
F[tup]=[tup[0],np.nan,tup[1]]
.....:
np.apply_along_axis和np.apply_over_axes是在分割其他轴时在一个或多个轴上生成索引的其他示例.
np.apply_along_axis and np.apply_over_axes are other examples of generating indices over 1 or more of the axes, while slicing others.
这篇关于像部分/选择性np.ndenumerate之类的东西?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！