(discussion) a good/convenient way to get a view of subset of a dataset

[zhangyingmath]

(please help me move it to the "discussion" session because I can't post there directly.)

As far as I understand, if I use a range index, start:stop:step, I got a view (no copy), and modifying the subset will modify its parent; if I use a fancy index I will get a copy. Here are my wishes:

1. Is there a way to get a view using some kind of fancy index? (I can use ds2=ds.filter(fancy_index, inplace=True), but this drops the rows of the parent that is not selected.)

(I suppose the answer to this question is impossible, but just want to ask.)

2. One reason I like a fancy index is because it can be assigned to a python variable, e.g. a list of int or a list of bools. On the other hand, I don't know how to do it to a range index. Thus I often have to type ds[start_index:(stop_index + 1), :], sometimes multiple times, leading to code duplication. I wonder if I can define a variable, a=numpy_range_index(start, stop, step), and just use this variable ds[a, :] to select a view. It should be nothing more than a syntactic sugar.

Moreover, it would be good if statements like "np.where" can accept this range_index object as well.