daru: Should we support duplicated keys in an index?

Daru::Vector.new([1,2,3], index: [1,1,1])
=> IndexError: Expected index size >= vector size. Index size : 1, vector size : 3

Daru::Index.new([1,1,1])
=> #<Daru::Index(1): {1}>

Index#initialize uses @relation_hash = index.each_with_index.to_h.freeze, this causes

[1,2,1].each_with_index.to_h
=> {1=>2, 2=>1}

About this issue

Original URL
State: open
Created 8 years ago
Comments: 16 (12 by maintainers)

Most upvoted comments

Since CategoricalIndex is an “index”, it can keep looking up items by category values faster than using #where.

mrkn on Sep 25, 2016

@gnilrets Nothing.

I think in fact we should optimize #where irrelevant of whether we have CategoricalIndex or not.

So the question rises whether we should drop CategoricalIndex because other things can accomplish the same? Inspite of my surrender to that all things which are accomplishable via CategoricalIndex can still be accomplished though other things, I’m still of the opinion that CategoricalIndex is required to fills the gap for need to have non-unique indexing.

lokeshh on Sep 20, 2016

What prevents us from refactoring #where to be as fast?

gnilrets on Sep 17, 2016