Sterling Paramore

So I understand an index as a unique identifier to the row. Isn't that how vectors in a data frame line up in to rows? But what can you accomplish...

I'm not sure I understand how the concept "guarantees" it will be faster, particularly if we start overloading it with functionality we don't strictly require to manipulate data.

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

I've never used pandas, so maybe I just don't get the utility of having categorical indices. But I think I've been able to handle selecting, joining, grouping, and aggregating data...

Intriguing. Would this only work with JRuby, or would it be possible to work with MRI too?

I'm still struggling with understanding why indexes more complex than sequential integers are really necessary for dataframes. Ideally, `#where` on a single vector should be as performance as any index...

We've had several unexpected bugs arise (see the last group_by issue last week) due to `#[]` serving two purposes. I strongly prefer using `#at` for *position* and `#[]` for *index...

I think that `[]` should be predictable. It currently isn't because if an index doesn't exist then it falls back to position and there is no predictable relationship between index...

We have a #concat method for vectors, but it also requires an index. You can fake it with something like this: ``` a = Daru::Vector.new(['a','b','c']).to_df b = Daru::Vector.new(['d','e']).to_df (a.concat b)[nil]...

Reread your question and I may have jumped too quickly. I assumed by stacking you meant concatenation. But it sounds like you're actually talking about transposing? But yes, an example...