cbind/cbind! and copies

[tshort]

Julia-dev discussion

cbind(df1, df2) currently makes copies of both df1 and df2 (expensive)
cbind!(df1, df2) currently makes no copies and modifies df1

We need something that doesn't copy columns but doesn't modify df1 or df2. I'd rather have:

cbind(df1, df2) be equivalent to DataFrame([df1.columns, df2.columns], newcolnames)

That doesn't change df1 or df2, but they share columns. If we have that, I don't see too much need for cbind!.

Creating a DataFrame from existing columns isn't expensive.

[HarlanH]

I completely concur. Let's go with your proposed design. We should specify, however, that any meta-data (if we had any) associated with DataFrames, would get copied from df1.

[tshort]

I just pushed a change to fix this plus added hcat and vcat. I also deleted the version of cbind that allowed name/column pairs. I think we should add that back in once Julia has keyword function arguments. As it is now, the easiest way to get that is assignment.

It looks like I should have referenced this issue in my push. Anyway, I'll wait to close this until someone else looks at it.

[HarlanH]

Looks good! (Yeah, referring to issue numbers is always a good idea...)