new f-method .as()

[st-pasha]

I think our current API lacks somewhat when it comes to renaming columns in the j section. The "normal" API is to use a dict: {"newA": f.A, "newB": f.B, ...}. However, a common use-case is when you want to select some columns as-is, and also some new columns with names. In this case repeating names of all columns may be too difficult or even impossible.

A possible workaround is to use the .extend() method. However, this has its own drawbacks:

• it is quite non-obvious;
• the resulting expression is still very complicated;
• if you start with an expression such as DT[:, [f.A, f.B, f.C]] and then decide that you need to rename, say, column B, then you end up having to completely rewrite the formula into DT[:, f.A.extend({"newB": f.B}).extend(f.C)].

The solution that I'm proposing is to have a new method .as() that could rename an individual f-expression. For example:

DT[:, [f.A, f.B.as("newB"), f.C]]

(the name is inspired by the equivalent SQL keyword).

Similarly, you can apply this to a columnset and rename multiple columns at once:

DT[:, f[:2].as("Price", "Quantity")]

[pradkrish]

That is a good idea. Also, modifiers - by(), sort() and join() - within the square-brackets should be able to operate on the alias.

[samukweku]

this has been resolved with the .alias method