Assigning a data.table slice in R

Question

In order to read a data.table slice, I can use the following syntax: ``` foo = DT[, 5:10, with=F] ``` but now I want to do: ``` foo = foo + 1 DT[, 5:10, with=F] = foo ``` This doesn't work; referring to the columns by names also doesn't seem to work. Any suggestions?

Accepted Answer

It's a little more subtle. This is how I read your question and how you're trying to do it at the moment ...

Your first line creates a new `data.table` object with the 6-column subset of columns :

```
foo = DT[, 5:10, with=F]
```

I'm immediately thinking of the memory implications. If each column is 1GB, that's a 6GB new object you just allocated.

then you +1 to everything in that 6GB :

```
foo = foo + 1   # or something like that, that works
```

That's a copy of that 6GB to another new 6GB.

Then you copy the 6GB `foo` back into where it was in `DT` in the first place :

```
DT[, 5:10, with=F] = foo    # or something like that, that works
```

That's really memory inefficient. It's a base R way of doing things.

In `data.table` you can loop, and you can `set`. I would just do it in an easy to read and easy to understand loop.

```
for (col in 5:10)
    set(DT, j=col, value=DT[[col]]+1)
```

This changes each column by reference one-by-one. `DT[[col]]` doesn't copy the column contents (that's nothing special in `data.table`, that's base R that doesn't copy). But the `+1` does create a new vector. However, that new vector is then plonked directly into the column pointer slot, so it's as efficient as it can be given that `+1` returns a new object.

Assigning a data.table slice in R

Solution

Problem