alter class of data.table object

[jangorecki]

I'm storing particular structure of data as data.table object. I would like to easily plot the object.

dt <- data.table(a = 1:10, b = rnorm(10), z = letters[1:10]) # my custom data stored as DT
plot.my_class <- function(x, ...) plot(x$a, x$b, type="l", main="my custom plot of custom object stored in DT")

Is it safe to alter class of data.table in the following way?

class(dt) <- c("my_class",class(dt))
plot(dt) # plot works fine

I would like to confirm is it safe the add another (my custom) class to DT object? will the rest of data.table functions works fine on my dt or maybe it is not recommended?

[rsaporta]

Try to avoid using class(DT) <- ....

Instead use setattr(DT, "class", c("New_class", class(DT))

[arunsrinivasan]

The idiomatic way would be to use setattr:

setattr(dt, c("myClass", class(dt)))

Why?

In R v3.1+, class() doesn't copy the entire table (like in previous versions). That's great!

But still, <-. in R results in data.table losing it's over allocation. Therefore, when you assign a column by reference with := or set the next time, you'll get a warning message that a shallow copy has been made to get the over-allocation back in track.

Try doing: data.table:::selfrefok(dt) before and after class(dt) <- ... to see the difference.

And this is not really an issue, so added the "question" tag. Feel free to re-open the issue if this doesn't answer (completely) (or if you've further questions).

[jangorecki]

Yeah, my bad I've forgot about setattr. I fully agree it is not issue but only a question.
You focus more on my class() than on the quesiton itself.
But from your comment I assume there should be no any issues related to altered class of data.table object.
Thanks

[arunsrinivasan]

Oh right. So to answer your questions specifically:

I would like to confirm is it safe the add another (my custom) class to DT object?

Yes, it's safe here. The downsides are: a) in versions of R <3.10, it'll copy the entire data.table, and b) in all versions of R, the over-allocation will be lost, which'll result in the warning explained above.

will the rest of data.table functions works fine on my dt or maybe it is not recommended?

Yes it will (and should). But the idiomatic way is setattr.