setDT modifies the class of the origin data.frame

Take this code:

> d1 <- data.frame(a=c(1,2,3,4), b=c(2,3,4,5))
> d2 <- d1
> setDT(d2)

At this point d2 is a shallow copy of d1, meaning they are not the same R object (although they point to the same columns).
d2's class is data.table, as expected, but so is d1

> class(d2)
[1] "data.table" "data.frame"
> class(d1)
[1] "data.table" "data.frame"

Unlike #4783 , I think this is an easy fix. In two places in setDT's source you see the pair of lines -

setattr(x, "class", .resetclass(x, "data.frame"))
setalloccol(x)

setalloccol is what initiates the shallow copy. If the order of the rows were reversed, the class of only the current shallow copy (d2 in the example above) would be modified, as expected.

# Output of sessionInfo()
R version 3.6.1 (2019-07-05)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19041)
Matrix products: default

The fix I proposed doesn't work well with magrittr pipes :(

df<-data.frame(a=1)
df %>% setDT
class(df)  # data.frame

So I can't offer a solution, but the problem remains.

It seems setDT never worked well with magrittr pipes...

> d <- data.frame(a=1:2)
> setDT(d)
> truelength(d)
[1] 1025
> d <- data.frame(a=1:2)
> d %>% setDT
> truelength(d)
[1] 0

Fix Rdatatable#4784: setDT inside a function no longer modifies the c…

setDT generates shallow copy earlier to avoid interfering with attrib…

Navigation Menu

Repository navigation

setDT modifies the class of the origin data.frame #4784

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

setDT modifies the class of the origin data.frame #4784

Description

Activity

OfekShilon commented on Nov 15, 2020

OfekShilon commented on Nov 19, 2020

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions