IDEA: Add dfenv()?

[HenrikBengtsson]

Would it make sense to also have a data frame environment that is an environment but emulates how a data frame works? For example,

df <- dfenv(a=1:3, b=c("x", "y", "z"))
df <- cbind(df, c=c(TRUE, FALSE, TRUE))
df$d <- 5:7

[HenrikBengtsson]

• A prerequisite for this may be adding support for dimensions to list environments, cf. Issue Support dimensions for list environments? #11.

[HenrikBengtsson]

This should now be much easier given that dimensions are supported together with subsetting along dimensions.

Here's the basic framework:

> df <- listenv()
> dim(df) <- c(10,2)
> colnames(df) <- c("a", "b")
> df[,"a"] <- 1:10
> df[,"b"] <- letters[1:10]
> A 'listenv' matrix with 20 unnamed elements arranged in 10x2 unnamed rows and columns ('a', 'b').
> unlist(df[,"a"])
 [1]  1  2  3  4  5  6  7  8  9 10
> unlist(df[,"b"])
 [1] "a" "b" "c" "d" "e" "f" "g" "h" "i" "j"

For a dfenv class inheriting from listenv, operators $ and $<- needs to be overridden. The should call corresponding single-column subsetting methods. The latter should in turn coerce such that the elements belonging to the same column should all have the same data type. The constructor dfenv() should emulate data.frame() in how it validates and interprets the arguments.

[HenrikBengtsson]

In the same vein, one could imagine a tibbleenv that emulates tibble objects while being environments.