Why does R allow new column creation via tail-slicing?

If the assignment is a vector, R appears to create the vector z to add to the data.frame, and that vector needs to have the same length as the number rows in the data.frame.

It's more clear what is happening if you work with a list object instead:

df <- list(a = 1:3, b = 4:6)
df$z1[1:2] <- 7:8
df$z2[2:3] <- 7:8
df$z3[c(1,3)] <- 7:8
df
#> $a
#> [1] 1 2 3
#> 
#> $b
#> [1] 4 5 6
#> 
#> $z1
#> [1] 7 8
#> 
#> $z2
#> [1] NA  7  8
#> 
#> $z3
#> [1]  7 NA  8
data.frame(df)
#> Error in (function (..., row.names = NULL, check.rows = FALSE, check.names = TRUE, : arguments imply differing number of rows: 3, 2
data.frame(df[-3])
#>   a b z2 z3
#> 1 1 4 NA  7
#> 2 2 5  7 NA
#> 3 3 6  8  8

The error is coming from the $<-.data.frame function

> `$<-.data.frame`
function (x, name, value) 
{
    cl <- oldClass(x)
    class(x) <- NULL
    nrows <- .row_names_info(x, 2L)
    if (!is.null(value)) {
        N <- NROW(value)
        if (N > nrows) 
            stop(sprintf(ngettext(N, "replacement has %d row, data has %d", 
                "replacement has %d rows, data has %d"), N, nrows), 
                domain = NA)
...

i.e. the condition for N > nrows is satisfied

> NROW(c(7, 8))
[1] 2
> .row_names_info(df, 2L)
[1] 3

which is confirmed by traceback() on the error

> traceback()
3: stop(sprintf(ngettext(N, "replacement has %d row, data has %d", 
       "replacement has %d rows, data has %d"), N, nrows), domain = NA)
2: `$<-.data.frame`(`*tmp*`, z, value = c(7, 8))
1: `$<-`(`*tmp*`, z, value = c(7, 8))

Of course, it's the $<-.data.frame method and its error message, but here we have N < nrows, not N > nrows as @akrun claims. Everying else is correct of course, it's just rather this part of the code (further below in the function body) that signals the error (lines 1195ff in the source <R>/src/library/base/R/dataframe.R):

        if (N < nrows)
            if (N > 0L && (nrows %% N == 0L) && length(dim(value)) <= 1L)
                value <- rep(value, length.out = nrows)
            else
                stop(sprintf(ngettext(N,
                                      "replacement has %d row, data has %d",
                                      "replacement has %d rows, data has %d"),
                             N, nrows), domain = NA)

Note that this also shows in which cases such replacement does not signal an error, namely when nrows is a multiple of N, i.e, if you had 4 rows instead of 3, then recycling would happen.

Just to add one remark, if you use this syntaxe it works :

df = data.frame(a=c(1, 2, 3, 4),
                b=c(4, 5, 6, 6))

df[c(2, 3), 3] <- 1

It appears that if you use df$something, you need to instanciate the last row.

df$z[4] <- 1 # works

But :

df$z[1/2/3] <- 1 # do not works

However, if you do it in two steps it works :

df$z <- 1
df$z[1/2/3] <- 1

So I have no good answer but it is maybe one step to the answer. Akrun gives the answer.