Data Frame Indexing Bug

data_frame_indexing_bug.r

# This is a reduction which seems to indicate an issue when 
# adding a column using an arbitrary set of indexes.

# First, let's create a data-frame with some random values:
s <- data.frame(x = runif(10), y = runif(10))

# Now, two randomly generated lists of numbers that we'll use to try to index
# This could be created thusly:
# wrong <- sample(1:nrow(s), nrow(s) * 0.8), etc. 
wrong <- c(3, 6, 7, 5, 1, 2, 9, 8)
right <- c(4, 10, 2, 6, 1, 5, 9, 3)

# Indexing the df with the vector without 10 (the length of the data.frame s) 
# generates an error when creating a new column:
s$z[wrong] <- 1
> Error in `$<-.data.frame`(`*tmp*`, "z", value = c(1, 1, 1, NA, 1, 1, 1,  : 
  replacement has 9 rows, data has 10

# Whereas using the vector *with* 10 does work when creating a new column:
s$z[right] <- 1
> s
            x          y  z
1  0.28089503 0.60456879  1
2  0.37238938 0.79486150  1
3  0.76933910 0.11040848  1
4  0.90518539 0.09314296  1
5  0.62716400 0.81500552  1
6  0.64327678 0.58516840  1
7  0.05474681 0.23425909 NA
8  0.83767040 0.73871715 NA
9  0.01475207 0.23033121  1
10 0.18314193 0.10045413  1

Strangely, when I define a function called f with the same content, and then call it with the arguments
f(wrong, z, 1)
the function works as expected, and returns the data frame

       x         y  z

1 0.6707523 0.6788389 1
2 0.7418809 0.4239940 1
3 0.5061432 0.2350612 1
4 0.3961475 0.8396474 NA
5 0.2268333 0.8911401 1
6 0.5864661 0.6671142 1
7 0.2140100 0.7586803 1
8 0.9915062 0.9027944 1
9 0.4577299 0.9190519 1
10 0.3907814 0.6885121 NA

Weird!