How to add multiple empty columns to a data.frame with 0 rows?

Question

I can add columns to data.frame:

x <- head(iris)
y <- x[1:nrow(x) > 7, ]
x[c('NewCol1', 'NewCol2')] <- rep(c('a', 'b'), each = nrow(x))

For a data.frame with 0 row, it does not work:

# > y
# [1] Sepal.Length Sepal.Width  Petal.Length Petal.Width  Species     
# <0 rows> (or 0-length row.names)

y[c('NewCol1', 'NewCol2')] <- rep(c('a', 'b'), each = nrow(y))

# Error in value[[jvseq[[jjj]]]] : subscript out of bounds

I found this, Add Columns to an empty data frame in R, but it doesn't help much.

expected output:

# > y
# [1] Sepal.Length Sepal.Width  Petal.Length Petal.Width  Species      NewCol1      NewCol2
# <0 rows> (or 0-length row.names)

Answer 1

Consider the following code which creates an empty data frame:

df <- data.frame(Ints=integer(),
                 Characters=character(),
                 stringsAsFactors=FALSE)

One way to add a new column to this empty data frame is to use cbind():

df2 <- cbind(df, data.frame(Stuff=character(),stringsAsFactors=FALSE))

> df2
[1] Ints       Characters Stuff     
<0 rows> (or 0-length row.names)

Then just add your data as you normally would, e.g.

> df2[1,] <- c(1, "hello", "world")
> df2
  Ints Characters Stuff
1    1      hello world

As you mentioned, this might cause a casting problem in the Ints column. Assigning each column by itself avoids this, e.g.

df2$Ints <- c(1:5)
df2$Stuff <- c("one", "two", "three", "four", "five")

Or, you could use something like read.table to bring in your data, and explicitly assign the classes that way.