Question

If I do something like this:

> df <- data.frame()
> rbind(df, c("A","B","C"))
  X.A. X.B. X.C.
1    A    B    C

You can see the row gets added to the empty data frame. However, the columns get named automatically based on the content of the data.

This causes problems if I later want to:

> df <- rbind(df, c("P", "D", "Q"))

Is there a way to control the names of the columns that get automatically created by rbind? Or some other way to do what I’m attempting to do here?

Answer 1

@baha-kev has a good answer regarding strings and factors.

I just want to point out the weird behavior of rbind for data.frame:

# This is "should work", but it doesn't:
# Create an empty data.frame with the correct names and types
df <- data.frame(A=numeric(), B=character(), C=character(), stringsAsFactors=FALSE)
rbind(df, list(42, 'foo', 'bar')) # Messes up names!
rbind(df, list(A=42, B='foo', C='bar')) # OK...

# If you have at least one row, names are kept...
df <- data.frame(A=0, B="", C="", stringsAsFactors=FALSE)
rbind(df, list(42, 'foo', 'bar')) # Names work now...

But if you only have strings then why not use a matrix instead? Then it works fine to start with an empty matrix:

# Create a 0x3 matrix:
m <- matrix('', 0, 3, dimnames=list(NULL, LETTERS[1:3]))

# Now add a row:
m <- rbind(m, c('foo','bar','baz')) # This works fine!
m

# Then optionally turn it into a data.frame at the end...
as.data.frame(m, stringsAsFactors=FALSE)