fill 0-row data.frame with 1 line of NAs

On 2012-07-10 08:50, Brian Diggs wrote:

On 7/10/2012 7:53 AM, Peter Ehlers wrote:

On 2012-07-10 06:57, Rui Barradas wrote:

Hello,

If you write a function, it becomes less convoluted...


empty <- function(x){
     if(NROW(x) == 0){
         y <- rep(NA, NCOL(x))
         names(y) <- names(x)
         y
     }else x
}

(.xb <- iris[ iris$Species=='zz', ])
empty(.xb)

Both this and Liviu's original solution destroy the
factor nature of 'Species' (which may not matter, of
course). How about

    (.xb <- iris[ iris$Species=='zz', ])
    .xb <- .xb[1, ]   # this probably shouldn't work, but it does.

Using NA subscripting seems even better

Yes, you can subset with NA or any real number greater than 1.

Peter Ehlers

empty <- function(x) {
    if(NROW(x) == 0) {
      x[NA,]
    } else {
      x
    }
}

It even preserves the factor nature of things:

  > empty(iris[iris$Specis=='zz',])

     Sepal.Length Sepal.Width Petal.Length Petal.Width Species
NA           NA          NA           NA          NA    <NA>

  > str(empty(iris[iris$Specis=='zz',]))

'data.frame':   1 obs. of  5 variables:
   $ Sepal.Length: num NA
   $ Sepal.Width : num NA
   $ Petal.Length: num NA
   $ Petal.Width : num NA
   $ Species     : Factor w/ 3 levels "setosa","versicolor",..: NA

?

Peter Ehlers

Hope this helps,

Rui Barradas

Em 10-07-2012 14:15, Liviu Andronic escreveu:

Dear all
Is there a simpler method to achieve the following: When I obtain an
empty data.frame after subsetting, I need for it to contain one line
of NAs. Here's a dummy example:

(.xb <- iris[ iris$Species=='zz', ])

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

dim(.xb)

[1] 0 5

(.xa <- data.frame(matrix(rep(NA, ncol(.xb)), 1)))

     X1 X2 X3 X4 X5
1 NA NA NA NA NA

names(.xa) <- names(.xb)
(.xb <- .xa)

     Sepal.Length Sepal.Width Petal.Length Petal.Width Species
1           NA          NA           NA          NA      NA


The solution I came up with is way too convoluted. Anything simpler?
Regards
Liviu