Why is my sapply function building a vector of integers?

Question

I have two related questions -- I'm trying to learn R properly, so I'm doing some homework problems from an R course. They have us writing a function to return a vector of correlations:

example.function <- function(threshold = 0) {
  example.vector <- vector()
  example.vector <- sapply(1:30, function(i) {
    complete.record.count <- # ... counts the complete records in each of the 30 files.
    ## Cutting for space and to avoid giving away answers.
    ## a few lines get the complete records in each 
    ## file and count them. 
    if(complete.record.count > threshold) {
      new.correlation <- cor(complete.record$val1, complete.record$val2)
      print(new.correlation)
      example.vector <- c(new.correlation, example.vector)
    }  
  })
  # more null value handling#
  return(example.vector)
}

As the function runs it prints the correlation value to stdout. The values it prints are accurate to six decimal points. So I know I'm getting a good value for new.correlation. The vector that is returned doesn't include those values. Instead, it is whole numbers in sequence.

> tmp <- example.function()
> head(tmp)
[1] 2 3 4 5 6 7

I can't figure out why sapply is pushing integers into the vector? What am I missing here?

I actually don't understand the core structure, which is more or less:

some.vector <- vector()
some.vector <- sapply(range, function(i) {
  some.vector <- c(new.value,some.vector)
}

that seems awfully un-R-like in its redundancy. Tips?

Answer 1

If you use sapply you don't need to create the vector yourself and you don't need to grow it (sapply takes care of all that). You probably want something like this:

example.function <- function(threshold = 0) {
  example.vector <- sapply(1:30, function(i) {
    ## Cutting for space and to avoid giving away answers.
    ## a few lines get the complete records in each 
    ## file and count them. 
    if(complete.record.count > threshold) {
      new.correlation <- cor(complete.record$val1, complete.record$val2)
      }  else {
        new.correlation <- NA   
      }
    new.correlation #return value of anonymous function
  })
  # more null value handling#
  example.vector #return value of example.function
}

However, it is unclear how the index i factors into the anonymous function and the question is not reproducible ...