head Return the First or Last Parts of an Object

Description

Returns the first or last parts of a vector, matrix, table, data frame or function. Since head() and tail() are generic functions, they may also have been extended to other classes.

Usage

head(x, ...)
## Default S3 method:
head(x, n = 6L, ...)

## S3 method for class 'matrix'
head(x, n = 6L, ...) # is exported as head.matrix()
## NB: The methods for 'data.frame' and 'array'  are identical to the 'matrix' one

## S3 method for class 'ftable'
head(x, n = 6L, ...)
## S3 method for class 'function'
head(x, n = 6L, ...)


tail(x, ...)
## Default S3 method:
tail(x, n = 6L, keepnums = FALSE, addrownums, ...)
## S3 method for class 'matrix'
tail(x, n = 6L, keepnums = TRUE, addrownums, ...) # exported as tail.matrix()
## NB: The methods for 'data.frame', 'array', and 'table'
##     are identical to the  'matrix'  one

## S3 method for class 'ftable'
tail(x, n = 6L, keepnums = FALSE, addrownums, ...)
## S3 method for class 'function'
tail(x, n = 6L, ...)

Arguments

x

an object

n

an integer vector of length up to dim(x) (or 1, for non-dimensioned objects). Values specify the indices to be selected in the corresponding dimension (or along the length) of the object. A positive value of n[i] includes the first/last n[i] indices in that dimension, while a negative value excludes the last/first abs(n[i]), including all remaining indices. NA or non-specified values (when length(n) < length(dim(x))) select all indices in that dimension. Must contain at least one non-missing value.

keepnums

in each dimension, if no names in that dimension are present, create them using the indices included in that dimension. Ignored if dim(x) is NULL or its length 1.

addrownums

deprecated - keepnums should be used instead. Taken as the value of keepnums if it is explicitly set when keepnums is not.

...

arguments to be passed to or from other methods.

Details

For vector/array based objects, head() (tail()) returns a subset of the same dimensionality as x, usually of the same class. For historical reasons, by default they select the first (last) 6 indices in the first dimension ("rows") or along the length of a non-dimensioned vector, and the full extent (all indices) in any remaining dimensions. head.matrix() and tail.matrix() are exported.

The default and array(/matrix) methods for head() and tail() are quite general. They will work as is for any class which has a dim() method, a length() method (only required if dim() returns NULL), and a [ method (that accepts the drop argument and can subset in all dimensions in the dimensioned case).

For functions, the lines of the deparsed function are returned as character strings.

When x is an array(/matrix) of dimensionality two and more, tail() will add dimnames similar to how they would appear in a full printing of x for all dimensions k where n[k] is specified and non-missing and dimnames(x)[[k]] (or dimnames(x) itself) is NULL. Specifically, the form of the added dimnames will vary for different dimensions as follows:

k=1 (rows):

"[n,]" (right justified with whitespace padding)

k=2 (columns):

"[,n]" (with no whitespace padding)

k>2 (higher dims):

"n", i.e., the indices as character values

Setting keepnums = FALSE suppresses this behaviour.

As data.frame subsetting (‘indexing’) keeps attributes, so do the head() and tail() methods for data frames.

Value

An object (usually) like x but generally smaller. Hence, for arrays, the result corresponds to x[.., drop=FALSE]. For ftable objects x, a transformed format(x).

Note

For array inputs the output of tail when keepnums is TRUE, any dimnames vectors added for dimensions >2 are the original numeric indices in that dimension as character vectors. This means that, e.g., for 3-dimensional array arr, tail(arr, c(2,2,-1))[ , , 2] and tail(arr, c(2,2,-1))[ , , "2"] may both be valid but have completely different meanings.

Author(s)

Patrick Burns, improved and corrected by R-Core. Negative argument added by Vincent Goulet. Multi-dimension support added by Gabriel Becker.

Examples

head(letters)
head(letters, n = -6L)

head(freeny.x, n = 10L)
head(freeny.y)

head(iris3)
head(iris3, c(6L, 2L))
head(iris3, c(6L, -1L, 2L))

tail(letters)
tail(letters, n = -6L)

tail(freeny.x)
## the bottom-right "corner" :
tail(freeny.x, n = c(4, 2))
tail(freeny.y)

tail(iris3)
tail(iris3, c(6L, 2L))
tail(iris3, c(6L, -1L, 2L))

## iris with dimnames stripped
a3d <- iris3 ; dimnames(a3d) <- NULL
tail(a3d, c(6, -1, 2)) # keepnums = TRUE is default here!
tail(a3d, c(6, -1, 2), keepnums = FALSE)

## data frame w/ a (non-standard) attribute:
treeS <- structure(trees, foo = "bar")
(n <- nrow(treeS))
stopifnot(exprs = { # attribute is kept
    identical(htS <- head(treeS), treeS[1:6, ])
    identical(attr(htS, "foo") , "bar")
    identical(tlS <- tail(treeS), treeS[(n-5):n, ])
    ## BUT if I use "useAttrib(.)", this is *not* ok, when n is of length 2:
    ## --- because [i,j]-indexing of data frames *also* drops "other" attributes ..
    identical(tail(treeS, 3:2), treeS[(n-2):n, 2:3] )
})

tail(library) # last lines of function

head(stats::ftable(Titanic))

## 1d-array (with named dim) :
a1 <- array(1:7, 7); names(dim(a1)) <- "O2"
stopifnot(exprs = {
  identical( tail(a1, 10), a1)
  identical( head(a1, 10), a1)
  identical( head(a1, 1), a1 [1 , drop=FALSE] ) # was a1[1] in R <= 3.6.x
  identical( tail(a1, 2), a1[6:7])
  identical( tail(a1, 1), a1 [7 , drop=FALSE] ) # was a1[7] in R <= 3.6.x
})

Copyright (©) 1999–2012 R Foundation for Statistical Computing.
Licensed under the GNU General Public License.