Sorting or Ordering Vectors

Sort (or order) a vector or factor (partially) into ascending or descending order. For ordering along more than one variable, e.g., for sorting data frames, see order.

manip, arith, univar
sort(x, decreasing = FALSE, ...)
"sort"(x, decreasing = FALSE, na.last = NA, ...), partial = NULL, na.last = NA, decreasing = FALSE, method = c("shell", "quick"), index.return = FALSE)
for sort an R object with a class or a numeric, complex, character or logical vector. For, a numeric, complex, character or logical vector, or a factor.
logical. Should the sort be increasing or decreasing? Not available for partial sorting.
arguments to be passed to or from methods or (for the default methods and objects without a class) to
for controlling the treatment of NAs. If TRUE, missing values in the data are put last; if FALSE, they are put first; if NA, they are removed.
NULL or a vector of indices for partial sorting.
character string specifying the algorithm used. Not available for partial sorting.
logical indicating if the ordering index vector should be returned as well; this is only available for a few cases, the default na.last = NA and full sorting of non-factors.

sort is a generic function for which methods can be written, and is the internal method which is compatible with S if only the first three arguments are used.

The default sort method makes use of order for classed objects, which in turn makes use of the generic function xtfrm (and can be slow unless a xtfrm method has been defined or is.numeric(x) is true).

Complex values are sorted first by the real part, then the imaginary part.

The sort order for character vectors will depend on the collating sequence of the locale in use: see Comparison. The sort order for factors is the order of their levels (which is particularly appropriate for ordered factors).

If partial is not NULL, it is taken to contain indices of elements of the result which are to be placed in their correct positions in the sorted array by partial sorting. For each of the result values in a specified position, any values smaller than that one are guaranteed to have a smaller index in the sorted array and any values which are greater are guaranteed to have a bigger index in the sorted array. (This is included for efficiency, and many of the options are not available for partial sorting. It is only substantially more efficient if partial has a handful of elements, and a full sort is done (a Quicksort if possible) if there are more than 10.) Names are discarded for partial sorting.

Method "shell" uses Shellsort (an $O(n^{4/3})$ variant from Sedgewick (1986)). If x has names a stable modification is used, so ties are not reordered. (This only matters if names are present.)

Method "quick" uses Singleton (1969)'s implementation of Hoare's Quicksort method and is only available when x is numeric (double or integer) and partial is NULL. (For other types of x Shellsort is used, silently.) It is normally somewhat faster than Shellsort (perhaps 50% faster on vectors of length a million and twice as fast at a billion) but has poor performance in the rare worst case. (Peto's modification using a pseudo-random midpoint is used to make the worst case rarer.) This is not a stable sort, and ties may be reordered.

Factors with less than 100,000 levels are sorted by radix sorting when method is not supplied: see sort.list.


For sort, the result depends on the S3 method which is dispatched. If x does not have a class is used and it description applies. For classed objects which do not have a specific method the default method will be used and is equivalent to x[order(x, ...)]: this depends on the class having a suitable method for [ (and also that order will work, which is not the case for a class based on a list).For the value is the sorted vector unless index.return is true, when the result is a list with components named x and ix containing the sorted numbers and the ordering index vector. In the latter case, if method == "quick" ties may be reversed in the ordering (unlike sort.list) as quicksort is not stable. NB: the index vector refers to element numbers after removal of NAs: see order if you want the original element numbers.All attributes are removed from the return value (see Becker et al, 1988, p.146) except names, which are sorted. (If partial is specified even the names are removed.) Note that this means that the returned value has no class, except for factors and ordered factors (which are treated specially and whose result is transformed back to the original class).


Becker, R. A., Chambers, J. M. and Wilks, A. R. (1988) The New S Language. Wadsworth & Brooks/Cole.

Sedgewick, R. (1986) A new upper bound for Shell sort. J. Algorithms 7, 159--173.

Singleton, R. C. (1969) An efficient algorithm for sorting with minimal storage: Algorithm 347. Communications of the ACM 12, 185--187.

See Also

Comparison’ for how character strings are collated.

order for sorting on or reordering multiple variables.

is.unsorted. rank.

  • sort
  • sort.default
  • sort.POSIXlt
library(base) require(stats) x <- swiss$Education[1:25] x; sort(x); sort(x, partial = c(10, 15)) ## illustrate 'stable' sorting (of ties): sort(c(10:3, 2:12), method = "sh", index.return = TRUE) # is stable ## $x : 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 12 ## $ix: 9 8 10 7 11 6 12 5 13 4 14 3 15 2 16 1 17 18 19 sort(c(10:3, 2:12), method = "qu", index.return = TRUE) # is not ## $x : 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 12 ## $ix: 9 10 8 7 11 6 12 5 13 4 14 3 15 16 2 17 1 18 19 x <- c(1:3, 3:5, 10) is.unsorted(x) # FALSE: is sorted is.unsorted(x, strictly = TRUE) # TRUE : is not (and cannot be) # sorted strictly ## Not run: # ## Small speed comparison simulation: # N <- 2000 # Sim <- 20 # rep <- 1000 # << adjust to your CPU # c1 <- c2 <- numeric(Sim) # for(is in seq_len(Sim)){ # x <- rnorm(N) # c1[is] <- system.time(for(i in 1:rep) sort(x, method = "shell"))[1] # c2[is] <- system.time(for(i in 1:rep) sort(x, method = "quick"))[1] # stopifnot(sort(x, method = "s") == sort(x, method = "q")) # } # rbind(ShellSort = c1, QuickSort = c2) # cat("Speedup factor of quick sort():\n") # summary({qq <- c1 / c2; qq[is.finite(qq)]}) # # ## A larger test # x <- rnorm(1e7) # system.time(x1 <- sort(x, method = "shell")) # system.time(x2 <- sort(x, method = "quick")) # stopifnot(identical(x1, x2)) # ## End(Not run)
Documentation reproduced from package base, version 3.1.3, License: Part of R 3.1.3

Community examples at Mar 16, 2018 base v3.4.3