formatc Formatting Using C-style Formats

Description

formatC() formats numbers individually and flexibly using C style format specifications.

prettyNum() is used for “prettifying” (possibly formatted) numbers, also in format.default.

.format.zeros(x), an auxiliary function of prettyNum(), re-formats the zeros in a vector x of formatted numbers.

Usage

formatC(x, digits = NULL, width = NULL,
        format = NULL, flag = "", mode = NULL,
        big.mark = "", big.interval = 3L,
        small.mark = "", small.interval = 5L,
        decimal.mark = getOption("OutDec"),
        preserve.width = "individual",
        zero.print = NULL, replace.zero = TRUE,
        drop0trailing = FALSE)

prettyNum(x, big.mark = "",   big.interval = 3L,
          small.mark  = "", small.interval = 5L,
          decimal.mark = getOption("OutDec"), input.d.mark = decimal.mark,
          preserve.width = c("common", "individual", "none"),
          zero.print = NULL, replace.zero = FALSE,
          drop0trailing = FALSE, is.cmplx = NA,
          ...)

.format.zeros(x, zero.print, nx = suppressWarnings(as.numeric(x)),
              replace = FALSE, warn.non.fitting = TRUE)

Arguments

x	an atomic numerical or character object, possibly complex only for prettyNum(), typically a vector of real numbers. Any class is discarded, with a warning.
digits	the desired number of digits after the decimal point (format = "f") or significant digits (format = "g", = "e" or = "fg"). Default: 2 for integer, 4 for real numbers. If less than 0, the C default of 6 digits is used. If specified as more than 50, 50 will be used with a warning unless format = "f" where it is limited to typically 324. (Not more than 15–21 digits need be accurate, depending on the OS and compiler used. This limit is just a precaution against segfaults in the underlying C runtime.)
width	the total field width; if both digits and width are unspecified, width defaults to 1, otherwise to digits + 1. width = 0 will use width = digits, width < 0 means left justify the number in this field (equivalent to flag = "-"). If necessary, the result will have more characters than width. For character data this is interpreted in characters (not bytes nor display width).
format	equal to "d" (for integers), "f", "e", "E", "g", "G", "fg" (for reals), or "s" (for strings). Default is "d" for integers, "g" for reals. "f" gives numbers in the usual xxx.xxx format; "e" and "E" give n.ddde+nn or n.dddE+nn (scientific format); "g" and "G" put x[i] into scientific format only if it saves space to do so and drop trailing zeros and decimal point - unless flag contains "#" which keeps trailing zeros for the "g", "G" formats. "fg" (our own hybrid format) uses fixed format as "f", but digits as the minimum number of significant digits. This can lead to quite long result strings, see examples below. Note that unlike signif this prints large numbers with more significant digits than digits. Trailing zeros are dropped in this format, unless flag contains "#".
flag	for formatC, a character string giving a format modifier as in Kernighan and Ritchie (1988, page 243) or the C+99 standard. "0" pads leading zeros; "-" does left adjustment, "+" ensures a sign in all cases, i.e., "+" for positive numbers , " " if the first character is not a sign, the space character " " will be used instead. "#" specifies “an alternative output form”, specifically depending on format. "'" on some platform–locale combination, activates “thousands' grouping” for decimal conversion, "I" in some versions of ‘glibc’ allow for integer conversion to use the locale's alternative output digits, if any. There can be more than one of these flags, in any order. Other characters used to have no effect for character formatting, but signal an error since R 3.4.0.
mode	"double" (or "real"), "integer" or "character". Default: Determined from the storage mode of x.
big.mark	character; if not empty used as mark between every big.interval decimals before (hence big) the decimal point.
big.interval	see big.mark above; defaults to 3.
small.mark	character; if not empty used as mark between every small.interval decimals after (hence small) the decimal point.
small.interval	see small.mark above; defaults to 5.
decimal.mark	the character to be used to indicate the numeric decimal point.
input.d.mark	if x is character, the character known to have been used as the numeric decimal point in x.
preserve.width	string specifying if the string widths should be preserved where possible in those cases where marks (big.mark or small.mark) are added. "common", the default, corresponds to format-like behavior whereas "individual" is the default in formatC(). Value can be abbreviated.
zero.print	logical, character string or NULL specifying if and how zeros should be formatted specially. Useful for pretty printing ‘sparse’ objects.
replace.zero, replace	logical; if zero.print is a character string, indicates if the exact zero entries in x should be simply replaced by zero.print. Otherwise, depending on the widths of the respective strings, the (formatted) zeroes are partly replaced by zero.print and then padded with " " to the right were applicable. In that case (false replace[.zero]), if the zero.print string does not fit, a warning is produced (if warn.non.fitting is true). This works via prettyNum(), which calls .format.zeros(*, replace=replace.zero) three times in this case, see the ‘Details’.
warn.non.fitting	logical; if it is true, replace[.zero] is false and the zero.print string does not fit, a warning is signalled.
drop0trailing	logical, indicating if trailing zeros, i.e., "0" after the decimal mark, should be removed; also drops "e+00" in exponential formats. This is simply passed to prettyNum(), see the ‘Details’.
is.cmplx	optional logical, to be used when x is "character" to indicate if it stems from complex vector or not. By default (NA), x is checked to ‘look like’ complex.
...	arguments passed to format.
nx	numeric vector of the same length as x, typically the numbers of which the character vector x is the pre-format.

Details

For numbers, formatC() calls prettyNum() when needed which itself calls .format.zeros(*, replace=replace.zero). (“when needed”: when zero.print is not NULL, drop0trailing is true, or one of big.mark, small.mark, or decimal.mark is not at default.)

If you set format it overrides the setting of mode, so formatC(123.45, mode = "double", format = "d") gives 123.

The rendering of scientific format is platform-dependent: some systems use n.ddde+nnn or n.dddenn rather than n.ddde+nn.

formatC does not necessarily align the numbers on the decimal point, so formatC(c(6.11, 13.1), digits = 2, format = "fg") gives c("6.1", " 13"). If you want common formatting for several numbers, use format.

prettyNum is the utility function for prettifying x. x can be complex (or format(<complex>)), here. If x is not a character, format(x[i], ...) is applied to each element, and then it is left unchanged if all the other arguments are at their defaults. Use the input.d.mark argument for prettyNum(x) when x is a character vector not resulting from something like format(<number>) with a period as decimal mark.

Because gsub is used to insert the big.mark and small.mark, special characters need escaping. In particular, to insert a single backslash, use "\\\\".

The C doubles used for R numerical vectors have signed zeros, which formatC may output as -0, -0.000 ....

There is a warning if big.mark and decimal.mark are the same: that would be confusing to those reading the output.

Value

A character object of same size and attributes as x (after discarding any class), in the current locale's encoding.

Unlike format, each number is formatted individually. Looping over each element of x, the C function sprintf(...) is called for numeric inputs (inside the C function str_signif).

formatC: for character x, do simple (left or right) padding with white space.

Note

The default for decimal.mark in formatC() was changed in R 3.2.0: for use within print methods in packages which might be used with earlier versions: use decimal.mark = getOption("OutDec") explicitly.

Author(s)

formatC was originally written by Bill Dunlap for S-PLUS, later much improved by Martin Maechler.

It was first adapted for R by Friedrich Leisch and since much improved by the R Core team.

References

Kernighan, B. W. and Ritchie, D. M. (1988) The C Programming Language. Second edition. Prentice Hall.

See Also

format.

sprintf for more general C-like formatting.

Examples

xx <- pi * 10^(-5:4)
cbind(format(xx, digits = 4), formatC(xx))
cbind(formatC(xx, width = 9, flag = "-"))
cbind(formatC(xx, digits = 5, width = 8, format = "f", flag = "0"))
cbind(format(xx, digits = 4), formatC(xx, digits = 4, format = "fg"))

f <- (-2:4); f <- f*16^f
# Default ("g") format:
formatC(pi*f)
# Fixed ("f") format, more than one flag ('width' partly "enlarged"):
cbind(formatC(pi*f, digits = 3, width=9, format = "f", flag = "0+"))

formatC(    c("a", "Abc", "no way"), width = -7)  # <=> flag = "-"
formatC(c((-1:1)/0,c(1,100)*pi), width = 8, digits = 1)

## note that some of the results here depend on the implementation
## of long-double arithmetic, which is platform-specific.
xx <- c(1e-12,-3.98765e-10,1.45645e-69,1e-70,pi*1e37,3.44e4)
##       1        2             3        4      5       6
formatC(xx)
formatC(xx, format = "fg")       # special "fixed" format.
formatC(xx[1:4], format = "f", digits = 75) #>> even longer strings

formatC(c(3.24, 2.3e-6), format = "f", digits = 11)
formatC(c(3.24, 2.3e-6), format = "f", digits = 11, drop0trailing = TRUE)

r <- c("76491283764.97430", "29.12345678901", "-7.1234", "-100.1","1123")
## American:
prettyNum(r, big.mark = ",")
## Some Europeans:
prettyNum(r, big.mark = "'", decimal.mark = ",")

(dd <- sapply(1:10, function(i) paste((9:0)[1:i], collapse = "")))
prettyNum(dd, big.mark = "'")

## examples of 'small.mark'
pN <- stats::pnorm(1:7, lower.tail = FALSE)
cbind(format (pN, small.mark = " ", digits = 15))
cbind(formatC(pN, small.mark = " ", digits = 17, format = "f"))

cbind(ff <- format(1.2345 + 10^(0:5), width = 11, big.mark = "'"))
## all with same width (one more than the specified minimum)

## individual formatting to common width:
fc <- formatC(1.234 + 10^(0:8), format = "fg", width = 11, big.mark = "'")
cbind(fc)
## Powers of two, stored exactly, formatted individually:
pow.2 <- formatC(2^-(1:32), digits = 24, width = 1, format = "fg")
## nicely printed (the last line showing 5^32 exactly):
noquote(cbind(pow.2))

## complex numbers:
r <- 10.0000001; rv <- (r/10)^(1:10)
(zv <- (rv + 1i*rv))
op <- options(digits = 7) ## (system default)
(pnv <- prettyNum(zv))
stopifnot(pnv == "1+1i", pnv == format(zv),
          pnv == prettyNum(zv, drop0trailing = TRUE))
## more digits change the picture:
options(digits = 8)
head(fv <- format(zv), 3)
prettyNum(fv)
prettyNum(fv, drop0trailing = TRUE) # a bit nicer
options(op)

## The  '  flag :
doLC <- FALSE # <= R warns, so change to TRUE manually if you want see the effect
if(doLC) {
  oldLC <- Sys.getlocale("LC_NUMERIC")
           Sys.setlocale("LC_NUMERIC", "de_CH.UTF-8") }
formatC(1.234 + 10^(0:4), format = "fg", width = 11, flag = "'")
## -->  .....  "      1'001" "     10'001"   on supported platforms
if(doLC) ## revert, typically to  "C"  :
  Sys.setlocale("LC_NUMERIC", oldLC)