deparseOpts: Options for Expression Deparsing

deparseOptsR Documentation

Options for Expression Deparsing


Process the deparsing options for deparse, dput and dump.






character vector of deparsing options.


..deparseOpts is the character vector of possible deparsing options used by .deparseOpts().

.deparseOpts() is called by deparse, dput and dump to process their control argument.

The control argument is a vector containing zero or more of the following strings (exactly those in ..deparseOpts). Partial string matching is used.


Either surround integer vectors by as.integer() or use suffix L, so they are not converted to type double when parsed. This includes making sure that integer NAs are preserved (via NA_integer_ if there are no non-NA values in the vector, unless "S_compatible" is set).


Surround unevaluated expressions, but not formulas, with quote(), so they are not evaluated when re-parsed.


If the object has attributes (other than a source attribute, see srcref), use structure() to display them as well as the object value unless the only such attribute is names and the "niceNames" option is set. This ("showAttributes") is the default for deparse and dput.


If the object has a source attribute (srcref), display that instead of deparsing the object. Currently only applies to function definitions.


Some exotic objects such as environments, external pointers, etc. can not be deparsed properly. This option causes a warning to be issued if the deparser recognizes one of these situations.

Also, the parser in R < 2.7.0 would only accept strings of up to 8192 bytes, and this option gives a warning for longer strings.


Integer, real and character NAs are surrounded by coercion functions where necessary to ensure that they are parsed to the same type. Since e.g. NA_real_ can be output in R, this is mainly used in connection with S_compatible.


If true, lists and atomic vectors with non-NA names (see names) are deparsed as e.g., c(A = 1) instead of structure(1, .Names = "A"), independently of the "showAttributes" setting.


An abbreviated way to specify all of the options listed above plus "digits17" (since R version 4.0.0). This is the default for dump, and, without "digits17", the options used by edit (which are fixed).


Deparse promises in the form <promise: expression> rather than evaluating them. The value and the environment of the promise will not be shown and the deparsed code cannot be sourced.


Make deparsing as far as possible compatible with S and R < 2.5.0. For compatibility with S, integer values of double vectors are deparsed with a trailing decimal point. Backticks are not used.


Real and finite complex numbers are output in "%a" format as binary fractions (coded as hexadecimal: see sprintf) with maximal opportunity to be recorded exactly to full precision. Complex numbers with one or both non-finite components are output as if this option were not set.

(This relies on that format being correctly supported: known problems on Windows are worked around as from R 3.1.2.)


Real and finite complex numbers are output using format "%.17g" which may give more precision than the default (but the output will depend on the platform and there may be loss of precision when read back). Complex numbers with one or both non-finite components are output as if this option were not set.


An abbreviated way to specify control = c("all", "hexNumeric") which is guaranteed to be exact for numbers, see also below.

For the most readable (but perhaps incomplete) display, use control = NULL. This displays the object's value, but not its attributes. The default in deparse is to display the attributes as well, but not to use any of the other options to make the result parseable. (dump uses more default options via control = "all", and printing of functions without sources uses c("keepInteger", "keepNA") to which one may add "warnIncomplete".)

Using control = "exact" (short for control = c("all", "hexNumeric")) comes closest to making deparse() an inverse of parse() (but we have not yet seen an example where "all", now including "digits17", would not have been as good). However, not all objects are deparse-able even with these options, and a warning will be issued if the function recognizes that it is being asked to do the impossible.

Only one of "hexNumeric" and "digits17" can be specified.


An integer value corresponding to the control options selected.


stopifnot(.deparseOpts("exact") == .deparseOpts(c("all", "hexNumeric")))
(iOpt.all <- .deparseOpts("all")) # a four digit integer

## one integer --> vector binary bits
int2bits <- function(x, base = 2L,
                     ndigits = 1 + floor(1e-9 + log(max(x,1), base))) {
    r <- numeric(ndigits)
    for (i in ndigits:1) {
        r[i] <- x%%base
        if (i > 1L)
            x <- x%/%base
    rev(r) # smallest bit at left
## What options does  "all" contain ? =========
(depO.indiv <- setdiff(..deparseOpts, c("all", "exact")))
(oa <- depO.indiv[int2bits(iOpt.all) == 1])# 8 strings
stopifnot(identical(iOpt.all, .deparseOpts(oa)))

## ditto for "exact" instead of "all":
(iOpt.X <- .deparseOpts("exact"))
data.frame(opts = depO.indiv,
           all  = int2bits(iOpt.all),
           exact= int2bits(iOpt.X))
(oX <- depO.indiv[int2bits(iOpt.X) == 1]) # 8 strings, too
diffXall <- oa != oX
stopifnot(identical(iOpt.X, .deparseOpts(oX)),
          identical(oX[diffXall], "hexNumeric"),
          identical(oa[diffXall], "digits17"))