Save R Objects

Description

save writes an external representation of R objects to the specified file. The objects can be read back from the file at a later date by using the function load or attach (or data in some cases).

save.image() is just a short-cut for ‘save my current workspace’, i.e., save(list = ls(all.names = TRUE), file = ".RData", envir = .GlobalEnv). It is also what happens with q("yes").

Usage

save(..., list = character(),
     file = stop("'file' must be specified"),
     ascii = FALSE, version = NULL, envir = parent.frame(),
     compress = isTRUE(!ascii), compression_level,
     eval.promises = TRUE, precheck = TRUE)

save.image(file = ".RData", version = NULL, ascii = FALSE,
           compress = !ascii, safe = TRUE)

Arguments

Details

The names of the objects specified either as symbols (or character strings) in ... or as a character vector in list are used to look up the objects from environment envir. By default promises are evaluated, but if eval.promises = FALSE promises are saved (together with their evaluation environments). (Promises embedded in objects are always saved unevaluated.)

All R platforms use the XDR (bigendian) representation of C ints and doubles in binary save-d files, and these are portable across all R platforms.

ASCII saves used to be useful for moving data between platforms but are now mainly of historical interest. They can be more compact than binary saves where compression is not used, but are almost always slower to both read and write: binary saves compress much better than ASCII ones. Further, decimal ASCII saves may not restore double/complex values exactly, and what value is restored may depend on the R platform.

Default values for the ascii, compress, safe and version arguments can be modified with the "save.defaults" option (used both by save and save.image), see also the ‘Examples’ section. If a "save.image.defaults" option is set it is used in preference to "save.defaults" for function save.image (which allows this to have different defaults). In addition, compression_level can be part of the "save.defaults" option.

A connection that is not already open will be opened in mode "wb". Supplying a connection which is open and not in binary mode gives an error.

Compression

Large files can be reduced considerably in size by compression. A particular 46MB R object was saved as 35MB without compression in 2 seconds, 22MB with gzip compression in 8 secs, 19MB with bzip2 compression in 13 secs and 9.4MB with xz compression in 40 secs. The load times were 1.3, 2.8, 5.5 and 5.7 seconds respectively. These results are indicative, but the relative performances do depend on the actual file: xz compressed unusually well here.

It is possible to compress later (with gzip, bzip2 or xz) a file saved with compress = FALSE: the effect is the same as saving with compression. Also, a saved file can be uncompressed and re-compressed under a different compression scheme (and see resaveRdaFiles for a way to do so from within R).

Parallel compression

That file can be a connection can be exploited to make use of an external parallel compression utility such as pigz (http://zlib.net/pigz/) or pbzip2 (https://launchpad.net/pbzip2) via a pipe connection. For example, using 8 threads,

    con <- pipe("pigz -p8 > fname.gz", "wb")
    save(myObj, file = con); close(con)

    con <- pipe("pbzip2 -p8 -9 > fname.bz2", "wb")
    save(myObj, file = con); close(con)

    con <- pipe("xz -T8 -6 -e > fname.xz", "wb")
    save(myObj, file = con); close(con)

where the last requires xz 5.1.1 or later built with support for multiple threads (and parallel compression is only effective for large objects: at level 6 it will compress in serialized chunks of 12MB).

Warnings

The ... arguments only give the names of the objects to be saved: they are searched for in the environment given by the envir argument, and the actual objects given as arguments need not be those found.

Saved R objects are binary files, even those saved with ascii = TRUE, so ensure that they are transferred without conversion of end-of-line markers and of 8-bit characters. The lines are delimited by LF on all platforms.

Although the default version has not changed since R 1.4.0, this does not mean that saved files are necessarily backwards compatible. You will be able to load a saved image into an earlier version of R unless use is made of later additions (for example, raw vectors, external pointers and some S4 objects).

One such ‘later addition’ was long vectors, introduced in R 3.0.0 and loadable only on 64-bit platforms.

Loading files saved with ASCII = NA requires a C99-compliant C function sscanf: this is a problem on Windows, first worked around in R 3.1.2: they should be readable in earlier versions of R on all other platforms.

Note

For saving single R objects, saveRDS() is mostly preferable to save(), notably because of the functional nature of readRDS(), as opposed to load().

The most common reason for failure is lack of write permission in the current directory. For save.image and for saving at the end of a session this will shown by messages like

    Error in gzfile(file, "wb") : unable to open connection
    In addition: Warning message:
    In gzfile(file, "wb") :
      cannot open compressed file '.RDataTmp',
      probable reason 'Permission denied'

See Also

dput, dump, load, data.

For other interfaces to the underlying serialization format, see serialize and saveRDS.

Examples

x <- stats::runif(20)
y <- list(a = 1, b = TRUE, c = "oops")
save(x, y, file = "xy.RData")
save.image() # creating ".RData" in current working directory
unlink("xy.RData")

# set save defaults using option:
options(save.defaults = list(ascii = TRUE, safe = FALSE))
save.image() # creating ".RData"
if(interactive()) withAutoprint({
   file.info(".RData")
   readLines(".RData", n = 7) # first 7 lines; first starts w/ "RDA"..
})
unlink(".RData")