memCompress: In-memory Compression and Decompression

memCompressR Documentation

In-memory Compression and Decompression


In-memory compression or decompression for raw vectors.


memCompress(from, type = c("gzip", "bzip2", "xz", "none"))

              type = c("unknown", "gzip", "bzip2", "xz", "none"),
              asChar = FALSE)



A raw vector. For memCompress a character vector will be converted to a raw vector with character strings separated by "\n". Types "gzip" and "xz" support long raw vectors as from R 4.0.0.


character string, the type of compression. May be abbreviated to a single letter, defaults to the first of the alternatives.


logical: should the result be converted to a character string? NB: character strings have a limit of 2^31 - 1 bytes, so raw vectors should be used for large inputs.


type = "none" passes the input through unchanged, but may be useful if type is a variable.

type = "unknown" attempts to detect the type of compression applied (if any): this will always succeed for bzip2 compression, and will succeed for other forms if there is a suitable header. It will auto-detect the ‘magic’ header ("\x1f\x8b") added to files by the gzip program (and to files written by gzfile), but memCompress does not add such a header. (It supports RFC 1950 format, sometimes known as ‘zlib’ format, for compression and decompression and RFC 1952 for decompression only.)

gzip compression uses whatever is the default compression level of the underlying library (usually 6).

bzip2 compression always adds a header ("BZh"). The underlying library only supports in-memory (de)compression of up to 2^31 - 1 elements. Compression is equivalent to bzip2 -9 (the default).

Compressing with type = "xz" is equivalent to compressing a file with xz -9e (including adding the ‘magic’ header): decompression should cope with the contents of any file compressed by xz version 4.999 and later, as well as by some versions of lzma. There are other versions, in particular ‘raw’ streams, that are not currently handled.

All the types of compression can expand the input: for "gzip" and "bzip2" the maximum expansion is known and so memCompress can always allocate sufficient space. For "xz" it is possible (but extremely unlikely) that compression will fail if the output would have been too large.


A raw vector or a character string (if asChar = TRUE).

See Also


extSoftVersion for the versions of the zlib, bzip2 and xz libraries in use. for background on data compression,,,,, and for references about the particular schemes used.


txt <- readLines(file.path(R.home("doc"), "COPYING"))
txt.gz <- memCompress(txt, "g")
txt2 <- strsplit(memDecompress(txt.gz, "g", asChar = TRUE), "\n")[[1]]
stopifnot(identical(txt, txt2))
txt.bz2 <- memCompress(txt, "b")
## can auto-detect bzip2:
txt3 <- strsplit(memDecompress(txt.bz2, asChar = TRUE), "\n")[[1]]
stopifnot(identical(txt, txt3))

## xz compression is only worthwhile for large objects
txt.xz <- memCompress(txt, "x")
txt3 <- strsplit(memDecompress(txt.xz, asChar = TRUE), "\n")[[1]]
stopifnot(identical(txt, txt3))