using R's internal tar function on specific files

R has a handy cross platform tar() function that can tar and gzip files. It seems this function was designed to tar up entire directories. I was hoping to use this function to tar and compress a subset of a directory, or a single file. I c开发者_运维百科an't seem to do this, however. I was expecting the following to tar up a single csv file in the current working directory:

tar( "tst.tgz", "myCsv.csv", compression="gzip" )

So is it only possible to use the tar() function on directories?

I've temporarily gotten around this by creating a temp directory, copying my files, and then tarring the whole temp dir. But I was hoping for a bit simpler solution. That would not require copying the files which is somewhat time consuming for large files.

I don't think that is possible as you describe JD. The files argument is passed to the path argument of list.files, and as a result it works by tarring up files in directories, not individual files.

If you are prepared to edit an internal function, tar() can be made to do what you want, by fiddling with the call to list.files() inside tar(). A bit of fiddling produced the tar2() function below, which has extra arguments to control what list.files() returns. Using this function, we can achieve what you want via a call like this:

tar2("foo.tar", path = ".", pattern = "bar.csv", recursive = FALSE, 
     full.names = FALSE, all.files = FALSE)

The all.files = FALSE is likely redundant unless you have hidden files with names containing "bar.csv".

The recursive = FALSE bit just stops the function searching anywhere but the current directory, which seems what you want and speeds the search up if the working dir has a lot of files and sub-folder.

The full.names = FALSE bit is key. If this if TRUE, list.files() returns the matched filename as "./bar.csv", which tar() would stick in a folder inside the tarball. If we set this to FALSE, list.files() returns "bar.csv", so we get a tarball with a single CSV file as requested.

If you have files with similar names and wish to only find the stated filename, peg it inside the pattern with ^ and $, eg:

tar2("foo.tar", path = ".", pattern = "^bar.csv$", recursive = FALSE, 
     full.names = FALSE, all.files = FALSE)

Here is the modified tar() function as tar2():

tar2 <- function (tarfile, files = NULL, compression = c("none", "gzip", 
    "bzip2", "xz"), compression_level = 6, tar = Sys.getenv("tar"),
    pattern = NULL, all.files = TRUE, recursive = TRUE, full.names = TRUE) 
{
    if (is.character(tarfile)) {
        TAR <- tar
        if (nzchar(TAR) && TAR != "internal") {
            flags <- switch(match.arg(compression), none = "cf", 
                gzip = "zcf", bzip2 = "jcf", xz = "Jcf")
            cmd <- paste(TAR, flags, shQuote(tarfile), paste(shQuote(files), 
                collapse = " "))
            return(invisible(system(cmd)))
        }
        con <- switch(match.arg(compression), none = file(tarfile, 
            "wb"), gzip = gzfile(tarfile, "wb", compress = compression_level), 
            bzip2 = bzfile(tarfile, "wb", compress = compression_level), 
            xz = xzfile(tarfile, "wb", compress = compression_level))
        on.exit(close(con))
    }
    else if (inherits(tarfile, "connection")) 
        con <- tarfile
    else stop("'tarfile' must be a character string or a connection")
    files <- list.files(files, recursive = recursive, all.files = all.files, 
        full.names = full.names, pattern = pattern)
    bf <- unique(dirname(files))
    files <- c(bf[!bf %in% c(".", files)], files)
    for (f in unique(files)) {
        info <- file.info(f)
        if (is.na(info$size)) {
            warning(gettextf("file '%s' not found", f), domain = NA)
            next
        }
        header <- raw(512L)
        if (info$isdir && !grepl("/$", f)) 
            f <- paste(f, "/", sep = "")
        name <- charToRaw(f)
        if (length(name) > 100L) {
            if (length(name) > 255L) 
                stop("file path is too long")
            s <- max(which(name[1:155] == charToRaw("/")))
            if (is.infinite(s) || s + 100 < length(name)) 
                stop("file path is too long")
            warning("storing paths of more than 100 bytes is not portable:\n  ", 
                sQuote(f), domain = NA)
            prefix <- name[1:(s - 1)]
            name <- name[-(1:s)]
            header[345 + seq_along(prefix)] <- prefix
        }
        header[seq_along(name)] <- name
        header[101:107] <- charToRaw(sprintf("%07o", info$mode))
        uid <- info$uid
        if (!is.null(uid) && !is.na(uid)) 
            header[109:115] <- charToRaw(sprintf("%07o", uid))
        gid <- info$gid
        if (!is.null(gid) && !is.na(gid)) 
            header[117:123] <- charToRaw(sprintf("%07o", gid))
        size <- ifelse(info$isdir, 0, info$size)
        header[137:147] <- charToRaw(sprintf("%011o", as.integer(info$mtime)))
        if (info$isdir) 
            header[157L] <- charToRaw("5")
        else {
            lnk <- Sys.readlink(f)
            if (is.na(lnk)) 
                lnk <- ""
            header[157L] <- charToRaw(ifelse(nzchar(lnk), "2", 
                "0"))
            if (nzchar(lnk)) {
                if (length(lnk) > 100L) 
                  stop("linked path is too long")
                header[157L + seq_len(nchar(lnk))] <- charToRaw(lnk)
                size <- 0
            }
        }
        header[125:135] <- charToRaw(sprintf("%011o", as.integer(size)))
        header[258:262] <- charToRaw("ustar")
        header[264:265] <- charToRaw("0")
        s <- info$uname
        if (!is.null(s) && !is.na(s)) {
            ns <- nchar(s, "b")
            header[265L + (1:ns)] <- charToRaw(s)
        }
        s <- info$grname
        if (!is.null(s) && !is.na(s)) {
            ns <- nchar(s, "b")
            header[297L + (1:ns)] <- charToRaw(s)
        }
        header[149:156] <- charToRaw(" ")
        checksum <- sum(as.integer(header))%%2^24
        header[149:154] <- charToRaw(sprintf("%06o", as.integer(checksum)))
        header[155L] <- as.raw(0L)
        writeBin(header, con)
        if (info$isdir || nzchar(lnk)) 
            next
        inf <- file(f, "rb")
        for (i in seq_len(ceiling(info$size/512L))) {
            block <- readBin(inf, "raw", 512L)
            writeBin(block, con)
            if ((n <- length(block)) < 512L) 
                writeBin(raw(512L - n), con)
        }
        close(inf)
    }
    block <- raw(512L)
    writeBin(block, con)
    writeBin(block, con)
    invisible(0L)
}

This has been revised again further simplifying the tar1 code. Also tar1 can now output one file without a directory or multiple files without a directory. Essentially we need to work around a bug in the use of list.files by R in tar and we do that by redefining the list.files that tar1 uses.

The line that manipulates the environment of tar actually creates a copy of tar whose environment is the environment within tar1 so that when the copied tar is run it first looks there for list.files. If we had not made a copy of tar with a new environment then it would have used the list.files in the base of R, ignoring our redefinition.

Below tar1 is a variation of the tar command that produces a tar file whose components having one level (i.e. a single file or set of files without directories). It is assumed that all files are in current directory.

Following the definition of tar1 we test it by creating two files and creating an archive with the first of those files and then with both of those files.

# tar a single file
tar1 <- function(...) {
    list.files <- function(...) ..1
    environment(tar) <- environment()
    tar(...)
}

# test - first create test files, then some test runs of tar1
cat("a", file = "a.csv")
cat("b", file = "a.csv")

tar1("tst.tgz", "a.csv", "gzip")
tar1("tst2.tgz", Sys.glob("*.csv"), "gzip")

What happens if you give files= the character vector it asks for, say by running list.files() (or its synonym dir()) with a suitable pattern? Methinks the help is clear there:

Arguments:

tarfile: The pathname of the tarfile: tilde expansion (see ‘path.expand’) will be performed. Alternatively, a connection that can be used for binary writes.

files: A character vector offilepaths to be archived: the default is to archive all files under the current directory.