tar-fs

filesystem bindings for tar-stream.

npm install tar-fs

Usage

tar-fs allows you to pack directories into tarballs and extract tarballs into directories.

It doesn't gunzip for you, so if you want to extract a .tar.gz with this you'll need to use something like gunzip-maybe in addition to this.

var tar = require('tar-fs')
var fs = require('fs')

// packing a directory
tar.pack('./my-directory').pipe(fs.createWriteStream('my-tarball.tar'))

// extracting a directory
fs.createReadStream('my-other-tarball.tar').pipe(tar.extract('./my-other-directory'))

To ignore various files when packing or extracting add a ignore function to the options. ignore is also an alias for filter. Additionally you get header if you use ignore while extracting. That way you could also filter by metadata.

var pack = tar.pack('./my-directory', {
  ignore: function(name) {
    return path.extname(name) === '.bin' // ignore .bin files when packing
  }
})

var extract = tar.extract('./my-other-directory', {
  ignore: function(name) {
    return path.extname(name) === '.bin' // ignore .bin files inside the tarball when extracing
  }
})

var extractFilesDirs = tar.extract('./my-other-other-directory', {
  ignore: function(_, header) {
    // pass files & directories, ignore e.g. symlinks
    return header.type !== 'file' && header.type !== 'directory'
  }
})

You can also specify which entries to pack using the entries option

var pack = tar.pack('./my-directory', {
  entries: ['file1', 'subdir/file2'] // only the specific entries will be packed
})

If you want to modify the headers when packing/extracting add a map function to the options

var pack = tar.pack('./my-directory', {
  map: function(header) {
    header.name = 'prefixed/'+header.name
    return header
  }
})

var extract = tar.extract('./my-directory', {
  map: function(header) {
    header.name = 'another-prefix/'+header.name
    return header
  }
})

Similarly you can use mapStream incase you wanna modify the input/output file streams

var pack = tar.pack('./my-directory', {
  mapStream: function(fileStream, header) {
    if (path.extname(header.name) === '.js') {
      return fileStream.pipe(someTransform)
    }
    return fileStream;
  }
})

var extract = tar.extract('./my-directory', {
  mapStream: function(fileStream, header) {
    if (path.extname(header.name) === '.js') {
      return fileStream.pipe(someTransform)
    }
    return fileStream;
  }
})

Set options.fmode and options.dmode to ensure that files/directories extracted have the corresponding modes

var extract = tar.extract('./my-directory', {
  dmode: parseInt(555, 8), // all dirs should be readable
  fmode: parseInt(444, 8) // all files should be readable
})

It can be useful to use dmode and fmode if you are packing/unpacking tarballs between *nix/windows to ensure that all files/directories unpacked are readable.

Alternatively you can set options.readable and/or options.writable to set the dmode and fmode to readable/writable.

var extract = tar.extract('./my-directory', {
  readable: true, // all dirs and files should be readable
  writable: true, // all dirs and files should be writable
})

Set options.strict to false if you want to ignore errors due to unsupported entry types (like device files)

To dereference symlinks (pack the contents of the symlink instead of the link itself) set options.dereference to true.

Copy a directory

Copying a directory with permissions and mtime intact is as simple as

tar.pack('source-directory').pipe(tar.extract('dest-directory'))

Interaction with `tar-stream`

Use finalize: false and the finish hook to leave the pack stream open for further entries (see tar-stream#pack), and use pack to pass an existing pack stream.

var mypack = tar.pack('./my-directory', {
  finalize: false,
  finish: function(sameAsMypack) {
    mypack.entry({name: 'generated-file.txt'}, "hello")
    tar.pack('./other-directory', {
      pack: sameAsMypack
    })
  }
})

Performance

Packing and extracting a 6.1 GB with 2496 directories and 2398 files yields the following results on my Macbook Air. See the benchmark here

tar-fs: 34.261 ms
node-tar: 366.123 ms (or 10x slower)

License

MIT