// Make sure working directory exists and is empty
val wd = os.pwd/"out/splash"
os.remove.all(wd)
os.makeDir.all(wd)
// Read/write files
os.write(wd/"file.txt", "hello")
os.read(wd/"file.txt") ==> "hello"
// Perform filesystem operations
os.copy(wd/"file.txt", wd/"copied.txt")
os.list(wd) ==> Seq(wd/"copied.txt", wd/"file.txt")
// Invoke subprocesses
val invoked = os.proc("cat", wd/"file.txt", wd/"copied.txt").call(cwd = wd)
invoked.out.trim ==> "hellohello"
// Chain multiple subprocess' stdin/stdout together
val curl = os.proc("curl", "-L" , "https://git.io/fpvpS").spawn(stderr = os.Inherit)
val gzip = os.proc("gzip", "-n").spawn(stdin = curl.stdout)
val sha = os.proc("shasum", "-a", "256").spawn(stdin = gzip.stdout)
sha.stdout.trim ==> "acc142175fa520a1cb2be5b97cbbe9bea092e8bba3fe2e95afa645615908229e -"OS-Lib is a simple Scala interface to common OS filesystem and subprocess APIs. OS-Lib aims to make working with files and processes in Scala as simple as any scripting language, while still providing the safety, flexibility and performance you would expect from Scala.
OS-Lib aims to be a complete replacement for the
java.nio.file.Files/java.nio.file.Paths, java.lang.ProcessBuilder
scala.io and scala.sys APIs. You should not need to drop down to underlying
Java APIs, as OS-Lib exposes all relevant capabilities in an intuitive and
performant way. OS-Lib has no dependencies and is unopinionated: it exposes the
underlying APIs in a concise but straightforward way, without introducing it’s
own idiosyncrasies, quirks, or clever DSLs.
If you use OS-Lib and like it, you will probably enjoy the following book by the Author:
Hands-on Scala has uses OS-Lib extensively throughout the book, and has the entirety of Chapter 7: Files and Subprocesses dedicated to OS-Lib. Hands-on Scala is a great way to level up your skills in Scala in general and OS-Lib in particular.
You can also support it by donating to our Patreon:
For a hands-on introduction to the library, take a look at these two blog posts:
To begin using OS-Lib, first add it as a dependency to your project’s build:
// Mill
ivy"com.lihaoyi::os-lib:0.11.3"
// SBT
"com.lihaoyi" %% "os-lib" % "0.11.3"Most operation in OS-Lib take place on os.Paths, which are
constructed from a base path or working directory wd. Most often, the first
thing to do is to define a wd path representing the folder you want to work
with:
val wd = os.pwd / "my-test-folder"You can of course multiple base paths, to use in different parts of your program
where convenient, or simply work with one of the pre-defined paths os.pwd,
os.root, or os.home.
// Find and concatenate all .txt files directly in the working directory
os.write(
wd / "all.txt",
os.list(wd).filter(_.ext == "txt").map(os.read)
)
os.read(wd / "all.txt") ==>
"""I am cowI am cow
|Hear me moo
|I weigh twice as much as you
|And I look good on the barbecue""".stripMargin// Find and concatenate all .txt files directly in the working directory using `cat`
os.proc("cat", os.list(wd).filter(_.ext == "txt")).call(stdout = wd / "all.txt")
os.read(wd / "all.txt") ==>
"""I am cowI am cow
|Hear me moo
|I weigh twice as much as you
|And I look good on the barbecue""".stripMargin// Curl to temporary file
val temp = os.temp()
os.proc("curl", "-L" , "https://git.io/fpfTs").call(stdout = temp)
os.size(temp) ==> 53814
// Curl to temporary file
val temp2 = os.temp()
val proc = os.proc("curl", "-L" , "https://git.io/fpfTJ").spawn()
os.write.over(temp2, proc.stdout)
os.size(temp2) ==> 53814// Line-count of all .txt files recursively in wd
val lineCount = os.walk(wd)
.filter(_.ext == "txt")
.map(os.read.lines)
.map(_.size)
.sum
lineCount ==> 9// Find the largest three files in the given folder tree
val largestThree = os.walk(wd)
.filter(os.isFile(_, followLinks = false))
.map(x => os.size(x) -> x).sortBy(-_._1)
.take(3)
largestThree ==> Seq(
(711, wd / "misc/binary.png"),
(81, wd / "Multi Line.txt"),
(22, wd / "folder1/one.txt")
)// Move all files inside the "misc" folder out of it
import os./
os.list(wd / "misc").map(os.move.matching { case p/"misc"/x => p/x } )// Calculate the word frequency of all the text files in the folder tree
def txt = os.walk(wd).filter(_.ext == "txt").map(os.read)
def freq(s: Seq[String]) = s.groupBy(x => x).mapValues(_.length).toSeq
val map = freq(txt.flatMap(_.split("[^a-zA-Z0-9_]"))).sortBy(-_._2)
mapos.read(arg: os.ReadablePath): String
os.read(arg: os.ReadablePath, charSet: Codec): String
os.read(arg: os.Path,
offset: Long = 0,
count: Int = Int.MaxValue,
charSet: Codec = java.nio.charset.StandardCharsets.UTF_8): StringReads the contents of a os.Path or other os.Source as a
java.lang.String. Defaults to reading the entire file as UTF-8, but you can
also select a different charSet to use, and provide an offset/count to
read from if the source supports seeking.
os.read(wd / "File.txt") ==> "I am cow"
os.read(wd / "folder1/one.txt") ==> "Contents of folder one"
os.read(wd / "Multi Line.txt") ==>
"""I am cow
|Hear me moo
|I weigh twice as much as you
|And I look good on the barbecue""".stripMarginos.read.bytes(arg: os.ReadablePath): Array[Byte]
os.read.bytes(arg: os.Path, offset: Long, count: Int): Array[Byte]Reads the contents of a os.Path or os.Source as an
Array[Byte]; you can provide an offset/count to read from if the source
supports seeking.
os.read.bytes(wd / "File.txt") ==> "I am cow".getBytes
os.read.bytes(wd / "misc/binary.png").length ==> 711os.read.chunks(p: ReadablePath, chunkSize: Int): os.Generator[(Array[Byte], Int)]
os.read.chunks(p: ReadablePath, buffer: Array[Byte]): os.Generator[(Array[Byte], Int)]Reads the contents of the given path in chunks of the given size; returns a generator which provides a byte array and an offset into that array which contains the data for that chunk. All chunks will be of the given size, except for the last chunk which may be smaller.
Note that the array returned by the generator is shared between each callback; make sure you copy the bytes/array somewhere else if you want to keep them around.
Optionally takes in a provided input buffer instead of a chunkSize,
allowing you to re-use the buffer between invocations.
val chunks = os.read.chunks(wd / "File.txt", chunkSize = 2)
.map{case (buf, n) => buf.take(n).toSeq } // copy the buffer to save the data
.toSeq
chunks ==> Seq(
Seq[Byte]('I', ' '),
Seq[Byte]('a', 'm'),
Seq[Byte](' ', 'c'),
Seq[Byte]('o', 'w')
)os.read.lines(arg: os.ReadablePath): IndexedSeq[String]
os.read.lines(arg: os.ReadablePath, charSet: Codec): IndexedSeq[String]Reads the given os.Path or other os.Source as a string
and splits it into lines; defaults to reading as UTF-8, which you can override
by specifying a charSet.
os.read.lines(wd / "File.txt") ==> Seq("I am cow")
os.read.lines(wd / "Multi Line.txt") ==> Seq(
"I am cow",
"Hear me moo",
"I weigh twice as much as you",
"And I look good on the barbecue"
)os.read.lines(arg: os.ReadablePath): os.Generator[String]
os.read.lines(arg: os.ReadablePath, charSet: Codec): os.Generator[String]Identical to os.read.lines, but streams the results back to you
in a os.Generator rather than accumulating them in memory.
Useful if the file is large.
os.read.lines.stream(wd / "File.txt").count() ==> 1
os.read.lines.stream(wd / "Multi Line.txt").count() ==> 4
// Streaming the lines to the console
for(line <- os.read.lines.stream(wd / "Multi Line.txt")){
println(line)
}os.read.inputStream(p: ReadablePath): java.io.InputStreamOpens a java.io.InputStream to read from the given file.
val is = os.read.inputStream(wd / "File.txt") // ==> "I am cow"
is.read() ==> 'I'
is.read() ==> ' '
is.read() ==> 'a'
is.read() ==> 'm'
is.read() ==> ' '
is.read() ==> 'c'
is.read() ==> 'o'
is.read() ==> 'w'
is.read() ==> -1
is.close()os.read.stream(p: ReadablePath): geny.ReadableOpens a geny.Readable to read from
the given file. This allows you to stream data to any other library that
supports Readable without buffering the data in memory, e.g. parsing it via
FastParse, deserializing it via uPickle, uploading it via Requests-Scala, etc.
val readable: geny.Readable = os.read.stream(wd / "File.json")
requests.post("https://httpbin.org/post", data = readable)
upickle.default.read(readable)
ujson.read(readable)os.write(target: Path,
data: os.Source,
perms: PermSet = null,
createFolders: Boolean = false): UnitWrites data from the given file or os.Source to a file at the
target os.Path. You can specify the filesystem permissions of the
newly created file by passing in a os.PermSet.
This throws an exception if the file already exists. To over-write or append to
an existing file, see os.write.over or
os.write.append.
By default, this doesn’t create enclosing folders; you can enable this
behavior by setting createFolders = true
os.write(wd / "New File.txt", "New File Contents")
os.read(wd / "New File.txt") ==> "New File Contents"
os.write(wd / "NewBinary.bin", Array[Byte](0, 1, 2, 3))
os.read.bytes(wd / "NewBinary.bin") ==> Array[Byte](0, 1, 2, 3)os.write.append(target: Path,
data: os.Source,
perms: PermSet = null,
createFolders: Boolean = false): UnitSimilar to os.write, except if the file already exists this appends
the written data to the existing file contents.
os.read(wd / "File.txt") ==> "I am cow"
os.write.append(wd / "File.txt", ", hear me moo")
os.read(wd / "File.txt") ==> "I am cow, hear me moo"
os.write.append(wd / "File.txt", ",\nI weigh twice as much as you")
os.read(wd / "File.txt") ==>
"I am cow, hear me moo,\nI weigh twice as much as you"
os.read.bytes(wd / "misc/binary.png").length ==> 711
os.write.append(wd / "misc/binary.png", Array[Byte](1, 2, 3))
os.read.bytes(wd / "misc/binary.png").length ==> 714os.write.over(target: Path,
data: os.Source,
perms: PermSet = null,
offset: Long = 0,
createFolders: Boolean = false,
truncate: Boolean = true): UnitSimilar to os.write, except if the file already exists this
over-writes the existing file contents. You can also pass in truncate = false
to avoid truncating the file if the new contents is shorter than the old
contents, and an offset to the file you want to write to.
os.read(wd / "File.txt") ==> "I am cow"
os.write.over(wd / "File.txt", "You are cow")
os.read(wd / "File.txt") ==> "You are cow"
os.write.over(wd / "File.txt", "We ", truncate = false)
os.read(wd / "File.txt") ==> "We are cow"
os.write.over(wd / "File.txt", "s", offset = 8, truncate = false)
os.read(wd / "File.txt") ==> "We are sow"os.write.outputStream(target: Path,
perms: PermSet = null,
createFolders: Boolean = false,
openOptions: Seq[OpenOption] = Seq(CREATE, WRITE))Open a java.io.OutputStream to write to the given file.
val out = os.write.outputStream(wd / "New File.txt")
out.write('H')
out.write('e')
out.write('l')
out.write('l')
out.write('o')
out.close()
os.read(wd / "New File.txt") ==> "Hello"os.list(p: Path): IndexedSeq[Path]
os.list(p: Path, sort: Boolean = true): IndexedSeq[Path]Returns all the files and folders directly within the given folder. If the given
path is not a folder, raises an error. Can be called via
os.list.stream to stream the results. To list files recursively,
use os.walk.
For convenience os.list sorts the entries in the folder before returning
them. You can disable sorted by passing in the flag sort = false.
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.list(wd / "folder2") ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedB"
)os.list.stream(p: Path): os.Generator[Path]Similar to os.list, except provides a os.Generator of
results rather than accumulating all of them in memory. Useful if the result set
is large.
os.list.stream(wd / "folder2").count() ==> 2
// Streaming the listed files to the console
for(line <- os.list.stream(wd / "folder2")){
println(line)
}os.walk(path: Path,
skip: Path => Boolean = _ => false,
preOrder: Boolean = true,
followLinks: Boolean = false,
maxDepth: Int = Int.MaxValue,
includeTarget: Boolean = false): IndexedSeq[Path]Recursively walks the given folder and returns the paths of every file or folder within.
You can pass in a skip callback to skip files or folders you are not
interested in. This can avoid walking entire parts of the folder hierarchy,
saving time as compared to filtering them after the fact.
By default, the paths are returned as a pre-order traversal: the enclosing
folder is occurs first before any of it’s contents. You can pass in preOrder =
false to turn it into a post-order traversal, such that the enclosing folder
occurs last after all it’s contents.
os.walk returns but does not follow symlinks; pass in followLinks = true to
override that behavior. You can also specify a maximum depth you wish to walk
via the maxDepth parameter.
os.walk does not include the path given to it as part of the traversal by
default. Pass in includeTarget = true to make it do so. The path appears at
the start of the traversal of preOrder = true, and at the end of the traversal
if preOrder = false.
os.walk(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.walk(wd / "folder1", includeTarget = true) ==> Seq(
wd / "folder1",
wd / "folder1/one.txt"
)
os.walk(wd / "folder2") ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedA/a.txt",
wd / "folder2/nestedB",
wd / "folder2/nestedB/b.txt"
)
os.walk(wd / "folder2", preOrder = false) ==> Seq(
wd / "folder2/nestedA/a.txt",
wd / "folder2/nestedA",
wd / "folder2/nestedB/b.txt",
wd / "folder2/nestedB"
)
os.walk(wd / "folder2", maxDepth = 1) ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedB"
)
os.walk(wd / "folder2", skip = _.last == "nestedA") ==> Seq(
wd / "folder2/nestedB",
wd / "folder2/nestedB/b.txt"
)os.walk.attrs(path: Path,
skip: (Path, os.StatInfo) => Boolean = (_, _) => false,
preOrder: Boolean = true,
followLinks: Boolean = false,
maxDepth: Int = Int.MaxValue,
includeTarget: Boolean = false): IndexedSeq[(Path, os.StatInfo)]Similar to os.walk, except it also provides the os.StatInfo
filesystem metadata of every path that it returns. Can save time by allowing you
to avoid querying the filesystem for metadata later. Note that os.StatInfo
does not include filesystem ownership and permissions data; use os.stat.posix on
the path if you need those attributes.
val filesSortedBySize = os.walk.attrs(wd / "misc", followLinks = true)
.sortBy{case (p, attrs) => attrs.size}
.collect{case (p, attrs) if attrsisFile => p}
filesSortedBySize ==> Seq(
wd / "misc/echo",
wd / "misc/file-symlink",
wd / "misc/echo_with_wd",
wd / "misc/folder-symlink/one.txt",
wd / "misc/binary.png"
)os.walk.stream(path: Path,
skip: Path => Boolean = _ => false,
preOrder: Boolean = true,
followLinks: Boolean = false,
maxDepth: Int = Int.MaxValue,
includeTarget: Boolean = false): os.Generator[Path]Similar to os.walk, except returns a os.Generator of
the results rather than accumulating them in memory. Useful if you are walking
very large folder hierarchies, or if you wish to begin processing the output
even before the walk has completed.
os.walk.stream(wd / "folder1").count() ==> 1
os.walk.stream(wd / "folder2").count() ==> 4
os.walk.stream(wd / "folder2", skip = _.last == "nestedA").count() ==> 2os.walk.stream.attrs(path: Path,
skip: (Path, os.StatInfo) => Boolean = (_, _) => false,
preOrder: Boolean = true,
followLinks: Boolean = false,
maxDepth: Int = Int.MaxValue,
includeTarget: Boolean = false): os.Generator[(Path, os.StatInfo)]Similar to os.walk.stream, except it also provides the filesystem
metadata of every path that it returns. Can save time by allowing you to avoid
querying the filesystem for metadata later.
def totalFileSizes(p: os.Path) = os.walk.stream.attrs(p)
.collect{case (p, attrs) if attrs.isFile => attrs.size}
.sum
totalFileSizes(wd / "folder1") ==> 22
totalFileSizes(wd / "folder2") ==> 40os.exists(p: Path, followLinks: Boolean = true): BooleanChecks if a file or folder exists at the specified path
os.exists(wd / "File.txt") ==> true
os.exists(wd / "folder1") ==> true
os.exists(wd / "doesnt-exist") ==> false
os.exists(wd / "misc/file-symlink") ==> true
os.exists(wd / "misc/folder-symlink") ==> true
os.exists(wd / "misc/broken-symlink") ==> false
os.exists(wd / "misc/broken-symlink", followLinks = false) ==> trueos.move(from: Path, to: Path): Unit
os.move(from: Path, to: Path, createFolders: Boolean): UnitMoves a file or folder from one path to another. Errors out if the destination path already exists, or is within the source path.
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.move(wd / "folder1/one.txt", wd / "folder1/first.txt")
os.list(wd / "folder1") ==> Seq(wd / "folder1/first.txt")
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedA", wd / "folder2/nestedB")
os.move(wd / "folder2/nestedA", wd / "folder2/nestedC")
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedB", wd / "folder2/nestedC")
os.read(wd / "File.txt") ==> "I am cow"
os.move(wd / "Multi Line.txt", wd / "File.txt", replaceExisting = true)
os.read(wd / "File.txt") ==>
"""I am cow
|Hear me moo
|I weigh twice as much as you
|And I look good on the barbecue""".stripMarginos.move.matching(t: PartialFunction[Path, Path]): PartialFunction[Path, Unit]os.move can also be used as a transformer, via os.move.matching. This lets
you use .map or .collect on a list of paths, and move all of them at once,
e.g. to rename all .txt files within a folder tree to .data:
import os.{GlobSyntax, /}
os.walk(wd / "folder2") ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedA/a.txt",
wd / "folder2/nestedB",
wd / "folder2/nestedB/b.txt"
)
os.walk(wd/'folder2).collect(os.move.matching{case p/g"$x.txt" => p/g"$x.data"})
os.walk(wd / "folder2") ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedA/a.data",
wd / "folder2/nestedB",
wd / "folder2/nestedB/b.data"
)os.move.into(from: Path, to: Path): UnitMove the given file or folder into the destination folder
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.move.into(wd / "File.txt", wd / "folder1")
os.list(wd / "folder1") ==> Seq(wd / "folder1/File.txt", wd / "folder1/one.txt")os.move.over(from: Path, to: Path): UnitMove a file or folder from one path to another, and overwrite any file or folder than may already be present at that path
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedA", wd / "folder2/nestedB")
os.move.over(wd / "folder1", wd / "folder2")
os.list(wd / "folder2") ==> Seq(wd / "folder2/one.txt")os.copy(from: Path, to: Path): Unit
os.copy(from: Path, to: Path, createFolders: Boolean): UnitCopy a file or folder from one path to another. Recursively copies folders with all their contents. Errors out if the destination path already exists, or is within the source path.
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.copy(wd / "folder1/one.txt", wd / "folder1/first.txt")
os.list(wd / "folder1") ==> Seq(wd / "folder1/first.txt", wd / "folder1/one.txt")
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedA", wd / "folder2/nestedB")
os.copy(wd / "folder2/nestedA", wd / "folder2/nestedC")
os.list(wd / "folder2") ==> Seq(
wd / "folder2/nestedA",
wd / "folder2/nestedB",
wd / "folder2/nestedC"
)
os.read(wd / "File.txt") ==> "I am cow"
os.copy(wd / "Multi Line.txt", wd / "File.txt", replaceExisting = true)
os.read(wd / "File.txt") ==>
"""I am cow
|Hear me moo
|I weigh twice as much as you
|And I look good on the barbecue""".stripMargin
```
`os.copy` can also be used as a transformer:
```scala
os.copy.matching(t: PartialFunction[Path, Path]): PartialFunction[Path, Unit]This lets you use .map or .collect on a list of paths, and copy all of them
at once:
paths.map(os.copy.matching{case p/"scala"/file => p/"java"/file})os.copy.into(from: Path, to: Path): UnitCopy the given file or folder into the destination folder
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.copy.into(wd / "File.txt", wd / "folder1")
os.list(wd / "folder1") ==> Seq(wd / "folder1/File.txt", wd / "folder1/one.txt")os.copy.over(from: Path, to: Path): UnitSimilar to os.copy, but if the destination file already exists then
overwrite it instead of erroring out.
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedA", wd / "folder2/nestedB")
os.copy.over(wd / "folder1", wd / "folder2")
os.list(wd / "folder2") ==> Seq(wd / "folder2/one.txt")Since 0.7.5
If you want to copy a directory over another but don’t want to overwrite the whole destination directory (and loose it’s content),
you can use the mergeFolders option of os.copy.
os.list(wd / "folder1") ==> Seq(wd / "folder1/one.txt")
os.list(wd / "folder2") ==> Seq(wd / "folder2/nestedA", wd / "folder2/nestedB")
os.copy(wd / "folder1", wd / "folder2", mergeFolders = true)
os.list(wd / "folder2") ==> Seq(wd / "folder2/one.txt", wd / "folder2/nestedA", wd / "folder2/nestedB")os.makeDir(path: Path): Unit
os.makeDir(path: Path, perms: PermSet): UnitCreate a single directory at the specified path. Optionally takes in a
os.PermSet to specify the filesystem permissions of the created
directory.
Errors out if the directory already exists, or if the parent directory of the
specified path does not exist. To automatically create enclosing directories and
ignore the destination if it already exists, using
os.makeDir.all
os.exists(wd / "new_folder") ==> false
os.makeDir(wd / "new_folder")
os.exists(wd / "new_folder") ==> trueos.makeDir.all(path: Path): Unit
os.makeDir.all(path: Path,
perms: PermSet = null,
acceptLinkedDirectory: Boolean = true): UnitSimilar to os.makeDir, but automatically creates any necessary
enclosing directories if they do not exist, and does not raise an error if the
destination path already contains a directory. Also does not raise an error if
the destination path contains a symlink to a directory, though you can force it
to error out in that case by passing in acceptLinkedDirectory = false
os.exists(wd / "new_folder") ==> false
os.makeDir.all(wd / "new_folder/inner/deep")
os.exists(wd / "new_folder/inner/deep") ==> trueos.remove(target: Path): Boolean
os.remove(target: Path, checkExists: Boolean = false): BooleanRemove the target file or folder. Folders need to be empty to be removed; if you
want to remove a folder tree recursively, use os.remove.all.
Returns true if the file was present before.
It will fail with an exception when the file is missing but checkExists is true,
or when the directory to remove is not empty.
os.exists(wd / "File.txt") ==> true
os.remove(wd / "File.txt")
os.exists(wd / "File.txt") ==> false
os.exists(wd / "folder1/one.txt") ==> true
os.remove(wd / "folder1/one.txt")
os.remove(wd / "folder1")
os.exists(wd / "folder1/one.txt") ==> false
os.exists(wd / "folder1") ==> falseWhen removing symbolic links, it is the link that gets removed, and not its destination:
os.remove(wd / "misc/file-symlink")
os.exists(wd / "misc/file-symlink", followLinks = false) ==> false
os.exists(wd / "File.txt", followLinks = false) ==> true
os.remove(wd / "misc/folder-symlink")
os.exists(wd / "misc/folder-symlink", followLinks = false) ==> false
os.exists(wd / "folder1", followLinks = false) ==> true
os.exists(wd / "folder1/one.txt", followLinks = false) ==> true
os.remove(wd / "misc/broken-symlink")
os.exists(wd / "misc/broken-symlink", followLinks = false) ==> falseIf you wish to remove the destination of a symlink, use
os.readLink.
os.remove.all(target: Path): UnitRemove the target file or folder; if it is a folder and not empty, recursively removing all it’s contents before deleting it.
os.exists(wd / "folder1/one.txt") ==> true
os.remove.all(wd / "folder1")
os.exists(wd / "folder1/one.txt") ==> false
os.exists(wd / "folder1") ==> falseWhen removing symbolic links, it is the links that gets removed, and not it’s destination:
os.remove.all(wd / "misc/file-symlink")
os.exists(wd / "misc/file-symlink", followLinks = false) ==> false
os.exists(wd / "File.txt", followLinks = false) ==> true
os.remove.all(wd / "misc/folder-symlink")
os.exists(wd / "misc/folder-symlink", followLinks = false) ==> false
os.exists(wd / "folder1", followLinks = false) ==> true
os.exists(wd / "folder1/one.txt", followLinks = false) ==> true
os.remove.all(wd / "misc/broken-symlink")
os.exists(wd / "misc/broken-symlink", followLinks = false) ==> falseIf you wish to remove the destination of a symlink, use
os.readLink.
os.hardlink(src: Path, dest: Path, perms): UnitCreate a hardlink to the source path from the destination path
os.hardlink(wd / "File.txt", wd / "Linked.txt")
os.exists(wd / "Linked.txt")
os.read(wd / "Linked.txt") ==> "I am cow"
os.isLink(wd / "Linked.txt") ==> falseos.symlink(link: Path, dest: FilePath, perms: PermSet = null): UnitCreate a symbolic to the source path from the destination path. Optionally takes
a os.PermSet to customize the filesystem permissions of the symbolic
link.
os.symlink(wd / "File.txt", wd / "Linked.txt")
os.exists(wd / "Linked.txt")
os.read(wd / "Linked.txt") ==> "I am cow"
os.isLink(wd / "Linked.txt") ==> trueYou can create symlinks with either absolute os.Paths or relative os.RelPaths:
os.symlink(wd / "File.txt", os.rel/ "Linked2.txt")
os.exists(wd / "Linked2.txt")
os.read(wd / "Linked2.txt") ==> "I am cow"
os.isLink(wd / "Linked2.txt") ==> trueCreating absolute and relative symlinks respectively. Relative symlinks are resolved relative to the enclosing folder of the link.
os.readLink(src: Path): os.FilePath
os.readLink.absolute(src: Path): os.PathReturns the immediate destination of the given symbolic link.
os.readLink(wd / "misc/file-symlink") ==> os.up / "File.txt"
os.readLink(wd / "misc/folder-symlink") ==> os.up / "folder1"
os.readLink(wd / "misc/broken-symlink") ==> os.rel / "broken"
os.readLink(wd / "misc/broken-abs-symlink") ==> os.root / "doesnt/exist"Note that symbolic links can be either absolute os.Paths or relative
os.RelPaths, represented by os.FilePath. You can also use os.readLink.absolute
to automatically resolve relative symbolic links to their absolute destination:
os.readLink.absolute(wd / "misc/file-symlink") ==> wd / "File.txt"
os.readLink.absolute(wd / "misc/folder-symlink") ==> wd / "folder1"
os.readLink.absolute(wd / "misc/broken-symlink") ==> wd / "misc/broken"
os.readLink.absolute(wd / "misc/broken-abs-symlink") ==> os.root / "doesnt/exist"os.followLink(src: Path): Option[Path]Attempts to any deference symbolic links in the given path, recursively, and return the
canonical path. Returns None if the path cannot be resolved (i.e. some
symbolic link in the given path is broken)
os.followLink(wd / "misc/file-symlink") ==> Some(wd / "File.txt")
os.followLink(wd / "misc/folder-symlink") ==> Some(wd / "folder1")
os.followLink(wd / "misc/broken-symlink") ==> Noneos.temp(contents: os.Source = null,
dir: Path = null,
prefix: String = null,
suffix: String = null,
deleteOnExit: Boolean = true,
perms: PermSet = null): PathCreates a temporary file. You can optionally provide a dir to specify where
this file lives, file-prefix and file-suffix to customize what it looks
like, and a os.PermSet to customize its filesystem permissions.
Passing in a os.Source will initialize the contents of that file to
the provided data; otherwise it is created empty.
By default, temporary files are deleted on JVM exit. You can disable that
behavior by setting deleteOnExit = false
val tempOne = os.temp("default content")
os.read(tempOne) ==> "default content"
os.write.over(tempOne, "Hello")
os.read(tempOne) ==> "Hello"os.temp.dir(dir: Path = null,
prefix: String = null,
deleteOnExit: Boolean = true,
perms: PermSet = null): PathCreates a temporary directory. You can optionally provide a dir to specify
where this file lives, a prefix to customize what it looks like, and a
os.PermSet to customize its filesystem permissions.
By default, temporary directories are deleted on JVM exit. You can disable that
behavior by setting deleteOnExit = false
val tempDir = os.temp.dir()
os.list(tempDir) ==> Nil
os.write(tempDir / "file", "Hello")
os.list(tempDir) ==> Seq(tempDir / "file")def apply(dest: os.Path,
sources: Seq[ZipSource] = List(),
excludePatterns: Seq[Regex] = List(),
includePatterns: Seq[Regex] = List(),
preserveMtimes: Boolean = false,
deletePatterns: Seq[Regex] = List(),
compressionLevel: Int = -1 /* 0-9 */): os.PathThe zip object provides functionality to create or modify zip archives. It supports:
-
Zipping Files and Directories: You can zip both individual files and entire directories.
-
Appending to Existing Archives: Files can be appended to an existing zip archive.
-
Exclude Patterns (-x): You can specify files or patterns to exclude while zipping.
-
Include Patterns (-i): You can include specific files or patterns while zipping.
-
Delete Patterns (-d): You can delete specific files from an existing zip archive.
-
Configuring whether or not to preserve filesyste mtimes and permissions
This will create a new zip archive at dest containing file1.txt and everything
inside sources. If dest already exists as a zip, the files will be appended to the
existing zip, and any existing zip entries matching deletePatterns will be removed.
Note that os.zip doesn’t support creating/unpacking symlinks or filesystem permissions
in Zip files, because the underlying java.util.zip.Zip*Stream doesn’t support them.
The example below demonstrates the core workflows: creating a zip, appending to it, and unzipping it:
// Zipping files and folders in a new zip file
val zipFileName = "zip-file-test.zip"
val zipFile1: os.Path = os.zip(
destination = wd / zipFileName,
sourcePaths = Seq(
wd / "File.txt",
wd / "folder1"
)
)
// Adding files and folders to an existing zip file
os.zip(
destination = zipFile1,
sourcePaths = Seq(
wd / "folder2",
wd / "Multi Line.txt"
)
)
// Unzip file to a destination folder
val unzippedFolder = os.unzip(
source = wd / zipFileName,
destination = wd / "unzipped folder"
)
val paths = os.walk(unzippedFolder)
val expected = Seq(
// Files get included in the zip root using their name
wd / "unzipped folder/File.txt",
wd / "unzipped folder/Multi Line.txt",
// Folder contents get included relative to the source root
wd / "unzipped folder/nestedA",
wd / "unzipped folder/nestedB",
wd / "unzipped folder/one.txt",
wd / "unzipped folder/nestedA/a.txt",
wd / "unzipped folder/nestedB/b.txt",
)
assert(paths.sorted == expected)You can also pass in a mapping to os.zip to specify exactly where in the zip each
input source file or folder should go:
val zipFileName = "zip-file-test.zip"
val zipFile1: os.Path = os.zip(
destination = wd / zipFileName,
sourcePaths = List(
// renaming files and folders
wd / "File.txt" -> os.sub / "renamed-file.txt",
wd / "folder1" -> os.sub / "renamed-folder"
)
)
val unzippedFolder = os.unzip(
source = zipFile1,
destination = wd / "unzipped folder"
)
val paths = os.walk(unzippedFolder)
val expected = Seq(
wd / "unzipped folder/renamed-file.txt",
wd / "unzipped folder/renamed-folder",
wd / "unzipped folder/renamed-folder/one.txt",
)
assert(paths.sorted == expected)You can specify files or folders to be excluded or included when creating the zip:
os.zip(
os.Path("/path/to/destination.zip"),
List(os.Path("/path/to/folder")),
excludePatterns = List(".*\\.log".r, "temp/.*".r), // Exclude log files and "temp" folder
includePatterns = List(".*\\.txt".r) // Include only .txt files
)This will include only .txt files, excluding any .log files and anything inside
the temp folder.
You can use os.zip.stream to write the final zip to an OutputStream rather than a
concrete os.Path. os.zip.stream returns a geny.Writable, which has a writeBytesToStream
method:
val zipFileName = "zipStreamFunction.zip"
val stream = os.write.outputStream(wd / "zipStreamFunction.zip")
val writable = zip.stream(sources = Seq(wd / "File.txt"))
writable.writeBytesTo(stream)
stream.close()
val unzippedFolder = os.unzip(
source = wd / zipFileName,
dest = wd / "zipStreamFunction"
)
val paths = os.walk(unzippedFolder)
assert(paths == Seq(unzippedFolder / "File.txt"))This can be useful for streaming the zipped data to places which are not files: over the network, over a pipe, etc.
os.unzip(os.Path("/path/to/archive.zip"), Some(os.Path("/path/to/destination")))This extracts the contents of archive.zip to the specified destination.
You can exclude certain files from being extracted using patterns:
os.unzip(
os.Path("/path/to/archive.zip"),
Some(os.Path("/path/to/destination")),
excludePatterns = List(".*\\.log".r, "temp/.*".r) // Exclude log files and the "temp" folder
)You can unzip a zip file from any arbitrary java.io.InputStream containing its binary data
using the os.unzip.stream method:
val readableZipStream: java.io.InputStream = ???
// Unzipping the stream to the destination folder
os.unzip.stream(
source = readableZipStream,
dest = unzippedFolder
)This can be useful if the zip file does not exist on disk, e.g. if it is received over the network or produced in-memory by application logic.
OS-Lib also provides the os.unzip.streamRaw API, which is a lower level API used internally
within os.unzip.stream but can also be used directly if lower-level control is necessary.
os.zip.open(path: Path): ZipRootos.zip.open allows you to treat zip files as filesystems, using normal os.* operations
on them. This provides a move flexible way to manipulate the contents of the zip in a fine-grained
manner when the normal os.zip or os.unzip operations do not suffice.
val zipFile = os.zip.open(wd / "zip-test.zip")
try {
os.copy(wd / "File.txt", zipFile / "File.txt")
os.copy(wd / "folder1", zipFile / "folder1")
os.copy(wd / "folder2", zipFile / "folder2")
}finally zipFile.close()
val zipFile2 = os.zip.open(wd / "zip-test.zip")
try{
os.list(zipFile2) ==> Vector(zipFile2 / "File.txt", zipFile2 / "folder1", zipFile2 / "folder2")
os.remove.all(zipFile2 / "folder2")
os.remove(zipFile2 / "File.txt")
}finally zipFile2.close()
val zipFile3 = os.zip.open(wd / "zip-test.zip")
try os.list(zipFile3) ==> Vector(zipFile3 / "folder1")
finally zipFile3.close()os.zip.open returns a ZipRoot, which is identical to os.Path except it references the root
of the zip file rather than a bare path on the filesystem. Note that you need to call ZipRoot#close()
when you are done with it to avoid leaking filesystem resources.
os.stat(p: os.Path, followLinks: Boolean = true): os.StatInfoReads in the basic filesystem metadata for the given file. By default, follows
symbolic links to read the metadata of whatever the link is pointing at; set
followLinks = false to disable that and instead read the metadata of the
symbolic link itself.
os.stat(wd / "File.txt").size ==> 8
os.stat(wd / "Multi Line.txt").size ==> 81
os.stat(wd / "folder1").fileType ==> os.FileType.Diros.stat.posix(p: os.Path, followLinks: Boolean = true): os.PosixStatInfoReads in the posix filesystem metadata for the given file, providing
information on permissions and ownership. By default, follows symbolic
links to read the metadata of whatever the link is pointing at; set
followLinks = false to disable that and instead read the metadata of
the symbolic link itself.
os.isFile(p: Path, followLinks: Boolean = true): BooleanReturns true if the given path is a file. Follows symbolic links by default,
pass in followLinks = false to not do so.
os.isFile(wd / "File.txt") ==> true
os.isFile(wd / "folder1") ==> false
os.isFile(wd / "misc/file-symlink") ==> true
os.isFile(wd / "misc/folder-symlink") ==> false
os.isFile(wd / "misc/file-symlink", followLinks = false) ==> falseos.isDir(p: Path, followLinks: Boolean = true): BooleanReturns true if the given path is a folder. Follows symbolic links by default,
pass in followLinks = false to not do so.
os.isDir(wd / "File.txt") ==> false
os.isDir(wd / "folder1") ==> true
os.isDir(wd / "misc/file-symlink") ==> false
os.isDir(wd / "misc/folder-symlink") ==> true
os.isDir(wd / "misc/folder-symlink", followLinks = false) ==> falseos.isLink(p: Path, followLinks: Boolean = true): BooleanReturns true if the given path is a symbolic link. Follows symbolic links by
default, pass in followLinks = false to not do so.
os.isLink(wd / "misc/file-symlink") ==> true
os.isLink(wd / "misc/folder-symlink") ==> true
os.isLink(wd / "folder1") ==> falseos.size(p: Path): LongReturns the size of the given file, in bytes
os.size(wd / "File.txt") ==> 8
os.size(wd / "Multi Line.txt") ==> 81os.mtime(p: Path): Long
os.mtime.set(p: Path, millis: Long): UnitGets or sets the last-modified timestamp of the given file, in milliseconds
os.mtime.set(wd / "File.txt", 0)
os.mtime(wd / "File.txt") ==> 0
os.mtime.set(wd / "File.txt", 90000)
os.mtime(wd / "File.txt") ==> 90000
os.mtime(wd / "misc/file-symlink") ==> 90000
os.mtime.set(wd / "misc/file-symlink", 70000)
os.mtime(wd / "File.txt") ==> 70000
os.mtime(wd / "misc/file-symlink") ==> 70000
assert(os.mtime(wd / "misc/file-symlink", followLinks = false) != 40000)os.perms(p: Path, followLinks: Boolean = true): PermSet
os.perms.set(p: Path, arg2: PermSet): UnitGets or sets the filesystem permissions of the given file or folder, as an
os.PermSet.
Note that if you want to create a file or folder with a given set of
permissions, you can pass in an os.PermSet to os.write
or os.makeDir. That will ensure the file or folder is created
atomically with the given permissions, rather than being created with the
default set of permissions and having os.perms.set over-write them later
os.perms.set(wd / "File.txt", "rwxrwxrwx")
os.perms(wd / "File.txt").toString() ==> "rwxrwxrwx"
os.perms(wd / "File.txt").toInt() ==> Integer.parseInt("777", 8)
os.perms.set(wd / "File.txt", Integer.parseInt("755", 8))
os.perms(wd / "File.txt").toString() ==> "rwxr-xr-x"
os.perms.set(wd / "File.txt", "r-xr-xr-x")
os.perms.set(wd / "File.txt", Integer.parseInt("555", 8))os.owner(p: Path, followLinks: Boolean = true): UserPrincipal
os.owner.set(arg1: Path, arg2: UserPrincipal): Unit
os.owner.set(arg1: Path, arg2: String): UnitGets or sets the owner of the given file or folder. Note that your process needs
to be running as the root user in order to do this.
val originalOwner = os.owner(wd / "File.txt")
os.owner.set(wd / "File.txt", "nobody")
os.owner(wd / "File.txt").getName ==> "nobody"
os.owner.set(wd / "File.txt", originalOwner)os.group(p: Path, followLinks: Boolean = true): GroupPrincipal
os.group.set(arg1: Path, arg2: GroupPrincipal): Unit
os.group.set(arg1: Path, arg2: String): UnitGets or sets the owning group of the given file or folder. Note that your
process needs to be running as the root user in order to do this.
val originalOwner = os.owner(wd / "File.txt")
os.owner.set(wd / "File.txt", "nobody")
os.owner(wd / "File.txt").getName ==> "nobody"
os.owner.set(wd / "File.txt", originalOwner)Subprocess are spawned using os.call(cmd: os.Shellable, ...) or
os.spawn(cmd: os.Shellable, ...) calls,
where the cmd: Shellable sets up the basic command you wish to run and
.foo(...) specifies how you want to run it. os.Shellable represents a value
that can make up part of your subprocess command, and the following values can
be used as os.Shellables:
-
java.lang.String -
scala.Symbol -
os.Path -
os.RelPath -
T: Numeric -
Iterable[T]s of any of the above -
TupleN[T1, T2, …Tn]s of any of the above
Most of the subprocess commands also let you redirect the subprocess’s
stdin/stdout/stderr streams via os.ProcessInput or os.ProcessOutput
values: whether to inherit them from the parent process, stream them into
buffers, or output them to files. The following values are common to both input
and output:
-
os.Pipe: the default, this connects the subprocess’s stream to the parent process via pipes; if used on its stdin this lets the parent process write to the subprocess viaos.SubProcess#stdin, and if used on its stdout it lets the parent process read from the subprocess viaos.SubProcess#stdoutandos.SubProcess#stderr. -
os.Inherit: inherits the stream from the parent process. This lets the subprocess read directly from the parent process’s standard input or write directly to the parent process’s standard output or error.os.Inheritcan be redirected on a threadlocal basis viaos.Inherit.in,.out, or.err. -
os.InheritRaw: identical toos.Inherit, but without being affected by redirects. -
os.Path: connects the subprocess’s stream to the given filesystem path, reading its standard input from a file or writing its standard output/error to the file.
In addition, you can pass any os.Sources to a Subprocess’s stdin
(Strings, InputStreams, Array[Byte]s, …), and pass in a
os.ProcessOutput value to stdout/stderr to register callbacks that are run
when output is received on those streams.
Often, if you are only interested in capturing the standard output of the
subprocess but want any errors sent to the console, you might set stderr =
os.Inherit while leaving stdout = os.Pipe.
os.call(cmd: os.Shellable,
cwd: Path = null,
env: Map[String, String] = null,
stdin: ProcessInput = Pipe,
stdout: ProcessOutput = Pipe,
stderr: ProcessOutput = Pipe,
mergeErrIntoOut: Boolean = false,
timeout: Long = Long.MaxValue,
check: Boolean = true,
propagateEnv: Boolean = true,
shutdownGracePeriod: Long = 100,
destroyOnExit: Boolean = true): os.CommandResultAlso callable via `os.proc(cmd).call(…)`
Invokes the given subprocess like a function, passing in input and returning a
CommandResult. You can then call result.exitCode to see how it exited, or
result.out.bytes or result.err.string to access the aggregated stdout and
stderr of the subprocess in a number of convenient ways.
call provides a number of parameters that let you configure how the subprocess
is run:
-
cwd: the working directory of the subprocess -
env: any additional environment variables you wish to set in the subprocess -
stdin: any data you wish to pass to the subprocess’s standard input -
stdout/stderr: these areos.Redirects that let you configure how the processes output/error streams are configured. -
mergeErrIntoOut: merges the subprocess’s stderr stream into it’s stdout -
timeout: how long to wait for the subprocess to complete -
check: disable this to avoid throwing an exception if the subprocess fails with a non-zero exit code -
propagateEnv: disable this to avoid passing in this parent process’s environment variables to the subprocess
Note that redirecting stdout/stderr elsewhere means that the respective
CommandResult#out/CommandResult#err values will be empty.
val res = os.call(cmd = ('ls, wd/"folder2"))
res.exitCode ==> 0
res.out.text() ==>
"""nestedA
|nestedB
|""".stripMargin
res.out.trim() ==>
"""nestedA
|nestedB""".stripMargin
res.out.lines() ==> Seq(
"nestedA",
"nestedB"
)
res.out.bytes
// Non-zero exit codes throw an exception by default
val thrown = intercept[os.SubprocessException]{
os.call(cmd = ('ls, "doesnt-exist"), cwd = wd)
}
assert(thrown.result.exitCode != 0)
// Though you can avoid throwing by setting `check = false`
val fail = os.call(cmd = ('ls, "doesnt-exist"), cwd = wd, check = false)
assert(fail.exitCode != 0)
fail.out.text() ==> ""
assert(fail.err.text().contains("No such file or directory"))
// You can pass in data to a subprocess' stdin
val hash = os.call(cmd = ("shasum", "-a", "256"), stdin = "Hello World")
hash.out.trim() ==> "a591a6d40bf420404a011733cfb7b190d62c65bf0bcda32b57b277d9ad9f146e -"
// Taking input from a file and directing output to another file
os.call(cmd = ("base64"), stdin = wd / "File.txt", stdout = wd / "File.txt.b64")
os.read(wd / "File.txt.b64") ==> "SSBhbSBjb3c="If you want to spawn an interactive subprocess, such as vim, less, or a
python shell, set all of stdin/stdout/stderr to os.Inherit:
os.proc("vim").call(stdin = os.Inherit, stdout = os.Inherit, stderr = os.Inherit)Note that by customizing stdout and stderr, you can use the results
of os.proc.call in a streaming fashion, either on groups of bytes:
var lineCount = 1
os.call(
cmd = ('find, "."),
cwd = wd,
stdout = os.ProcessOutput(
(buf, len) => lineCount += buf.slice(0, len).count(_ == '\n')
),
)Or on lines of output:
lineCount ==> 22
var lineCount = 1
os.call(
cmd = ('find, "."),
cwd = wd,
stdout = os.ProcessOutput.Readlines(
line => lineCount += 1
),
)
lineCount ==> 22os.spawn(cmd: os.Shellable,
cwd: Path = null,
env: Map[String, String] = null,
stdin: os.ProcessInput = os.Pipe,
stdout: os.ProcessOutput = os.Pipe,
stderr: os.ProcessOutput = os.Pipe,
mergeErrIntoOut: Boolean = false,
propagateEnv: Boolean = true,
shutdownGracePeriod: Long = 100,
destroyOnExit: Boolean = true): os.SubProcessAlso callable via `os.proc(cmd).spawn(…)`
The most flexible of the os.proc calls, os.spawn simply configures and
starts a subprocess, and returns it as a os.SubProcess. os.SubProcess is a
simple wrapper around java.lang.Process, which provides stdin, stdout, and
stderr streams for you to interact with however you like. e.g. You can sending
commands to it’s stdin and reading from it’s stdout.
To implement pipes, you can spawn a process, take its stdout, and pass it as the stdin of a second spawned process.
Note that if you provide ProcessOutput callbacks to stdout/stderr, the
calls to those callbacks take place on newly spawned threads that execute in
parallel with the main thread. Thus make sure any data processing you do in
those callbacks is thread safe!
stdin, stdout and stderr are java.lang.OutputStreams and
java.lang.InputStreams enhanced with the .writeLine(s: String)/.readLine()
methods for easy reading and writing of character and line-based data.
// Start a long-lived python process which you can communicate with
val sub = os.spawn(
cmd = ("python", "-u", "-c", "while True: print(eval(raw_input()))"),
cwd = wd
)
// Sending some text to the subprocess
sub.stdin.write("1 + 2")
sub.stdin.writeLine("+ 4")
sub.stdin.flush()
sub.stdout.readLine() ==> "7"
sub.stdin.write("'1' + '2'")
sub.stdin.writeLine("+ '4'")
sub.stdin.flush()
sub.stdout.readLine() ==> "124"
// Sending some bytes to the subprocess
sub.stdin.write("1 * 2".getBytes)
sub.stdin.write("* 4\n".getBytes)
sub.stdin.flush()
sub.stdout.read() ==> '8'.toByte
sub.destroy()
// You can chain multiple subprocess' stdin/stdout together
val curl = os.spawn(cmd = ("curl", "-L" , "https://git.io/fpfTs"), stderr = os.Inherit)
val gzip = os.spawn(cmd = ("gzip", "-n"), stdin = curl.stdout)
val sha = os.spawn(cmd = ("shasum", "-a", "256"), stdin = gzip.stdout)
sha.stdout.trim ==> "acc142175fa520a1cb2be5b97cbbe9bea092e8bba3fe2e95afa645615908229e -"Client-server CLI applications sometimes want to run subprocesses on the server based on the environment of the client.
It is possible to customize the default environment passed to subprocesses by setting the os.SubProcess.env threadlocal:
val clientEnvironment: Map[String, String] = ???
os.SubProcess.env.withValue(clientEnvironment) {
os.call(command) // clientEnvironment is passed by default instead of the system environment
}After constructing a subprocess with os.proc, you can use the pipeTo method
to pipe its output to another subprocess:
val wc = os.proc("ls", "-l")
.pipeTo(os.proc("wc", "-l"))
.call()
.out.text()This is equivalent to the shell command ls -l | wc -l. You can chain together
as many subprocesses as you like. Note that by using this API you can utilize
the broken pipe behaviour of Unix systems. For example, you can take 10 first elements
of output from the yes command, and after the head command terminates, the yes
command will be terminated as well:
val yes10 = os.proc("yes")
.pipeTo(os.proc("head", "-n", "10"))
.call()
.out.text()This feature is implemented inside the library and will terminate any process reading the
stdin of other process in pipeline on every IO error. This behavior can be disabled via the
handleBrokenPipe flag on call and spawn methods. Note that Windows does not support
broken pipe behaviour, so a command like`yes` would run forever. handleBrokenPipe is set
to false by default on Windows.
Both call and spawn correspond in their behavior to their counterparts in the os.proc,
but spawn returns the os.ProcessPipeline instance instead. It offers the same
API as SubProcess, but will operate on the set of processes instead of a single one.
Pipefail is enabled by default, so if any of the processes in the pipeline fails, the whole
pipeline will have a non-zero exit code. This behavior can be disabled via the pipefail flag
on call and spawn methods. Note that the pipefail does not kill the processes in the pipeline,
it just sets the exit code of the pipeline to the exit code of the failed process.
os.watch.watch(roots: Seq[os.Path], onEvent: Set[os.Path] => Unit): Unit// Mill
ivy"com.lihaoyi::os-lib-watch:0.11.3"
// SBT
"com.lihaoyi" %% "os-lib-watch" % "0.11.3"Efficiently watches the given roots folders for changes. Any time the
filesystem is modified within those folders, the onEvent callback is
called with the paths to the changed files or folders. Note that
os.watch.watch is under a different artifact than the rest of the
os.* functions, and you need to add a separate dependency to
os-lib-watch in order to pull it in.
Once the call to watch returns, onEvent is guaranteed to receive a
an event containing the path for:
-
Every file or folder that gets created, deleted, updated or moved within the watched folders
-
For copied or moved folders, the path of the new folder as well as every file or folder within it.
-
For deleted or moved folders, the root folder which was deleted/moved, but without the paths of every file that was within it at the original location
Note that watch does not provide any additional information about the
changes happening within the watched roots folder, apart from the path
at which the change happened. It is up to the onEvent handler to query
the filesystem and figure out what happened, and what it wants to do.
Here is an example of use from the Ammonite REPL:
@ import $ivy.`com.lihaoyi::os-lib-watch:0.11.3`
@ os.watch.watch(Seq(os.pwd / "out"), paths => println("paths changed: " + paths.mkString(", ")))
@ os.write(os.pwd / "out/i am", "cow")
paths changed: /Users/lihaoyi/Github/Ammonite/out/i am
@ os.move(os.pwd / "out/i am", os.pwd / "out/hear me")
paths changed: /Users/lihaoyi/Github/Ammonite/out/i am,/Users/lihaoyi/Github/Ammonite/out/hear me
@ os.remove.all(os.pwd / "out/version")
paths changed: /Users/lihaoyi/Github/Ammonite/out/version/log,/Users/lihaoyi/Github/Ammonite/out/version/meta.json,/Users/lihaoyi/Github/Ammonite/out/versionOS-Lib uses strongly-typed data-structures to represent filesystem paths. The two basic versions are:
-
os.Path: an absolute path, starting from the root -
os.RelPath: a relative path, not rooted anywhere -
os.SubPath: a sub path, without any..segments, not rooted anywhere
Generally, almost all commands take absolute os.Paths. These are
basically java.nio.file.Paths with additional guarantees:
-
os.Paths are always absolute. Relative paths are a separate typeos.RelPath -
os.Paths are always canonical. You will never find.or..segments in them, and never need to worry about calling.normalizebefore operations.
Absolute paths can be created in a few ways:
// Get the process' Current Working Directory. As a convention
// the directory that "this" code cares about (which may differ
// from the pwd) is called `wd`
val wd = os.pwd
// A path nested inside `wd` in multiple segments
wd / "folder" / "file"
// The RHS of `/` can have multiple segments if-and-only-if it is a literal string
wd / "folder/file"
// A path starting from the root
os.root / "folder/file"
// A path with spaces or other special characters
wd / "My Folder/My File.txt"
// Up one level from the wd
wd / os.up
// Up two levels from the wd
wd / os.up / os.upWhen constructing os.Path`s, the right-hand-side of the `/ operator must be either a non-literal
a string expression containing a single path segment or a literal string containing one-or-more
path segments. If a non-literal string expression on the RHS contains multiple segments, you need
to wrap the RHS in an explicit os.RelPath(…) or os.SubPath(…) constructor to tell OS-Lib
how to interpret it. The single-segment limitation is intended to avoid the developer accidentally
introducing Directory Traversal Attacks
or other related bugs when naively constructing paths out of dynamic and potentially untrusted
inputs, which is not an issue for literal string since the string value is directly written in
the source code and immediately visible.
os.pwd can be modified in certain scopes via the os.dynamicPwd dynamic variable, but
best practice is not to change it. Instead simply define a new path, e.g.
val target = os.pwd / "target"Should be sufficient for most needs.
Above, we made use of the os.pwd built-in path. There are a number of Paths
built into OS-Lib:
-
os.pwd: The current working directory of the process. This can’t be changed in Java, so if you need another path to work with the convention is to define awdvariable. -
os.root: The root of the filesystem. -
os.home: The home directory of the current user. -
os.temp()/os.temp.dir(): Creates a temporary file/folder and returns the path.
os.RelPaths represent relative paths. These are basically defined as:
class RelPath private[ops] (segments0: Array[String], val ups: Int)The same data structure as Paths, except that they can represent a number of ups before the relative path is applied. They can be created in the following ways:
// The path "folder/file" in multiple segments
val rel1 = os.rel / "folder" / "file"
// RHS of `/` can have multiple segments if-and-only-if it is a literal string
val rel2 = os.rel / "folder/file"
// The path "file"
val rel3 = os.rel / "file"
// The relative difference between two paths
val target = os.pwd / "target/file"
assert((target.relativeTo(os.pwd)) == os.rel / "target/file")
// `up`s get resolved automatically
val minus = os.pwd.relativeTo(target)
val ups = os.up / os.up
assert(minus == ups)In general, very few APIs take relative paths. Their main purpose is to be combined with absolute paths in order to create new absolute paths. e.g.
val target = os.pwd / "target/file"
val difference = target.relativeTo(os.pwd)
val newBase = os.root / "code/server"
assert(newBase / difference == os.root / "code/server/target/file")os.up is a relative path that comes in-built:
val target = os.root / "target/file"
assert(target / os.up == os.root / "target")Note that all paths, both relative and absolute, are always expressed in a canonical manner:
assert((os.root / "folder/file" / os.up).toString == "/folder")
// not "/folder/file/.."
assert((os.rel / "folder/file" / os.up).toString == "folder")
// not "folder/file/.."So you don’t need to worry about canonicalizing your paths before comparing them for equality or otherwise manipulating them.
os.SubPaths represent relative paths without any .. segments. These
are basically defined as:
class SubPath private[ops] (segments0: Array[String])They can be created in the following ways:
// The path "folder/file" in multiple segments
val sub1 = os.sub / "folder" / "file"
// RHS of `/` can have multiple segments if-and-only-if it is a literal string
val sub2 = os.sub / "folder/file"
// The relative difference between two paths
val target = os.pwd / "out/scratch/file"
assert((target subRelativeTo os.pwd) == os.sub / "out/scratch/file")
// Converting os.RelPath to os.SubPath
val rel3 = os.rel / "folder/file"
val sub3 = rel3.asSubPathos.SubPaths are useful for representing paths within a particular
folder or directory. You can combine them with absolute os.Paths to
resolve paths within them, without needing to worry about
Directory Traversal Attacks
du to accidentally accessing paths outside the destination folder.
val target = os.pwd / "target/file"
val difference = target.relativeTo(os.pwd)
val newBase = os.root / "code/server"
assert(newBase / difference == os.root / "code/server/target/file")Attempting to construct an os.SubPath with .. segments results in an
exception being thrown:
val target = os.pwd / "out/scratch" /
// `up`s are not allowed in sub paths
intercept[Exception](os.pwd subRelativeTo target)Like os.Paths and os.RelPath, os.SubPaths are always canonicalized
and can be compared for equality without worrying about different
representations.
OS-Lib’s paths are transparent data-structures, and you can always access the segments and ups directly. Nevertheless, OS-Lib defines a number of useful operations that handle the common cases of dealing with these paths:
In this definition, ThisType represents the same type as the current path; e.g. a Path’s / returns a Path while a RelPath’s / returns a RelPath. Similarly, you can only compare or subtract paths of the same type.
Apart from os.RelPaths themselves, a number of other data
structures are convertible into os.RelPaths when spliced into a
path using /:
-
Strings -
Symbols -
Array[T]s whereTis convertible into a RelPath -
Seq[T]s whereTis convertible into a RelPath
Apart from built-ins like os.pwd or os.root or os.home, you can also
construct Paths from Strings, java.io.Files or java.nio.file.Paths:
val relStr = "hello/cow/world/.."
val absStr = "/hello/world"
assert(
RelPath(relStr) == "hello/cow",
// Path(...) also allows paths starting with ~,
// which is expanded to become your home directory
Path(absStr) == os.root / "hello/world"
)
// You can also pass in java.io.File and java.nio.file.Path
// objects instead of Strings when constructing paths
val relIoFile = new java.io.File(relStr)
val absNioFile = java.nio.file.Paths.get(absStr)
assert(
RelPath(relIoFile) == "hello/cow",
Path(absNioFile) == os.root / "hello/world",
Path(relIoFile, root / "base") == os.root / "base/hello/cow"
)Trying to construct invalid paths fails with exceptions:
val relStr = "hello/.."
intercept[java.lang.IllegalArgumentException]{
Path(relStr)
}
val absStr = "/hello"
intercept[java.lang.IllegalArgumentException]{
RelPath(absStr)
}
val tooManyUpsStr = "/hello/../.."
intercept[PathError.AbsolutePathOutsideRoot.type]{
Path(tooManyUpsStr)
}As you can see, attempting to parse a relative path with os.Path or
an absolute path with os.RelPath throws an exception. If you’re
uncertain about what kind of path you are getting, you could use BasePath to
parse it :
val relStr = "hello/cow/world/.."
val absStr = "/hello/world"
assert(
FilePath(relStr) == "hello/cow",
FilePath(absStr) == os.root / "hello/world"
)This converts it into a BasePath, which is either a os.Path or
os.RelPath. It’s then up to you to pattern-match on the types and
decide what you want to do in each case.
You can also pass in a second argument to Path(..., base). If the path being
parsed is a relative path, this base will be used to coerce it into an absolute
path:
val relStr = "hello/cow/world/.."
val absStr = "/hello/world"
val basePath: FilePath = FilePath(relStr)
assert(
os.Path(relStr, os.root / "base") == os.root / "base/hello/cow",
os.Path(absStr, os.root / "base") == os.root / "hello/world",
os.Path(basePath, os.root / "base") == os.root / "base/hello/cow",
os.Path(".", os.pwd).last != ""
)For example, if you wanted the common behavior of converting relative paths to
absolute based on your current working directory, you can pass in os.pwd as
the second argument to Path(...). Apart from passing in Strings or
java.io.Files or java.nio.file.Paths, you can also pass in BasePaths you parsed
early as a convenient way of converting it to a absolute path, if it isn’t
already one.
In general, OS-Lib is very picky about the distinction between relative and absolute paths, and doesn’t allow "automatic" conversion between them based on current-working-directory the same way many other filesystem APIs (Bash, Java, Python, …) do. Even in cases where it’s uncertain, e.g. you’re taking user input as a String, you have to either handle both possibilities with BasePath or explicitly choose to convert relative paths to absolute using some base.
If you are using a system that supports different roots of paths, e.g. Windows,
you can use the argument of os.root to specify which root you want to use.
If not specified, the default root will be used (usually, C on Windows, / on Unix).
val root = os.root("C:\\") / "Users/me"
assert(root == os.Path("C:\\Users\\me"))Additionally, custom filesystems can be specified by passing a FileSystem to
os.root. This allows you to use OS-Lib with non-standard filesystems, such as
jar filesystems or in-memory filesystems.
val uri = new URI("jar", Paths.get("foo.jar").toURI().toString, null);
val env = new HashMap[String, String]();
env.put("create", "true");
val fs = FileSystems.newFileSystem(uri, env);
val path = os.root("/", fs) / "dir"Note that the jar file system operations suchs as writing to a file are supported
only on JVM 11+. Depending on the filesystem, some operations may not be supported -
for example, running an os.proc with pwd in a jar file won’t work. You may also
meet limitations imposed by the implementations - in jar file system, the files are
created only after the file system is closed. Until that, the ones created in your
program are kept in memory.
In addition to manipulating paths on the filesystem, you can also manipulate
os.ResourcePath in order to read resources off of the Java classpath. By
default, the path used to load resources is absolute, using the
Thread.currentThread().getContextClassLoader.
val contents = os.read(os.resource / "test/ammonite/ops/folder/file.txt")
assert(contents.contains("file contents lols"))You can also pass in a classloader explicitly to the resource call:
val cl = getClass.getClassLoader
val contents2 = os.read(os.resource(cl)/ "test/ammonite/ops/folder/file.txt")
assert(contents2.contains("file contents lols"))If you want to load resources relative to a particular class, pass in a class for the resource to be relative, or getClass to get something relative to the current class.
val cls = classOf[test.os.Testing]
val contents = os.read(os.resource(cls) / "folder/file.txt")
assert(contents.contains("file contents lols"))
val contents2 = os.read(os.resource(getClass) / "folder/file.txt")
assert(contents2.contains("file contents lols"))In both cases, reading resources is performed as if you did not pass a leading
slash into the getResource("foo/bar") call. In the case of
ClassLoader#getResource, passing in a leading slash is never valid, and in the
case of Class#getResource, passing in a leading slash is equivalent to calling
getResource on the ClassLoader.
OS-Lib ensures you only use the two valid cases in the API, without a leading
slash, and not the two cases with a leading slash which are redundant (in the
case of Class#getResource, which can be replaced by ClassLoader#getResource)
or invalid (a leading slash with ClassLoader#getResource)
Note that you can only use os.read from resource paths; you can’t write to them or
perform any other filesystem operations on them, since they’re not really files.
Note also that resources belong to classloaders, and you may have multiple classloaders in your application e.g. if you are running in a servlet or REPL. Make sure you use the correct classloader (or a class belonging to the correct classloader) to load the resources you want, or else it might not find them.
Many operations in OS-Lib operate on os.Sources. These represent values that
can provide data which you can then use to write, transmit, etc.
By default, the following types of values can be used where-ever os.Sources
are required:
-
Any
geny.Writabledata type:-
Array[Byte] -
java.lang.String(these are treated as UTF-8) -
java.io.InputStream
-
-
java.nio.channels.SeekableByteChannel -
Any
TraversableOnce[T]of the above: e.g.Seq[String],List[Array[Byte]], etc.
Some operations only work on os.SeekableSource, because they need the ability
to seek to specific offsets in the data. Only the following types of values can
be used where os.SeekableSource is required:
-
java.nio.channels.SeekableByteChannel
os.Source also supports anything that implements the
Writable interface, such as
ujson.Values,
uPickle's upickle.default.writable values,
or Scalatags's Tags
You can also convert an os.Path or os.ResourcePath to an os.Source via
.toSource.
Taken from the geny library, os.Generators
are similar to iterators except instead of providing:
-
def hasNext(): Boolean -
def next(): T
os.Generators provide:
-
def generate(handleItem: A => Generator.Action): Generator.Action
In general, you should not notice much of a difference using Generators vs
using Iterators: you can use the same .map/.filter/.reduce/etc.
operations on them, and convert them to collections via the same
.toList/.toArray/etc. conversions. The main difference is that Generators
can enforce cleanup after traversal completes, so we can ensure open files are
closed and resources are released without any accidental leaks.
os.PermSets represent the filesystem permissions on a single file or folder.
Anywhere an os.PermSet is required, you can pass in values of these types:
-
java.lang.Strings of the form"rw-r-xrwx", withr/w/xrepresenting the permissions that are present or dashes-representing the permissions which are absent -
Octal
Ints of the formInteger.parseInt("777", 8), matching the octal755or666syntax used on the command line -
Set[PosixFilePermission]
In places where os.PermSets are returned to you, you can then extract the
string, int or set representations of the os.PermSet via:
-
perms.toInt(): Int -
perms.toString(): String -
perms.value: Set[PosixFilePermission]
-
Add ability to instrument path based operations using hooks #325
-
SubProcessspawning operations now take andestroyOnExit = trueflag to try and shut them down when the host JVM exits,SubProcess#destroynow takes a configurable(shutdownGracePeriod: Long, async: Boolean)flags to configure the behavior (superseding the olddestroy()/destroyForcibly()methods), andtimeoutGracePeriodhas been renamed toshutdownGracePeriod#324
-
Use
java.nio.files.Files.newOutputStreaminstead ofjava.io.FileOutputStreamto try and avoid problems with windows open file deletion #323
-
Propagate content length from filesystem through
geny.Writableandos.Source#320
-
Added APIs to Zip & Unzip Files via
os.zip,os.unzip,os.zip.stream,os.unzip.stream,os.unzip.list,os.unzip.streamRaw,os.zip.open#317 -
Minimum officially supported Java version raised from 8 to 11
-
Allow multi-segment paths segments for literals #297: You can now write
os.pwd / "foo/bar/qux"rather thanos.pwd / "foo" / "bar" / "qux". Note that this is only allowed for string literals, and non-literal path segments still need to be wrapped e.g.def myString = "foo/bar/qux"; os.pwd / os.SubPath(myString)for security and safety purposes
-
Make
os.pwdmodifiable via theos.dynamicPwddynamic variable #298
-
Introduce
os.SubProcess.envDynamicVariableto override defaultenv(#295)
-
os.Inheritnow can be redirected on a threadlocal basis viaos.Inherit.in,.out, or.err.os.InheritRawis available if you do not want the redirects to take effect
-
Support
os.procon Scala Native (#257)
-
Support for Scala-Native 0.5.0
-
Dropped support for Scala 2.11.x
-
Minimum version of Scala 3 increased to 3.3.1
-
Fix
os.watchon Windows (#236) -
Fix propagateEnv = false to not propagate env (#238)
-
Make os.home a def (#239)
-
Added new convenience API to create pipes between processes with
.pipeTo -
Fixed issue with leading
../os.upin path segments created from aSeq -
Fixed Windows-specific issues with relative paths with leading (back)slashes
-
Removed some internal use of deprecated API
-
ScalaDoc now maps some external references to their online sites
-
Dependency updates: sourcecode 0.3.1
-
Tooling updates: acyclic 0.3.9, Mill 0.11.5, mill-mima 0.0.24, mill-vcs-version 0.4.0, scalafmt 3.7.15
-
Refined return types when constructing paths with
/and get rid of longThisType#ThisTypecascades. -
Added a new
PathConvertibleto support `URI`s when constructing paths.
-
os.procnow also supportsCharSequence(s)asShellable -
ProcessResultnow also contains the actual used command -
Fixed handling of
atimeandctimeinStatInfo -
Deleted
ConcurrentLinkedQueuefrom Scala Native jars, as it is now provided by Scala Native 0.4 itself -
Enabled MiMa checks to CI setup and officially support early semantic versioning since this release
-
Documentation improvements
-
Added support for Scala Native on Scala 3
-
Restored binary compatibility in
os.copyandos.copy.intoto os-lib versions before 0.7.5
-
Add support for Scala 3.0.0
-
Add support for Scala 3.0.0-RC3
-
Re-added support for Scala 2.11
-
Added new option
mergeFolderstoos.copy -
os.copy now honors
followLinkswhen copying symbolic links to directories
-
Add support for Scala 3.0.0-RC2
-
Add support for Scala 3.0.0-RC1
-
Migration of the CI system from Travis CI to GitHub Actions
-
Add support for Scala 3.0.0-M3
-
Improve performance of
os.writeby buffering output stream to files
-
Moved the
os.Bytes,os.StreamValue(now namedByteData) interfaces intogenypackage, for sharing with Requests-Scala -
Add
os.read.streamfunction, that returns ageny.Readable
-
os.Sourcenow supports any data type that isgeny.Writable
-
Added a new
os.SubPathdata type, for safer handling of sub-paths within a directory. -
Removed
os.proc.stream, since you can now customize thestdoutorstderrofos.proc.callto handle output in a streaming fashion -
stderrinos.proc.callandos.proc.spawndefaults toos.Inheritrather thanos.Pipe; pass instderr = os.Pipeexplicitly to get back the old behavior -
Fix timeout not working with
os.proc.call#27 -
Attempt to fix crasher accessing
os.pwd#24 -
Added an os-lib-watch package, which can be used to efficiently recursively watch folders for updates #23
-
os.statno longer provides POSIX owner/permissions related metadata by default #15, useos.stat.posixto fetch that separately -
os.stat.fullhas been superseded byos.statandos.stat.posix -
Removed
os.BasicStatInfo, which has been superseded byos.StatInfo
-
Support for Scala 2.13.0 final
-
os.ProcessOutputtrait is no longer sealed
-
Narrow return type of
readLink.absolutefromFilePathtoPath -
Fix handling of standaline
\rinos.SubProcess#stdout.readLine
-
Remove
os.StatInfo#name,os.BasicStatInfo#nameandos.FullStatInfo#name, since it is just the last path segment of the stat call and doesn’t properly reflect the actual name of the file on disk (e.g. on case-insensitive filesystems) -
os.walk.attrsandos.walk.stream.attrsnow provides aos.BasicFileInfoto theskippredicate. -
Add
os.BasePath#baseName, which returns the section of the path before theos.BasePath#extextension.
-
New
os.readLink/os.readLink.absolutemethods to read the contents of symbolic links without dereferencing them. -
New
os.read.chunked(p: Path, chunkSize: Int): os.Generator[(Array[Byte], Int)]method for conveniently iterating over chunks of a file -
New
os.truncate(p: Path, size: Int)method -
SubProcessstreams now implementjava.io.DataInput/DataOutputfor convenience -
SubProcessstreams are now synchronized for thread-safety -
os.writenow hascreateFoldersdefault tofalse -
os.Generatornow has a.withFiltermethod -
os.symlinknow allows relative paths -
os.remove.allnow properly removes broken symlinks, and no longer recurses into the symlink’s contents -
os.SubProcessnow implementsjava.lang.AutoCloseable -
New
write.channelcounterpart toread.channel(andwrite.over.channelandwrite.append.channel) -
os.PermSetis now modelled internally as a boxedIntfor performance, and is a case class with properequals/hashcode -
os.read.bytes(arg: Path, offset: Long, count: Int)no longer leaks open file channels -
Reversed the order of arguments in
os.symlinkandos.hardlink, to match the order of the underlying java NIO functions.
-
Allow chaining of multiple subprocesses
stdin/stdout
-
First release