better-files
is a dependency-free pragmatic thin Scala wrapper around Java NIO.
Imagine you have to write the following method:
- List all
.csv
files in a directory by increasing order of file size - Drop the first line of each file and concat the rest into a single output file
- Split the above output file into
n
smaller files without breaking up the lines in the input files gzip
each of the smaller output files
Note: Your program should work when files are much bigger than memory in your JVM and must close all open resources correctly
The above task is not that easy to write in Java or shell or Python without a certain amount of Googling. Using better-files, the above problem can be solved in a fairly straightforward way:
import better.files._
def run(inputDir: File, outputDir: File, n: Int) = {
val count = new AtomicInteger()
val outputs = Vector.tabulate(n)(i => outputDir / s"part-$i.csv.gz")
for {
writers <- outputs.map(_.newGzipOutputStream().printWriter()).autoClosed
inputFile <- inputDir.list(_.extension == Some(".csv")).toSeq.sorted(File.Order.bySize)
line <- inputFile.lineIterator.drop(1)
} writers(count.incrementAndGet() % n).println(line)
}
![ScalaDays NYC 2016: Introduction to better-files](https://raw.githubusercontent.com/NoriakiHoriuchi/better-files/master/site/tech_talk_preview.png)
- Instantiation
- Simple I/O
- Resource APIs
- Streams
- Encodings
- Java serialization utils
- Java compatibility
- Globbing
- File system operations
- Checksums
- Temporary files
- UNIX DSL
- File attributes
- File comparison
- Zip/GZip
- Automatic Resource Management
- Scanner
- File Monitoring
- Reactive File Watcher
In your build.sbt
, add this:
libraryDependencies += "com.github.pathikrit" %% "better-files" % version
To use the Akka based file monitor, also add this:
libraryDependencies ++= Seq(
"com.github.pathikrit" %% "better-files-akka" % version,
"com.typesafe.akka" %% "akka-actor" % "2.6.13"
)
Although this library is currently only actively developed for Scala 2.11, 2.12 and 2.13, you can find reasonably recent versions of this library for Scala 2.10 above.
The following are all equivalent:
import better.files._
import File._
import java.io.{File => JFile}
val f = File("/User/johndoe/Documents") // using constructor
val f1: File = file"/User/johndoe/Documents" // using string interpolator
val f2: File = "/User/johndoe/Documents".toFile // convert a string path to a file
val f3: File = new JFile("/User/johndoe/Documents").toScala // convert a Java file to Scala
val f4: File = root/"User"/"johndoe"/"Documents" // using root helper to start from root
val f5: File = `~` / "Documents" // also equivalent to `home / "Documents"`
val f6: File = "/User"/"johndoe"/"Documents" // using file separator DSL
val f7: File = "/User"/'johndoe/'Documents // same as above but using Symbols instead of Strings
val f8: File = home/"Documents"/"presentations"/`..` // use `..` to navigate up to parent
Note: Rename the import if you think the usage of the class File
may confuse your teammates:
import better.files.{File => ScalaFile, _}
import java.io.File
I personally prefer renaming the Java crap instead:
import better.files._
import java.io.{File => JFile}
Dead simple I/O:
val file = root/"tmp"/"test.txt"
file.overwrite("hello")
file.appendLine().append("world")
assert(file.contentAsString == "hello\nworld")
If you are someone who likes symbols, then the above code can also be written as:
import better.files.Dsl.SymbolicOperations
file < "hello" // same as file.overwrite("hello")
file << "world" // same as file.appendLines("world")
assert(file! == "hello\nworld")
Or even, right-associatively:
import better.files.Dsl.SymbolicOperations
"hello" `>:` file
"world" >>: file
val bytes: Array[Byte] = file.loadBytes
(root/"tmp"/"diary.txt")
.createIfNotExists()
.appendLine()
.appendLines("My name is", "Inigo Montoya")
.moveToDirectory(home/"Documents")
.renameTo("princess_diary.txt")
.changeExtensionTo(".md")
.lines
Confused by the various ways to load resources in Java? Worry no more:
val resource : InputStream = Resource.getAsStream("foo.txt") //Same as this.getClass().getResource("foo.txt")
val resourceURL : java.net.URL = Resource.getUrl("foo.txt")
val rootResourceURL : java.net.URL = Resource.getUrl()
val resourceAsStr : String = Resource.getAsString("foo.txt")
The above APIs can load from custom ClassLoader
s too:
val resource : InputStream = Resource.at[MyClass].getAsStream("foo.txt")
Various ways to slurp a file without loading its contents into memory:
val bytes : Iterator[Byte] = file.bytes
val chars : Iterator[Char] = file.chars
val lines : Iterator[String] = file.lineIterator //file.lines loads all lines in memory
Note: The above APIs can be traversed at most once e.g. file.bytes
is a Iterator[Byte]
which only allows TraversableOnce
.
To traverse it multiple times without creating a new iterator instance, convert it into some other collection e.g. file.bytes.toStream
You can write an Iterator[Byte]
or an Iterator[String]
back to a file:
file.writeBytes(bytes)
file.printLines(lines)
tee
multiple outputstreams:
val s3 = s1 tee s2
s3.printWriter.println(s"Hello world") // gets written to both s1 and s2
You can supply your own charset too for anything that does a read/write (it assumes java.nio.charset.Charset.defaultCharset()
if you don't provide one):
val content: String = file.contentAsString // default charset
// custom charset:
import java.nio.charset.Charset
file.contentAsString(charset = Charset.forName("US-ASCII"))
//or simply using implicit conversion from Strings
file.write("hello world")(charset = "US-ASCII")
Note: By default, better-files
correctly handles BOMs while decoding.
If you wish to have the incorrect JDK behaviour,
you would need to supply Java's UTF-8 charset e.g.:
file.contentAsString(charset = Charset.forName("UTF-8")) // Default incorrect JDK behaviour for UTF-8 (see: JDK-4508058)
If you also wish to write BOMs while encoding, you would need to supply it as:
file.write("hello world")(charset = UnicodeCharset("UTF-8", writeByteOrderMarkers = true))
Some common utils to serialize/deserialize using Java's serialization util
case class Person(name: String, age: Int)
val person = new Person("Chris", 24)
// Write
file.newOutputStream.asObjectOutputStream.serialize(obj).flush()
// Read
val person2 = file.newInputStream.asObjectInputStream.deserialize[Person]
assert(person == person2)
// Read using custom class loader:
file.newInputStream.asObjectInputStreamUsingClassLoader(classLoader = myClassLoader).deserialize[Person]
The above can be simply written as:
val person2: Person = file.writeSerialized(person).readDeserialized[Person]()
assert(person == person2)
You can always access the Java I/O classes:
val file: File = tmp / "hello.txt"
val javaFile : java.io.File = file.toJava
val uri : java.net.URI = file.uri
val url : java.net.URL = file.url
val reader : java.io.BufferedReader = file.newBufferedReader
val outputstream : java.io.OutputStream = file.newOutputStream
val writer : java.io.BufferedWriter = file.newBufferedWriter
val inputstream : java.io.InputStream = file.newInputStream
val path : java.nio.file.Path = file.path
val fs : java.nio.file.FileSystem = file.fileSystem
val channel : java.nio.channel.FileChannel = file.newFileChannel
val ram : java.io.RandomAccessFile = file.newRandomAccess
val fr : java.io.FileReader = file.newFileReader
val fw : java.io.FileWriter = file.newFileWriter(append = true)
val printer : java.io.PrintWriter = file.newPrintWriter
The library also adds some useful implicits to above classes e.g.:
file1.reader > file2.writer // pipes a reader to a writer
System.in > file2.out // pipes an inputstream to an outputstream
src.pipeTo(sink) // if you don't like symbols
val bytes : Iterator[Byte] = inputstream.bytes
val bis : BufferedInputStream = inputstream.buffered
val bos : BufferedOutputStream = outputstream.buffered
val reader : InputStreamReader = inputstream.reader
val writer : OutputStreamWriter = outputstream.writer
val printer : PrintWriter = outputstream.printWriter
val br : BufferedReader = reader.buffered
val bw : BufferedWriter = writer.buffered
val mm : MappedByteBuffer = fileChannel.toMappedByteBuffer
val str : String = inputstream.asString //Read a string from an InputStream
val in : InputStream = str.inputStream
val reader : Reader = str.reader
val lines : Seq[String] = str.lines
better-files
also supports certain conversions that are not supported out of the box by the JDK
No need to port this to Scala:
val dir = "src"/"test"
val matches: Iterator[File] = dir.glob("**/*.{java,scala}")
// above code is equivalent to:
dir.listRecursively.filter(f => f.extension == Some(".java") || f.extension == Some(".scala"))
You can even use more advanced regex syntax instead of glob syntax:
val matches = dir.globRegex("^\\w*$".r) //equivalent to dir.glob("^\\w*$")(syntax = File.PathMatcherSyntax.regex)
By default, glob syntax in better-files
is different from
the default JDK glob behaviour since it always includes path. To use the default behaviour:
dir.glob("**/*.txt", includePath = false) // JDK default
//OR
dir.glob("*.txt", includePath = true) // better-files default
You can also extend the File.PathMatcherSyntax
to create your own matchers.
For custom cases:
dir.collectChildren(_.isSymbolicLink) // collect all symlinks in a directory
For simpler cases, you can always use dir.list
or dir.walk(maxDepth: Int)
Utilities to ls
, cp
, rm
, mv
, ln
, md5
, touch
, cat
etc:
file.touch()
file.delete() // unlike the Java API, also works on directories as expected (deletes children recursively)
file.clear() // If directory, deletes all children; if file clears contents
file.renameTo(newName: String)
file.moveTo(destination)
file.moveToDirectory(destination)
file.copyTo(destination) // unlike the default API, also works on directories (copies recursively)
file.copyToDirectory(destination)
file.linkTo(destination) // ln destination file
file.symbolicLinkTo(destination) // ln -s destination file
file.setOwner(user: String) // chown user file
file.setGroup(group: String) // chgrp group file
Seq(file1, file2) `>:` file3 // same as cat file1 file2 > file3 (must import import better.files.Dsl.SymbolicOperations)
Seq(file1, file2) >>: file3 // same as cat file1 file2 >> file3 (must import import better.files.Dsl.SymbolicOperations)
file.isReadLocked; file.isWriteLocked; file.isLocked
File.numberOfOpenFileDescriptors // number of open file descriptors
One liner checksum for files:
file.md5 // equivalent to file.checksum("md5")
file.sha1
file.sha256
file.sha512
Note: The above also works for directories (it recursively computes for each file in the directory).
For input/output streams:
val md5: String = inputstream.md5.hexDigest()
val md5: String = outputstream.md5.hexDigest()
The above consumes and closes the inputstream
. If you want to write it to a file AND also compute the sha512
, you can do:
val md5: String = inputstream.sha512.hexDigest(drainTo = someFile)
Utils to create temporary files:
File.newTemporaryDirectory()
File.newTemporaryFile()
The above APIs allow optional specifications of prefix
, suffix
and parentDir
.
These files are not deleted automatically on exit by the JVM (you have to set deleteOnExit
which adds to shutdownHook
).
A cleaner alternative is to use self-deleting file contexts which deletes the file immediately when done:
for {
tempFile <- File.temporaryFile()
} doSomething(tempFile) // tempFile is auto deleted at the end of this block - even if an exception happens
OR equivalently:
File.usingTemporaryFile() {tempFile =>
//do something
} // tempFile is auto deleted at the end of this block - even if an exception happens
You can make any files temporary (i.e. delete after use) by doing this:
val foo = File.home / "Downloads" / "foo.txt"
for {
temp <- foo.toTemporary
} doSomething(temp) // foo is deleted at the end of this block - even if an exception happens
All the above can also be expressed using methods reminiscent of the command line:
import better.files._
import better.files.Dsl._ // must import Dsl._ to bring in these utils
pwd / cwd // current dir
cp(file1, file2)
mv(file1, file2)
rm(file) /*or*/ del(file)
ls(file) /*or*/ dir(file)
ln(file1, file2) // hard link
ln_s(file1, file2) // soft link
cat(file1)
cat(file1) >>: file
touch(file)
mkdir(file)
mkdirs(file) // mkdir -p
chown(owner, file)
chgrp(owner, file)
chmod_+(permission, files) // add permission
chmod_-(permission, files) // remove permission
md5(file); sha1(file); sha256(file); sha512(file)
unzip(zipFile)(targetDir)
zip(file*)(targetZipFile)
ungzip(gzipFile)(targetFile)
gzip(file)(targetGZipFile)
Query various file attributes e.g.:
file.name // simpler than java.io.File#getName
file.extension
file.contentType
file.lastModifiedTime // returns JSR-310 time
file.owner
file.group
file.isDirectory; file.isSymbolicLink; file.isRegularFile
file.isHidden
file.hide(); file.unhide()
file.isOwnerExecutable; file.isGroupReadable // etc. see file.permissions
file.size // for a directory, computes the directory size
file.posixAttributes; file.dosAttributes // see file.attributes
file.isEmpty // true if file has no content (or no children if directory) or does not exist
file.isParentOf; file.isChildOf; file.isSiblingOf; file.siblings
file("dos:system") = true // set custom meta-data for file (similar to Files.setAttribute)
All the above APIs let you specify the LinkOption
either directly:
file.isDirectory(LinkOption.NOFOLLOW_LINKS)
Or using the File.LinkOptions
helper:
file.isDirectory(File.LinkOptions.noFollow)
chmod
:
import java.nio.file.attribute.PosixFilePermission
file.addPermission(PosixFilePermission.OWNER_EXECUTE) // chmod +X file
file.removePermission(PosixFilePermission.OWNER_WRITE) // chmod -w file
assert(file.permissionsAsString == "rw-r--r--")
// The following are all equivalent:
assert(file.permissions contains PosixFilePermission.OWNER_EXECUTE)
assert(file.testPermission(PosixFilePermission.OWNER_EXECUTE))
assert(file.isOwnerExecutable)
Use ==
to check for path-based equality and ===
for content-based equality:
file1 == file2 // equivalent to `file1.isSamePathAs(file2)`
file1 === file2 // equivalent to `file1.isSameContentAs(file2)` (works for regular-files and directories)
file1 != file2 // equivalent to `!file1.isSamePathAs(file2)`
file1 !== file2 // equivalent to `!file1.isSameContentAs(file2)`
There are also various Ordering[File]
instances included, e.g.:
val files = myDir.list.toSeq
files.sorted(File.Order.byName)
files.max(File.Order.bySize)
files.min(File.Order.byDepth)
files.max(File.Order.byModificationTime)
files.sorted(File.Order.byDirectoriesFirst)
You don't have to lookup on StackOverflow "How to zip/unzip/gzip in Java/Scala?":
// Unzipping:
val zipFile: File = file"path/to/research.zip"
val research: File = zipFile.unzipTo(destination = home/"Documents"/"research")
// Zipping:
val zipFile: File = directory.zipTo(destination = home/"Desktop"/"toEmail.zip")
// Zipping in:
val zipFile = File("countries.zip").zipIn(Iterator(file"usa.txt", file"russia.txt"))()
// Zipping/Unzipping to temporary files/directories:
val someTempZipFile: File = directory.zip()
val someTempDir: File = someTempZipFile.unzip()
assert(directory === someTempDir)
Mapping over each ZipEntry
:
val fileNames = zipFile.newZipInputStream.mapEntries(_.getName) // gets the file names inside the zip file
GZIP handling:
File("big-data.csv").gzipTo(File("big-data.csv.gz"))
File("big-data.csv.gz").unGzipTo(File("big-data.csv"))
// GZIP stream handling:
File("countries.gz").newInputStream.asGzipInputStream().lines.take(10).foreach(println)
def write(out: OutputStream, countries: Seq[String]) =
out.asGzipOutputStream().printWriter().printLines(countries).close()
Auto-close Java closeables:
for {
in <- file1.newInputStream.autoClosed
out <- file2.newOutputStream.autoClosed
} in.pipeTo(out)
// The input and output streams are auto-closed once out of scope
better-files
provides convenient managed versions of all the Java closeables e.g. instead of writing:
for {
reader <- file.newBufferedReader.autoClosed
} foo(reader)
You can write:
for {
reader <- file.bufferedReader // returns Dispose[BufferedReader]
} foo(reader)
// or simply:
file.bufferedReader.foreach(foo)
Similarly:
for {
reader <- file.bufferedReader
} yield foo(reader)
// Simpler
file.bufferedReader.map(foo).get()
// Even simpler
file.bufferedReader.apply(foo)
If foo
itself is lazy and depends on reader
being open, you should flatMap
instead of apply
:
def lines(reader: BufferedReader): Iterator[String] = ???
for {
reader <- file.bufferedReader
line <- lines(reader)
} yield line
// or simply
file.bufferedReader.flatMap(lines)
You can also define your own custom disposable resources e.g.:
trait Shutdownable {
def shutdown(): Unit = ()
}
object Shutdownable {
implicit val disposable: Disposable[Shutdownable] = Disposable(_.shutdown())
}
val s: Shutdownable = ....
for {
instance <- new Dispose(s)
} doSomething(s) // s is disposed after this
using
syntax:
val lines: List[String] = using(file.newInputStream) { stream =>
stream.lines.toList // Must be eager so .toList
}
Although java.util.Scanner
has a feature-rich API, it only allows parsing primitives.
It is also notoriously slow since it uses regexes and does un-Scala things like returns nulls and throws exceptions.
better-files
provides a faster, richer, safer, more idiomatic and compossible Scala replacement
that does not use regexes, allows peeking, accessing line numbers, returns Option
s whenever possible and lets the user mixin custom parsers:
val f1 = File("/tmp/temp.txt")
val data = f1.overwrite(s"""Hello World
| 1 true
| 2 3
""".stripMargin)
val scanner: Scanner = data.newScanner()
assert(scanner.next[String] == "Hello")
assert(scanner.lineNumber == 1)
assert(scanner.next[String] == "World")
assert(scanner.next[(Int, Boolean)] == (1, true))
assert(scanner.nextLine() == " 2 3")
assert(!scanner.hasNext)
If you are simply interested in tokens, you can use file.tokens()
Writing your own custom scanners:
sealed trait Animal
case class Dog(name: String) extends Animal
case class Cat(name: String) extends Animal
implicit val animalParser: Scannable[Animal] = Scannable {scanner =>
val name = scanner.next[String]
if (name == "Garfield") Cat(name) else Dog(name)
}
val scanner = file.newScanner()
println(scanner.next[Animal])
The shapeless-scanner lets you scan HList
s:
val in = Scanner("""
12 Bob True
13 Mary False
26 Rick True
""")
import shapeless._
type Row = Int :: String :: Boolean :: HNil
val out = Seq.fill(3)(in.next[Row])
assert(out == Seq(
12 :: "Bob" :: true :: HNil,
13 :: "Mary" :: false :: HNil,
26 :: "Rick" :: true :: HNil
))
case class Person(id: Int, name: String, isMale: Boolean)
val out2 = Seq.fill(3)(in.next[Person])
Simple CSV reader:
val file = """
23,foo
42,bar
"""
val csvScanner = file.newScanner(StringSplitter.on(','))
csvScanner.next[Int] //23
csvScanner.next[String] //foo
Vanilla Java watchers:
import java.nio.file.{StandardWatchEventKinds => EventType}
val service: java.nio.file.WatchService = myDir.newWatchService
myDir.register(service, events = Seq(EventType.ENTRY_CREATE, EventType.ENTRY_DELETE))
The above APIs are cumbersome to use (involves a lot of type-casting and null-checking), are based on a blocking polling-based model, does not easily allow recursive watching of directories and nor does it easily allow watching regular files without writing a lot of Java boilerplate.
better-files
abstracts all the above ugliness behind a simple interface:
val watcher = new FileMonitor(myDir, recursive = true) {
override def onCreate(file: File, count: Int) = println(s"$file got created")
override def onModify(file: File, count: Int) = println(s"$file got modified $count times")
override def onDelete(file: File, count: Int) = println(s"$file got deleted")
}
watcher.start()
Thread.sleep(60 * 1000) // The above line starts the monitoring asynchronously
Sometimes, instead of overwriting each of the 3 methods above, it is more convenient to override the dispatcher itself:
import java.nio.file.{Path, StandardWatchEventKinds => EventType, WatchEvent}
val watcher = new FileMonitor(myDir, recursive = true) {
override def onEvent(eventType: WatchEvent.Kind[Path], file: File, count: Int) = eventType match {
case EventType.ENTRY_CREATE => println(s"$file got created")
case EventType.ENTRY_MODIFY => println(s"$file got modified $count")
case EventType.ENTRY_DELETE => println(s"$file got deleted")
}
}
There is also an external module which gives high performance file monitoring and interpolates with better-files. See: https://github.com/gmethvin/directory-watcher#better-files-integration-scala
better-files
also provides a powerful yet concise reactive file watcher
based on Akka actors that supports dynamic dispatches:
import akka.actor.{ActorRef, ActorSystem}
import better.files.File.home
import better.files.FileWatcher._
import java.nio.file.{StandardWatchEventKinds => EventType}
implicit val system = ActorSystem("mySystem")
val watcher: ActorRef = (home/"Downloads").newWatcher(recursive = true)
// register partial function for an event
watcher ! on(EventType.ENTRY_DELETE) {
case file if file.isDirectory => println(s"directory $file got deleted")
case file => println(s"$file got deleted")
}
// watch for multiple events
watcher ! when(events = EventType.ENTRY_CREATE, EventType.ENTRY_MODIFY) {
case (EventType.ENTRY_CREATE, file) => println(s"$file got created")
case (EventType.ENTRY_MODIFY, file) => println(s"$file got modified")
}
> sbt "core/testOnly better.files.benchmarks.*"
JavaScanner : 2191 ms
StringBuilderScanner : 1325 ms
CharBufferScanner : 1117 ms
StreamingScanner : 212 ms
IterableScanner : 365 ms
IteratorScanner : 297 ms
BetterFilesScanner : 272 ms
ArrayBufferScanner : 220 ms
FastJavaIOScanner2 : 181 ms
FastJavaIOScanner : 179 ms
YourKit supports better-files with its full-featured Java Profiler. YourKit, LLC is the creator of YourKit Java Profiler and YourKit .NET Profiler, innovative and intelligent tools for profiling Java and .NET applications.