播放播放2.5时在运行中创建的zip文件以及带有反压的akka流 [英] stream a zip created on the fly with play 2.5 and akka stream with backpressure
问题描述
我想使用带有反压的akka流,在播放框架2.5中流式传输动态创建的zip(而不是将其完全存储在内存中).这是我的代码,带有即时创建的小zip(16KB).当客户端下载与操作相关联的url时,下载不会开始.
I would like to stream a zip created on the fly (without putting it entirely in memory) with the play-framework 2.5 using akka stream with backpressure. Here my code, with a small zip created on the fly(16KB).When the client download the url associated with the action, the download does not start.
import java.util.zip.{ ZipEntry, ZipOutputStream, GZIPOutputStream }
import akka.stream.scaladsl._
import akka.util.ByteString
import play.api.mvc._
import scala.concurrent.duration._
import java.io.{ BufferedOutputStream, ByteArrayOutputStream }
import scala.concurrent.{ Promise, Future }
import akka.stream.OverflowStrategy
class ZipController extends Controller {
def getStreamedZip = Action {
val source: Source[ByteString, java.io.OutputStream] = StreamConverters.asOutputStream()
val result = source.mapMaterializedValue(x => {
val zip = new ZipOutputStream(x)
(0 to 100).map { i =>
zip.putNextEntry(new ZipEntry("test-zip/README-" + i + ".txt"))
zip.write("This is the line:\n".map(_.toByte).toArray)
zip.closeEntry()
}
zip.close
x
})
Ok.chunked(result).withHeaders(
"Content-Type" -> "application/zip",
"Content-Disposition" -> "attachment; filename=test.zip"
)
}
}
基本上,我想在1 GB内存服务器上流式传输2GB的zip文件.并且此zip文件将包含大约15MB的文件.是否可以在不将每个文件都完全加载到内存的情况下编写zip文件?如果说3个客户端以1MB/秒的速度下载zip.这些下载大约需要多少内存?预先谢谢你.
Basically I want to stream a zip file of 2GB on a 1 GB memory server. And this zip will be composed of files of about 15MB. Is it possible to write the zip without loading entirely each file in memory ? If let say 3 clients download the zip at 1MB/second speed. Approximatively how much memory these downloads will take ? Thank you in advance.
推荐答案
Here is an implementation sourced at https://gist.github.com/kirked/412b5156f94419e71ce4a84ec1d54761
/* License: MIT */
import com.typesafe.scalalogging.slf4j.StrictLogging
import java.io.{ByteArrayOutputStream, InputStream, IOException}
import java.util.zip.{ZipEntry, ZipOutputStream}
import play.api.libs.iteratee.{Enumeratee, Enumerator}
import scala.concurrent.{Future, ExecutionContext}
/**
* Play iteratee-based reactive zip-file generation.
*/
object ZipEnumerator extends StrictLogging {
/**
* A source to zip.
*
* @param filepath The zip-file path at which to store the data.
* @param stream The data stream provider.
*/
case class Source(filepath: String, stream: () => Future[Option[InputStream]])
/**
* Given sources, returns an Enumerator that feeds a zip-file of the source contents.
*/
def apply(sources: Iterable[Source])(implicit ec: ExecutionContext): Enumerator[Array[Byte]] = {
val resolveSources: Enumerator[ResolvedSource] = Enumerator.unfoldM(sources) { sources =>
sources.headOption match {
case None => Future(None)
case Some(Source(filepath, futureStream)) =>
futureStream().map { _.map(stream => (sources.tail, ResolvedSource(filepath, stream)) ) }
}
}
val buffer = new ZipBuffer(8192)
val writeCentralDirectory = Enumerator.generateM(Future {
if (buffer.isClosed) None
else {
buffer.close
Some(buffer.bytes)
}
})
resolveSources &> zipeach(buffer) andThen writeCentralDirectory
}
private def zipeach(buffer: ZipBuffer)(implicit ec: ExecutionContext): Enumeratee[ResolvedSource, Array[Byte]] = {
Enumeratee.mapConcat[ResolvedSource] { source =>
buffer.zipStream.putNextEntry(new ZipEntry(source.filepath))
var done = false
def entryDone: Unit = {
done = true
buffer.zipStream.closeEntry
source.stream.close
}
def restOfStream: Stream[Array[Byte]] = {
if (done) Stream.empty
else {
while (!done && !buffer.full) {
try {
val byte = source.stream.read
if (byte == -1) entryDone
else buffer.zipStream.write(byte)
}
catch {
case e: IOException =>
logger.error(s"reading/zipping stream [${source.filepath}]", e)
entryDone
}
}
buffer.bytes #:: restOfStream
}
}
restOfStream
}
}
private case class ResolvedSource(filepath: String, stream: InputStream)
private class ZipBuffer(capacity: Int) {
private val buf = new ByteArrayOutputStream(capacity)
private var closed = false
val zipStream = new ZipOutputStream(buf)
def close(): Unit = {
if (!closed) {
closed = true
reset
zipStream.close // writes central directory
}
}
def isClosed = closed
def reset: Unit = buf.reset
def full: Boolean = buf.size >= capacity
def bytes: Array[Byte] = {
val result = buf.toByteArray
reset
result
}
}
}
用法看起来像这样:
val s3 = ...
val sources = items.map(item => ZipEnumerator.Source(item.filename, { () => s3.getInputStream(item.storagePath) }))
Ok.chunked(ZipEnumerator(sources))(play.api.http.Writeable.wBytes).withHeaders(
CONTENT_TYPE -> "application/zip",
CONTENT_DISPOSITION -> s"attachment; filename=MyFiles.zip; filename*=UTF-8''My%20Files.zip"
)
这篇关于播放播放2.5时在运行中创建的zip文件以及带有反压的akka流的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!