Github user zsxwing commented on a diff in the pull request:
https://github.com/apache/spark/pull/11925#discussion_r57406100
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/HDFSMetadataLog.scala
---
@@ -196,4 +195,148 @@ class HDFSMetadataLog[T: ClassTag](sqlContext:
SQLContext, path: String)
}
None
}
+
+ private def createFileManager(): FileManager = {
+ val hadoopConf = sqlContext.sparkContext.hadoopConfiguration
+ try {
+ new FileContextManager(metadataPath, hadoopConf)
+ } catch {
+ case e: UnsupportedFileSystemException =>
+ logWarning("Could not use FileContext API for managing metadata
log file. The log may be" +
+ "inconsistent under failures.", e)
+ new FileSystemManager(metadataPath, hadoopConf)
+ }
+ }
+}
+
+object HDFSMetadataLog {
+
+ /** A simple trait to abstract out the file management operations needed
by HDFSMetadataLog. */
+ trait FileManager {
+
+ /** List the files in a path that matches a filter. */
+ def list(path: Path, filter: PathFilter): Array[FileStatus]
+
+ /** Make directory at the give path and all its parent directories as
needed. */
+ def mkdirs(path: Path): Unit
+
+ /** Whether path exists */
+ def exists(path: Path): Boolean
+
+ /** Open a file for reading, or throw exception if it does not exist.
*/
+ def open(path: Path): FSDataInputStream
+
+ /** Create path, or throw exception if it already exists */
+ def create(path: Path): FSDataOutputStream
+
+ /**
+ * Atomically ename path, or throw exception if it cannot be done.
+ * Should throw FileAlreadyExistsException if file already exists.
+ * Should throw FileNotFound exception if the file does not exist.
+ */
+ def rename(srcPath: Path, destPath: Path): Unit
+
+ /** Recursively delete a path if it exists. Should not throw exception
if file doesn't exist. */
+ def delete(path: Path): Unit
+ }
+
+ /**
+ * Default implementation of FileManager using newer FileContext API.
+ */
+ class FileContextManager(path: Path, hadoopConf: Configuration) extends
FileManager {
+ private val fc = if (path.toUri.getScheme == null) {
+ FileContext.getFileContext(hadoopConf)
+ } else {
+ FileContext.getFileContext(path.toUri, hadoopConf)
+ }
+
+ override def list(path: Path, filter: PathFilter): Array[FileStatus] =
{
+ fc.util.listStatus(path, filter)
+ }
+
+ override def rename(srcPath: Path, destPath: Path): Unit = {
+ fc.rename(srcPath, destPath)
+ }
+
+ override def mkdirs(path: Path): Unit = {
+ fc.mkdir(path, FsPermission.getDirDefault, true)
+ }
+
+ override def open(path: Path): FSDataInputStream = {
+ fc.open(path)
+ }
+
+ override def create(path: Path): FSDataOutputStream = {
+ fc.create(path, EnumSet.of(CreateFlag.CREATE))
+ }
+
+ override def exists(path: Path): Boolean = {
+ fc.util().exists(path)
+ }
+
+ override def delete(path: Path): Unit = {
+ try {
+ fc.delete(path, true)
+ } catch {
+ case e: FileNotFoundException =>
+ // ignore if file has already been deleted
+ }
+ }
+ }
+
+ /**
+ * Implementation of FileManager using older FileSystem API. Note that
this implementation
+ * cannot provide atomic renaming of paths, hence can lead to
consistency issues. This
+ * should be used only as a backup option, when FileContextManager
cannot be used.
+ */
+ class FileSystemManager(path: Path, hadoopConf: Configuration) extends
FileManager {
+ private val fs = path.getFileSystem(hadoopConf)
+
+ override def list(path: Path, filter: PathFilter): Array[FileStatus] =
{
+ fs.listStatus(path, filter)
+ }
+
+ /**
+ * Rename a path. Note that this implementation is not atomic.
+ * @throws FileNotFoundException if source path does not exist.
+ * @throws FileAlreadyExistsException if destination path already
exists.
+ * @throws IOException if renaming fails for some unknown reason.
+ */
+ override def rename(srcPath: Path, destPath: Path): Unit = {
+ if (!fs.exists(srcPath)) {
+ throw new FileNotFoundException(s"Source path already exists:
$srcPath")
--- End diff --
nit: Source path already exists -> Cannot find ...
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
