mukund-thakur commented on a change in pull request #2038:
URL: https://github.com/apache/hadoop/pull/2038#discussion_r438192115
##########
File path:
hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3AFileSystem.java
##########
@@ -4181,79 +4181,114 @@ public LocatedFileStatus next() throws IOException {
Path path = qualify(f);
LOG.debug("listFiles({}, {})", path, recursive);
try {
- // if a status was given, that is used, otherwise
- // call getFileStatus, which triggers an existence check
- final S3AFileStatus fileStatus = status != null
- ? status
- : (S3AFileStatus) getFileStatus(path);
- if (fileStatus.isFile()) {
+ // if a status was given and it is a file.
+ if (status != null && status.isFile()) {
// simple case: File
LOG.debug("Path is a file");
return new Listing.SingleStatusRemoteIterator(
- toLocatedFileStatus(fileStatus));
- } else {
- // directory: do a bulk operation
- String key = maybeAddTrailingSlash(pathToKey(path));
- String delimiter = recursive ? null : "/";
- LOG.debug("Requesting all entries under {} with delimiter '{}'",
- key, delimiter);
- final RemoteIterator<S3AFileStatus> cachedFilesIterator;
- final Set<Path> tombstones;
- boolean allowAuthoritative = allowAuthoritative(f);
- if (recursive) {
- final PathMetadata pm = metadataStore.get(path, true);
- // shouldn't need to check pm.isDeleted() because that will have
- // been caught by getFileStatus above.
- MetadataStoreListFilesIterator metadataStoreListFilesIterator =
- new MetadataStoreListFilesIterator(metadataStore, pm,
- allowAuthoritative);
- tombstones = metadataStoreListFilesIterator.listTombstones();
- // if all of the below is true
- // - authoritative access is allowed for this metadatastore for
this directory,
- // - all the directory listings are authoritative on the client
- // - the caller does not force non-authoritative access
- // return the listing without any further s3 access
- if (!forceNonAuthoritativeMS &&
- allowAuthoritative &&
- metadataStoreListFilesIterator.isRecursivelyAuthoritative()) {
- S3AFileStatus[] statuses = S3Guard.iteratorToStatuses(
- metadataStoreListFilesIterator, tombstones);
- cachedFilesIterator = listing.createProvidedFileStatusIterator(
- statuses, ACCEPT_ALL, acceptor);
- return
listing.createLocatedFileStatusIterator(cachedFilesIterator);
- }
- cachedFilesIterator = metadataStoreListFilesIterator;
- } else {
- DirListingMetadata meta =
- S3Guard.listChildrenWithTtl(metadataStore, path, ttlTimeProvider,
- allowAuthoritative);
- if (meta != null) {
- tombstones = meta.listTombstones();
- } else {
- tombstones = null;
- }
- cachedFilesIterator = listing.createProvidedFileStatusIterator(
- S3Guard.dirMetaToStatuses(meta), ACCEPT_ALL, acceptor);
- if (allowAuthoritative && meta != null && meta.isAuthoritative()) {
- // metadata listing is authoritative, so return it directly
- return
listing.createLocatedFileStatusIterator(cachedFilesIterator);
- }
+ toLocatedFileStatus(status));
+ }
+ // Assuming the path to be a directory
+ // do a bulk operation.
+ RemoteIterator<S3ALocatedFileStatus> listFilesAssumingDir =
+ getListFilesAssumingDir(path,
+ recursive,
+ acceptor,
+ collectTombstones,
+ forceNonAuthoritativeMS);
+ // If there are no list entries present, we
+ // fallback to file existence check as the path
+ // can be a file or empty directory.
+ if (!listFilesAssumingDir.hasNext()) {
+ final S3AFileStatus fileStatus = (S3AFileStatus) getFileStatus(path);
+ if (fileStatus.isFile()) {
+ return new Listing.SingleStatusRemoteIterator(
+ toLocatedFileStatus(fileStatus));
}
- return listing.createTombstoneReconcilingIterator(
- listing.createLocatedFileStatusIterator(
- listing.createFileStatusListingIterator(path,
- createListObjectsRequest(key, delimiter),
- ACCEPT_ALL,
- acceptor,
- cachedFilesIterator)),
- collectTombstones ? tombstones : null);
}
+ // If we have reached here, it means either there are files
+ // in this directory or it is empty.
+ return listFilesAssumingDir;
} catch (AmazonClientException e) {
// TODO S3Guard: retry on file not found exception
throw translateException("listFiles", path, e);
}
}
+ /**
+ * List files under a path assuming the path to be a directory.
+ * @param path input path.
+ * @param recursive recursive listing?
+ * @param acceptor file status filter
+ * @param collectTombstones should tombstones be collected from S3Guard?
+ * @param forceNonAuthoritativeMS forces metadata store to act like non
+ * authoritative. This is useful when
+ * listFiles output is used by import tool.
+ * @return an iterator over listing.
+ * @throws IOException any exception.
+ */
+ private RemoteIterator<S3ALocatedFileStatus> getListFilesAssumingDir(
+ Path path,
+ boolean recursive, Listing.FileStatusAcceptor acceptor,
+ boolean collectTombstones,
+ boolean forceNonAuthoritativeMS) throws IOException {
+
+ String key = maybeAddTrailingSlash(pathToKey(path));
+ String delimiter = recursive ? null : "/";
+ LOG.debug("Requesting all entries under {} with delimiter '{}'",
+ key, delimiter);
+ final RemoteIterator<S3AFileStatus> cachedFilesIterator;
+ final Set<Path> tombstones;
+ boolean allowAuthoritative = allowAuthoritative(path);
+ if (recursive) {
+ final PathMetadata pm = metadataStore.get(path, true);
+ // shouldn't need to check pm.isDeleted() because that will have
Review comment:
Should we take care of this as now we are not calling getFileStatus() in
the start?? @steveloughran
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
