bool MaprInputCodedBlockFile::Open(std::string uri) {
CHECK(!is_open_) << "File already open.";
if (!IsValidUri(uri, "maprfs")) {
LOG(ERROR) << "failed to validate uri: " << uri;
return false;
}
std::string scheme, path;
CHECK(ParseUri(uri, &scheme, &path)) << "Invalid uri format: " << uri;
file_ = hdfsOpenFile(fs_, path.c_str(), O_RDONLY, 0, 0, 0);
if (file_ == NULL) {
LOG(ERROR) << "Failed to open file: " << path;
return false;
}
is_open_ = true;
path_ = path;
// Cache file size
hdfsFileInfo* info = hdfsGetPathInfo(fs_, path_.c_str());
CHECK(info);
size_ = info->mSize;
hdfsFreeFileInfo(info, 1);
return true;
}
block_id_counter FileManagerHdfs::getMaxUsedBlockCounter(const block_id_domain block_domain) const {
int num_files = 0;
hdfsFileInfo *file_infos = hdfsListDirectory(hdfs_, storage_path_.c_str(), &num_files);
if (file_infos == nullptr) {
if (errno != ENOENT) {
LOG_WARNING("Failed to list file info with error: " << strerror(errno));
}
return 0;
}
string filename_pattern("/qsblk_");
filename_pattern.append(ToZeroPaddedString(block_domain, kBlockIdDomainLengthInDigits));
filename_pattern.append("_%");
filename_pattern.append(SCNu64);
filename_pattern.append(".qsb");
block_id_counter counter_max = 0, counter;
for (int i = 0; i < num_files; ++i) {
// NOTE(zuyu): mName looks like
// "/user/<username>/<storage_path_>/qsblk_<block_domain>_[0-9]*.qsb".
const char *filename = std::strrchr(file_infos[i].mName, '/');
if (filename != nullptr
&& sscanf(filename, filename_pattern.c_str(), &counter) == 1
&& counter > counter_max) {
counter_max = counter;
}
}
hdfsFreeFileInfo(file_infos, num_files);
return counter_max;
}
off64_t HdfsFile::size()
{
off64_t ret = 0;
int savedErrno;
if( ( m_flags & O_RDONLY ) != 0 )
{
hdfsFileInfo* fileinfo;
fileinfo = hdfsGetPathInfo(m_fs,m_fname.c_str());
ret = (fileinfo ? fileinfo->mSize : -1);
if( fileinfo )
hdfsFreeFileInfo(fileinfo,1);
}
else
{
// if file is open for either WRITE or APPEND then we know that
// size is always the current file offset since HDFS can only
// write at the end
ret = tell();
}
savedErrno = errno;
if( IDBLogger::isEnabled() )
IDBLogger::logSize(m_fname, this, ret);
errno = savedErrno;
return ret;
}
bool MaprFileSystem::ListDirectory(const std::string& uri, std::vector<std::string>* contents){
CHECK(contents);
contents->clear();
std::string path = GetUriPathOrDie(uri);
std::string host = "default";
hdfsFS fs = hdfsConnect(host.c_str(), 0); // use default config file settings
int num_entries;
hdfsFileInfo* entries = hdfsListDirectory(fs, path.c_str(), &num_entries);
hdfsFileInfo* cur_entry = entries;
for (int i=0; i < num_entries; ++i) {
// Sometimes the list directory command returns paths with the scheme and sometimes it doesn't
// Strange.
// Anyway, we need to consistently output uris with a proper scheme prefix.
std::string cur_scheme, cur_path, error;
if (ParseUri(cur_entry->mName, &cur_scheme, &cur_path, &error)){
CHECK_EQ(cur_scheme, "maprfs"); // if it has a scheme prefix, make sure it is maprfs as expected
}
else{
// this doesn't have a uri scheme prefix, so assume it is just the path portion
cur_path = cur_entry->mName;
}
contents->push_back(Uri("maprfs", cur_path));
cur_entry++;
}
hdfsFreeFileInfo(entries, num_entries);
CHECK_EQ(hdfsDisconnect(fs), 0);
return true;
}
bs_file_info_t *hdfs_list_dir(struct back_storage *storage, \
const char *dir_path, uint32_t *num_entries){
HLOG_DEBUG("hdfs -- enter func %s", __func__);
char full_path[256];
build_hdfs_path(full_path, storage->dir, storage->fs_name, dir_path);
int num;
hdfsFileInfo *hinfos = \
hdfsListDirectory((hdfsFS)storage->fs_handler, full_path, &num);
if (NULL == hinfos) {
//HLOG_ERROR("hdfsListDirectory error");
return NULL;
}
hdfsFileInfo *hinfo = hinfos;
bs_file_info_t *infos = \
(bs_file_info_t*)g_malloc0(sizeof(bs_file_info_t)*8192);
if (NULL == infos) {
//HLOG_ERROR("Allocate Error!");
return NULL;
}
bs_file_info_t *info = infos;
int i;
for (i = 0;i < num;i++) {
strcpy((char *)info->name, \
(const char *)g_path_get_basename(hinfo->mName));
info->is_dir = 0;
info->size = hinfo->mSize;
info->lmtime = hinfo->mLastMod;
info++;
hinfo++;
}
hdfsFreeFileInfo(hinfos, num);
*num_entries = num;
HLOG_DEBUG("hdfs -- leave func %s", __func__);
return infos;
}
HDFSChunkReaderIterator::HDFSChunkReaderIterator(const ChunkID& chunk_id, unsigned& chunk_size,const unsigned& block_size)
:ChunkReaderIterator(chunk_id,block_size,chunk_size){
block_buffer_=new Block(block_size_);
fs_=hdfsConnect(Config::hdfs_master_ip.c_str(),Config::hdfs_master_port);
hdfs_fd_=hdfsOpenFile(fs_,chunk_id.partition_id.getName().c_str(),O_RDONLY,0,0,0);
if(!hdfs_fd_){
printf("fails to open HDFS file [%s]\n",chunk_id.partition_id.getName().c_str());
number_of_blocks_=0;
}
const unsigned start_pos=start_pos+CHUNK_SIZE*chunk_id_.chunk_off;
if(hdfsSeek(fs_,hdfs_fd_,start_pos)==-1){
printf("fails to set the start offset %d for [%s]\n",start_pos,chunk_id.partition_id.getName().c_str());
number_of_blocks_=0;
}
hdfsFileInfo *file_info=hdfsGetPathInfo(fs_,"/imdb/");// to be refined after communicating with Zhang Lei
if(start_pos+CHUNK_SIZE<file_info->mSize){
number_of_blocks_=CHUNK_SIZE/block_size_;
}
else{
number_of_blocks_=(file_info->mSize-start_pos)/block_size_;
}
hdfsFreeFileInfo(file_info,1);
}
int HdfsFileSystem::listDirectory(const char* pathname, std::list<std::string>& contents) const
{
// clear the return list
contents.erase( contents.begin(), contents.end() );
int numEntries;
hdfsFileInfo* fileinfo;
if( !exists( pathname ) ) {
errno = ENOENT;
return -1;
}
// hdfs not happy if you call list directory on a path that does not exist
fileinfo = hdfsListDirectory(m_fs,pathname, &numEntries);
for( int i = 0; i < numEntries && fileinfo; ++i )
{
// hdfs returns a fully specified path name but we want to
// only return paths relative to the directory passed in.
boost::filesystem::path filepath( fileinfo[i].mName );
contents.push_back( filepath.filename().c_str() );
}
if( fileinfo )
hdfsFreeFileInfo(fileinfo, numEntries);
return 0;
}
void dir_cpi_impl::sync_is_dir (bool & is_dir,
saga::url url)
{
instance_data idata (this);
is_dir = false;
saga::url dir_url(idata->location_);
boost::filesystem::path name (url.get_path(), boost::filesystem::native);
boost::filesystem::path path (idata->location_.get_path(), boost::filesystem::native);
if ( ! name.has_root_path () )
path /= name;
else
path = name;
if(hdfsExists(fs_, path.string().c_str()) == 0)
{
//Check to see if it is a directory
hdfsFileInfo *info;
instance_data idata(this);
info = hdfsGetPathInfo(fs_, path.string().c_str());
if(info == NULL)
{
SAGA_ADAPTOR_THROW("file_cpi_impl::init failed",
saga::NoSuccess);
}
if(info->mKind == kObjectKindDirectory)
is_dir = true;
hdfsFreeFileInfo(info, 1);
}
}
int dfs_getattr(const char *path, struct stat *st)
{
TRACE1("getattr", path)
// retrieve dfs specific data
dfs_context *dfs = (dfs_context*)fuse_get_context()->private_data;
// check params and the context var
assert(dfs);
assert(path);
assert(st);
// if not connected, try to connect and fail out if we can't.
if (NULL == dfs->fs && NULL == (dfs->fs = hdfsConnect(dfs->nn_hostname,dfs->nn_port))) {
syslog(LOG_ERR, "ERROR: could not connect to %s:%d %s:%d\n", dfs->nn_hostname, dfs->nn_port,__FILE__, __LINE__);
return -EIO;
}
// call the dfs API to get the actual information
hdfsFileInfo *info = hdfsGetPathInfo(dfs->fs,path);
if (NULL == info) {
return -ENOENT;
}
fill_stat_structure(&info[0], st);
// setup hard link info - for a file it is 1 else num entries in a dir + 2 (for . and ..)
if (info[0].mKind == kObjectKindDirectory) {
int numEntries = 0;
hdfsFileInfo *info = hdfsListDirectory(dfs->fs,path,&numEntries);
if (info) {
hdfsFreeFileInfo(info,numEntries);
}
st->st_nlink = numEntries + 2;
} else {
// not a directory
st->st_nlink = 1;
}
// free the info pointer
hdfsFreeFileInfo(info,1);
return 0;
}
bool HdfsFileSystem::isDir(const char* pathname) const
{
hdfsFileInfo* fileinfo;
fileinfo = hdfsGetPathInfo(m_fs,pathname);
bool retval = (fileinfo ? fileinfo->mKind == kObjectKindDirectory : false);
if( fileinfo )
hdfsFreeFileInfo(fileinfo,1);
return retval;
}
static struct libhdfs_data *libhdfs_data_create(const struct options *opts)
{
struct libhdfs_data *ldata = NULL;
struct hdfsBuilder *builder = NULL;
hdfsFileInfo *pinfo = NULL;
ldata = calloc(1, sizeof(struct libhdfs_data));
if (!ldata) {
fprintf(stderr, "Failed to allocate libhdfs test data.\n");
goto error;
}
builder = hdfsNewBuilder();
if (!builder) {
fprintf(stderr, "Failed to create builder.\n");
goto error;
}
hdfsBuilderSetNameNode(builder, opts->rpc_address);
hdfsBuilderConfSetStr(builder,
"dfs.client.read.shortcircuit.skip.checksum", "true");
ldata->fs = hdfsBuilderConnect(builder);
if (!ldata->fs) {
fprintf(stderr, "Could not connect to default namenode!\n");
goto error;
}
pinfo = hdfsGetPathInfo(ldata->fs, opts->path);
if (!pinfo) {
int err = errno;
fprintf(stderr, "hdfsGetPathInfo(%s) failed: error %d (%s). "
"Attempting to re-create file.\n",
opts->path, err, strerror(err));
if (libhdfs_data_create_file(ldata, opts))
goto error;
} else if (pinfo->mSize != opts->length) {
fprintf(stderr, "hdfsGetPathInfo(%s) failed: length was %lld, "
"but we want length %lld. Attempting to re-create file.\n",
opts->path, (long long)pinfo->mSize, (long long)opts->length);
if (libhdfs_data_create_file(ldata, opts))
goto error;
}
ldata->file = hdfsOpenFile(ldata->fs, opts->path, O_RDONLY, 0, 0, 0);
if (!ldata->file) {
int err = errno;
fprintf(stderr, "hdfsOpenFile(%s) failed: error %d (%s)\n",
opts->path, err, strerror(err));
goto error;
}
ldata->length = opts->length;
return ldata;
error:
if (pinfo)
hdfsFreeFileInfo(pinfo, 1);
if (ldata)
libhdfs_data_free(ldata);
return NULL;
}
void Hdfs3Glob(const std::string& _path, const GlobType& gtype,
FileList& filelist) {
std::string path = _path;
// crop off hdfs://
die_unless(common::StartsWith(path, "hdfs://"));
path = path.substr(7);
// split uri into host/path
std::vector<std::string> splitted = common::Split(path, '/', 2);
hdfsFS fs = Hdfs3FindConnection(splitted[0]);
std::string hosturi = "hdfs://" + splitted[0];
// prepend root /
splitted[1] = "/" + splitted[1];
// list directory
int num_entries = 0;
hdfsFileInfo* list = hdfsListDirectory(
fs, splitted[1].c_str(), &num_entries);
if (!list) return;
for (int i = 0; i < num_entries; ++i) {
FileInfo fi;
fi.path = list[i].mName;
// remove leading slashes
while (fi.path.size() >= 2 && fi.path[0] == '/' && fi.path[1] == '/')
fi.path.erase(fi.path.begin(), fi.path.begin() + 1);
// prepend host uri
fi.path = hosturi + fi.path;
if (list[i].mKind == kObjectKindFile) {
if (gtype == GlobType::All || gtype == GlobType::File) {
// strangely full file name globs return the file with a / at
// the end.
while (fi.path.back() == '/')
fi.path.resize(fi.path.size() - 1);
fi.type = Type::File;
fi.size = list[i].mSize;
filelist.emplace_back(fi);
}
}
else if (list[i].mKind == kObjectKindDirectory) {
if (gtype == GlobType::All || gtype == GlobType::Directory) {
fi.type = Type::Directory;
fi.size = list[i].mSize;
filelist.emplace_back(fi);
}
}
}
hdfsFreeFileInfo(list, num_entries);
}
int hdfsExists(hdfsFS fs, const char *path)
{
hdfsFileInfo *fileInfo = hdfsGetPathInfoImpl(fs, path, 0);
if (!fileInfo) {
// (errno will have been set by hdfsGetPathInfo)
return -1;
}
hdfsFreeFileInfo(fileInfo, 1);
return 0;
}
NABoolean HHDFSListPartitionStats::populate(hdfsFS fs,
const NAString &dir,
Int32 numOfBuckets,
NABoolean doEstimation,
char recordTerminator,
NABoolean isSequenceFile)
{
NABoolean result = TRUE;
int numFiles = 0;
// remember parameters
partitionDir_ = dir;
defaultBucketIdx_ = (numOfBuckets >= 1) ? numOfBuckets : 0;
doEstimation_ = doEstimation;
recordTerminator_ = recordTerminator;
isSequenceFile_ = isSequenceFile;
// list all the files in this directory, they all belong
// to this partition and either belong to a specific bucket
// or to the default bucket
hdfsFileInfo *fileInfos = hdfsListDirectory(fs,
dir.data(),
&numFiles);
// populate partition stats
for (int f=0; f<numFiles && result; f++)
if (fileInfos[f].mKind == kObjectKindFile)
{
// the default (unbucketed) bucket number is
// defaultBucketIdx_
Int32 bucketNum = determineBucketNum(fileInfos[f].mName);
HHDFSBucketStats *bucketStats = NULL;
if (! bucketStatsList_.used(bucketNum))
{
bucketStats = new(heap_) HHDFSBucketStats(heap_);
bucketStatsList_.insertAt(bucketNum, bucketStats);
}
else
bucketStats = bucketStatsList_[bucketNum];
if (! bucketStats->addFile(fs, &fileInfos[f], doEstimation, recordTerminator, isSequenceFile))
result = FALSE;
}
hdfsFreeFileInfo(fileInfos, numFiles);
// aggregate statistics over all buckets
for (Int32 b=0; b<=defaultBucketIdx_; b++)
if (bucketStatsList_.used(b))
add(bucketStatsList_[b]);
return result;
}
inline std::vector<std::string> list_files(const std::string& path) {
int num_files = 0;
hdfsFileInfo* hdfs_file_list_ptr =
hdfsListDirectory(filesystem, path.c_str(), &num_files);
// copy the file list to the string array
std::vector<std::string> files(num_files);
for(int i = 0; i < num_files; ++i)
files[i] = std::string(hdfs_file_list_ptr[i].mName);
// free the file list pointer
hdfsFreeFileInfo(hdfs_file_list_ptr, num_files);
return files;
} // end of list_files
/**
* call-seq:
* hdfs.stat(path) -> file_info
*
* Stats the file or directory at the supplied path, returning a
* Hadoop::DFS::FileInfo object corresponding to it. If this fails, raises a
* DFSException.
*/
VALUE HDFS_File_System_stat(VALUE self, VALUE path) {
FSData* data = get_FSData(self);
hdfsFileInfo* info = hdfsGetPathInfo(data->fs, StringValuePtr(path));
if (info == NULL) {
rb_raise(e_dfs_exception, "Failed to stat file %s: %s",
StringValuePtr(path), get_error(errno));
return Qnil;
}
VALUE file_info = new_HDFS_File_Info(info);
hdfsFreeFileInfo(info, 1);
return file_info;
}
time_t HdfsFile::mtime()
{
boost::mutex::scoped_lock lock(m_mutex);
time_t ret = 0;
hdfsFileInfo* fileinfo;
fileinfo = hdfsGetPathInfo(m_fs,m_fname.c_str());
ret = (fileinfo ? fileinfo->mLastMod : -1);
if( fileinfo )
hdfsFreeFileInfo(fileinfo,1);
return ret;
}
off64_t HdfsFileSystem::size(const char* path) const
{
hdfsFileInfo* fileinfo;
fileinfo = hdfsGetPathInfo(m_fs,path);
off64_t retval = (fileinfo ? fileinfo->mSize : -1);
if( fileinfo )
hdfsFreeFileInfo(fileinfo,1);
if( IDBLogger::isEnabled() )
IDBLogger::logFSop( HDFS, "fs:size", path, this, retval);
return retval;
}
int dfs_rmdir(const char *path)
{
struct hdfsConn *conn = NULL;
hdfsFS fs;
int ret;
dfs_context *dfs = (dfs_context*)fuse_get_context()->private_data;
int numEntries = 0;
hdfsFileInfo *info = NULL;
TRACE1("rmdir", path)
assert(path);
assert(dfs);
assert('/' == *path);
if (is_protected(path)) {
ERROR("Trying to delete protected directory %s", path);
ret = -EACCES;
goto cleanup;
}
ret = fuseConnectAsThreadUid(&conn);
if (ret) {
fprintf(stderr, "fuseConnectAsThreadUid: failed to open a libhdfs "
"connection! error %d.\n", ret);
ret = -EIO;
goto cleanup;
}
fs = hdfsConnGetFs(conn);
info = hdfsListDirectory(fs, path, &numEntries);
if (numEntries) {
ret = -ENOTEMPTY;
goto cleanup;
}
if (hdfsDeleteWithTrash(fs, path, dfs->usetrash)) {
ERROR("Error trying to delete directory %s", path);
ret = -EIO;
goto cleanup;
}
ret = 0;
cleanup:
if (info) {
hdfsFreeFileInfo(info, numEntries);
}
if (conn) {
hdfsConnRelease(conn);
}
return ret;
}
void dir_cpi_impl::sync_remove (saga::impl::void_t & ret,
saga::url url,
int flags)
{
instance_data idata (this);
saga::url dir_url(idata->location_);
boost::filesystem::path src_location (idata->location_.get_path(), boost::filesystem::native);
// complete paths
boost::filesystem::path src_path (url.get_path(), boost::filesystem::native);
if ( ! src_path.has_root_path () )
src_location /= src_path;
else
src_location = src_path;
bool is_src_dir = false;
if(hdfsExists(fs_, src_location.string().c_str()) != 0)
{
SAGA_ADAPTOR_THROW("directory::remove: Can't remove directory: "
"Does not exist", saga::DoesNotExist);
}
else
{
hdfsFileInfo *info;
info = hdfsGetPathInfo(fs_, src_location.string().c_str());
if(info == NULL)
{
SAGA_ADAPTOR_THROW("file_cpi_impl::init failed",
saga::NoSuccess);
}
if(info->mKind == kObjectKindDirectory)
is_src_dir = true;
else
is_src_dir = false;
hdfsFreeFileInfo(info, 1);
}
if (is_src_dir) {
if (saga::name_space::Recursive != flags)
{
SAGA_ADAPTOR_THROW("directory::remove: Can't remove directory. "
"Please use recursive mode!", saga::BadParameter);
}
else {
saga_hdfs_delete(fs_, src_location.string().c_str());
}
}
else {
saga_hdfs_delete(fs_, src_location.string().c_str());
}
}
bool MaprFileSystem::IsDirectory(const std::string& uri){
std::string path = GetUriPathOrDie(uri);
std::string host = "default";
hdfsFS fs = hdfsConnect(host.c_str(), 0); // use default config file settings
CHECK(fs);
hdfsFileInfo* info = hdfsGetPathInfo(fs, path.c_str());
bool is_directory = false;
if (info){
is_directory = (info->mKind == kObjectKindDirectory);
hdfsFreeFileInfo(info,1);
}
else{
LOG(FATAL) << "uri does not exist: " << uri;
}
CHECK_EQ(hdfsDisconnect(fs), 0);
return is_directory;
}
int libhdfsconnector::streamInFile(const char * rfile, int bufferSize)
{
if (!fs)
{
fprintf(stderr, "Could not connect to hdfs on");
return RETURN_FAILURE;
}
unsigned long fileTotalSize = 0;
hdfsFileInfo *fileInfo = NULL;
if ((fileInfo = hdfsGetPathInfo(fs, rfile)) != NULL)
{
fileTotalSize = fileInfo->mSize;
hdfsFreeFileInfo(fileInfo, 1);
}
else
{
fprintf(stderr, "Error: hdfsGetPathInfo for %s - FAILED!\n", rfile);
return RETURN_FAILURE;
}
hdfsFile readFile = hdfsOpenFile(fs, rfile, O_RDONLY, bufferSize, 0, 0);
if (!readFile)
{
fprintf(stderr, "Failed to open %s for writing!\n", rfile);
return RETURN_FAILURE;
}
unsigned char buff[bufferSize + 1];
buff[bufferSize] = '\0';
for (unsigned long bytes_read = 0; bytes_read < fileTotalSize;)
{
unsigned long read_length = hdfsRead(fs, readFile, buff, bufferSize);
bytes_read += read_length;
for (unsigned long i = 0; i < read_length; i++)
fprintf(stdout, "%c", buff[i]);
}
hdfsCloseFile(fs, readFile);
return 0;
}
int dfsList(const char* path){
hdfsFS fs = hdfsConnect("default", 0);
int i, entries;
hdfsFileInfo *files, *head;
/* Get the list info */
files = hdfsListDirectory(fs, path, &entries);
if (!files){
perror("Get directory info");
exit(-1);
}
head = files;
/* Print the info */
fprintf(stdout, "%s %-50s %-9s %s\n",
"Kind", "Name", "Size", "Replicas");
for (i = 0; i < entries; i++){
const char* unit[] = {" B", "KB", "MB", "GB", "TB", "PB"};
double size = files->mSize;
unsigned int u = 0;
while (size > 1024){
u++;
size /= 1024;
}
assert(u < 6);
fprintf(stdout, "%4c %-50s %-7.2lf%s %8d\n",
files->mKind, files->mName,
size, unit[u],
files->mReplication);
files += 1;
}
/* List ends */
hdfsFreeFileInfo(head, entries);
hdfsDisconnect(fs);
return 0;
}
/**
* call-seq:
* hdfs.ls(path) -> file_infos
*
* Lists the directory at the supplied path, returning an Array of
* HDFS::FileInfo objects. If this fails, raises a DFSException.
*/
VALUE HDFS_File_System_ls(VALUE self, VALUE path) {
FSData* data = get_FSData(self);
VALUE file_infos = rb_ary_new();
int num_files = -1;
hdfsFileInfo* infos = hdfsListDirectory(data->fs, StringValuePtr(path),
&num_files);
if (infos == NULL && num_files == -1) {
rb_raise(e_dfs_exception, "Failed to list directory %s: %s",
StringValuePtr(path), get_error(errno));
return Qnil;
}
int i;
for (i = 0; i < num_files; i++) {
hdfsFileInfo* cur_info = infos + i;
rb_ary_push(file_infos, new_HDFS_File_Info(cur_info));
}
hdfsFreeFileInfo(infos, num_files);
return file_infos;
}
size_t FileManagerHdfs::numSlots(const block_id block) const {
string filename(blockFilename(block));
hdfsFileInfo *file_info = hdfsGetPathInfo(hdfs_, filename.c_str());
if (file_info == nullptr) {
if (errno != ENOENT) {
LOG_WARNING("Failed to get size of file " << filename << " with error: " << strerror(errno));
}
return 0;
}
size_t file_size = file_info->mSize;
hdfsFreeFileInfo(file_info, 1);
if ((file_size % kSlotSizeBytes) != 0) {
throw CorruptPersistentStorage();
}
return file_size / kSlotSizeBytes;
}
FileStatus Hdfs::getFileStatus(string path)
{
hdfsFileInfo* fi = hdfsGetPathInfo(_getFs(), path.data());
if (fi == NULL)
{
throw ios_base::failure("Error retrieving file status. (" + path + ")");
}
FileStatus result;
result._setAccessTime(fi->mLastAccess);
result._setBlockSize(fi->mBlockSize);
result._setModificationTime(fi->mLastMod);
result._setLen(fi->mSize);
result._setKind(fi->mKind);
result._setPath(fi->mName);
hdfsFreeFileInfo(fi, 1);
return result;
}
int hdfsSeek(hdfsFS fs, hdfsFile file, tOffset desiredPos)
{
struct webhdfsFileHandle *wf;
hdfsFileInfo *fileInfo = NULL;
int ret = 0;
if (!fs || !file || (file->type == OUTPUT) || (desiredPos < 0)) {
ret = EINVAL;
goto done;
}
wf = file->file;
if (!wf) {
ret = EINVAL;
goto done;
}
fileInfo = hdfsGetPathInfo(fs, wf->absPath);
if (!fileInfo) {
ret = errno;
goto done;
}
if (desiredPos > fileInfo->mSize) {
fprintf(stderr,
"hdfsSeek for %s failed since the desired position %" PRId64
" is beyond the size of the file %" PRId64 "\n",
wf->absPath, desiredPos, fileInfo->mSize);
ret = ENOTSUP;
goto done;
}
file->offset = desiredPos;
done:
if (fileInfo) {
hdfsFreeFileInfo(fileInfo, 1);
}
if (ret) {
errno = ret;
return -1;
}
return 0;
}
std::vector<FileStatus> Hdfs::listStatus(string path, const bool sortByPath)
{
if (exists(path) == false)
{
throw ios_base::failure("Error retrieving status on non-existant path (" + path + ")");
}
std::vector<FileStatus> result;
int numEntries;
hdfsFileInfo* fis = hdfsListDirectory(_getFs(), path.data(), &numEntries);
if (fis == NULL)
{
throw ios_base::failure("Error listing directory contents. (" + path + ")");
}
result.resize(numEntries);
for (int i = 0; i < numEntries; i++)
{
hdfsFileInfo* fi = fis + i;
FileStatus& r = result[i];
r._setAccessTime(fi->mLastAccess);
r._setModificationTime(fi->mLastMod);
r._setLen(fi->mSize);
r._setKind(fi->mKind);
r._setPath(fi->mName);
}
hdfsFreeFileInfo(fis, numEntries);
if (sortByPath)
{
sort(result.begin(), result.end(), _fileStatusPathCompare);
}
return result;
}
long libhdfsconnector::getFileSize(const char * filename)
{
if (!fs)
{
fprintf(stderr, "Could not connect to hdfs");
return RETURN_FAILURE;
}
hdfsFileInfo *fileInfo = NULL;
if ((fileInfo = hdfsGetPathInfo(fs, filename)) != NULL)
{
long fsize = fileInfo->mSize;
hdfsFreeFileInfo(fileInfo, 1);
return fsize;
}
else
{
fprintf(stderr, "Error: hdfsGetPathInfo for %s - FAILED!\n", filename);
return RETURN_FAILURE;
}
return RETURN_FAILURE;
}
int main(int argc, char **argv) {
hdfsFS fs = hdfsConnect("default", 0);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs!\n");
exit(-1);
}
hdfsFS lfs = hdfsConnect(NULL, 0);
if(!lfs) {
fprintf(stderr, "Oops! Failed to connect to 'local' hdfs!\n");
exit(-1);
}
const char* writePath = "/tmp/testfile.txt";
{
//Write tests
hdfsFile writeFile = hdfsOpenFile(fs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", writePath);
char* buffer = "Hello, World!";
tSize num_written_bytes = hdfsWrite(fs, writeFile, (void*)buffer, strlen(buffer)+1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
tOffset currentPos = -1;
if ((currentPos = hdfsTell(fs, writeFile)) == -1) {
fprintf(stderr,
"Failed to get current file position correctly! Got %ld!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %ld\n", currentPos);
if (hdfsFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", writePath);
hdfsCloseFile(fs, writeFile);
}
{
//Read tests
const char* readPath = "/tmp/testfile.txt";
int exists = hdfsExists(fs, readPath);
if (exists) {
fprintf(stderr, "Failed to validate existence of %s\n", readPath);
exit(-1);
}
hdfsFile readFile = hdfsOpenFile(fs, readPath, O_RDONLY, 0, 0, 0);
if (!readFile) {
fprintf(stderr, "Failed to open %s for reading!\n", readPath);
exit(-1);
}
fprintf(stderr, "hdfsAvailable: %d\n", hdfsAvailable(fs, readFile));
tOffset seekPos = 1;
if(hdfsSeek(fs, readFile, seekPos)) {
fprintf(stderr, "Failed to seek %s for reading!\n", readPath);
exit(-1);
}
tOffset currentPos = -1;
if((currentPos = hdfsTell(fs, readFile)) != seekPos) {
fprintf(stderr,
"Failed to get current file position correctly! Got %ld!\n",
currentPos);
exit(-1);
}
fprintf(stderr, "Current position: %ld\n", currentPos);
static char buffer[32];
tSize num_read_bytes = hdfsRead(fs, readFile, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
num_read_bytes = hdfsPread(fs, readFile, 0, (void*)buffer,
sizeof(buffer));
fprintf(stderr, "Read following %d bytes:\n%s\n",
num_read_bytes, buffer);
hdfsCloseFile(fs, readFile);
}
int totalResult = 0;
int result = 0;
{
//Generic file-system operations
const char* srcPath = "/tmp/testfile.txt";
const char* dstPath = "/tmp/testfile2.txt";
fprintf(stderr, "hdfsCopy(remote-local): %s\n", ((result = hdfsCopy(fs, srcPath, lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(local-local): %s\n", ((result = hdfsMove(lfs, srcPath, lfs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsMove(remote-local): %s\n", ((result = hdfsMove(fs, srcPath, lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsRename: %s\n", ((result = hdfsRename(fs, dstPath, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsCopy(remote-remote): %s\n", ((result = hdfsCopy(fs, srcPath, fs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
const char* slashTmp = "/tmp";
const char* newDirectory = "/tmp/newdir";
fprintf(stderr, "hdfsCreateDirectory: %s\n", ((result = hdfsCreateDirectory(fs, newDirectory)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsSetReplication: %s\n", ((result = hdfsSetReplication(fs, srcPath, 2)) ? "Failed!" : "Success!"));
totalResult += result;
char buffer[256];
const char *resp;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ? buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsSetWorkingDirectory: %s\n", ((result = hdfsSetWorkingDirectory(fs, slashTmp)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsGetWorkingDirectory: %s\n", ((resp = hdfsGetWorkingDirectory(fs, buffer, sizeof(buffer))) ? buffer : "Failed!"));
totalResult += (resp ? 0 : 1);
fprintf(stderr, "hdfsGetDefaultBlockSize: %ld\n", hdfsGetDefaultBlockSize(fs));
fprintf(stderr, "hdfsGetCapacity: %ld\n", hdfsGetCapacity(fs));
fprintf(stderr, "hdfsGetUsed: %ld\n", hdfsGetUsed(fs));
hdfsFileInfo *fileInfo = NULL;
if((fileInfo = hdfsGetPathInfo(fs, slashTmp)) != NULL) {
fprintf(stderr, "hdfsGetPathInfo - SUCCESS!\n");
fprintf(stderr, "Name: %s, ", fileInfo->mName);
fprintf(stderr, "Type: %c, ", (char)(fileInfo->mKind));
fprintf(stderr, "Replication: %d, ", fileInfo->mReplication);
fprintf(stderr, "BlockSize: %ld, ", fileInfo->mBlockSize);
fprintf(stderr, "Size: %ld, ", fileInfo->mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileInfo->mLastMod));
fprintf(stderr, "Owner: %s, ", fileInfo->mOwner);
fprintf(stderr, "Group: %s, ", fileInfo->mGroup);
char permissions[10];
permission_disp(fileInfo->mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileInfo->mPermissions, permissions);
hdfsFreeFileInfo(fileInfo, 1);
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetPathInfo for %s - FAILED!\n", slashTmp);
}
hdfsFileInfo *fileList = 0;
int numEntries = 0;
if((fileList = hdfsListDirectory(fs, slashTmp, &numEntries)) != NULL) {
int i = 0;
for(i=0; i < numEntries; ++i) {
fprintf(stderr, "Name: %s, ", fileList[i].mName);
fprintf(stderr, "Type: %c, ", (char)fileList[i].mKind);
fprintf(stderr, "Replication: %d, ", fileList[i].mReplication);
fprintf(stderr, "BlockSize: %ld, ", fileList[i].mBlockSize);
fprintf(stderr, "Size: %ld, ", fileList[i].mSize);
fprintf(stderr, "LastMod: %s", ctime(&fileList[i].mLastMod));
fprintf(stderr, "Owner: %s, ", fileList[i].mOwner);
fprintf(stderr, "Group: %s, ", fileList[i].mGroup);
char permissions[10];
permission_disp(fileList[i].mPermissions, permissions);
fprintf(stderr, "Permissions: %d (%s)\n", fileList[i].mPermissions, permissions);
}
hdfsFreeFileInfo(fileList, numEntries);
} else {
if (errno) {
totalResult++;
fprintf(stderr, "waah! hdfsListDirectory - FAILED!\n");
} else {
fprintf(stderr, "Empty directory!\n");
}
}
char*** hosts = hdfsGetHosts(fs, srcPath, 0, 1);
if(hosts) {
fprintf(stderr, "hdfsGetHosts - SUCCESS! ... \n");
int i=0;
while(hosts[i]) {
int j = 0;
while(hosts[i][j]) {
fprintf(stderr,
"\thosts[%d][%d] - %s\n", i, j, hosts[i][j]);
++j;
}
++i;
}
} else {
totalResult++;
fprintf(stderr, "waah! hdfsGetHosts - FAILED!\n");
}
char *newOwner = "root";
// setting tmp dir to 777 so later when connectAsUser nobody, we can write to it
short newPerm = 0666;
// chown write
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, NULL, "users")) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChown: %s\n", ((result = hdfsChown(fs, writePath, newOwner, NULL)) ? "Failed!" : "Success!"));
totalResult += result;
// chmod write
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, writePath, newPerm)) ? "Failed!" : "Success!"));
totalResult += result;
sleep(2);
tTime newMtime = time(NULL);
tTime newAtime = time(NULL);
// utime write
fprintf(stderr, "hdfsUtime: %s\n", ((result = hdfsUtime(fs, writePath, newMtime, newAtime)) ? "Failed!" : "Success!"));
totalResult += result;
// chown/chmod/utime read
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, writePath);
fprintf(stderr, "hdfsChown read: %s\n", ((result = (strcmp(finfo->mOwner, newOwner) != 0)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsChmod read: %s\n", ((result = (finfo->mPermissions != newPerm)) ? "Failed!" : "Success!"));
totalResult += result;
// will later use /tmp/ as a different user so enable it
fprintf(stderr, "hdfsChmod: %s\n", ((result = hdfsChmod(fs, "/tmp/", 0777)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr,"newMTime=%ld\n",newMtime);
fprintf(stderr,"curMTime=%ld\n",finfo->mLastMod);
fprintf(stderr, "hdfsUtime read (mtime): %s\n", ((result = (finfo->mLastMod != newMtime)) ? "Failed!" : "Success!"));
totalResult += result;
// No easy way to turn on access times from hdfs_test right now
// fprintf(stderr, "hdfsUtime read (atime): %s\n", ((result = (finfo->mLastAccess != newAtime)) ? "Failed!" : "Success!"));
// totalResult += result;
hdfsFreeFileInfo(finfo, 1);
// Clean up
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, newDirectory)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(fs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, srcPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsDelete: %s\n", ((result = hdfsDelete(lfs, dstPath)) ? "Failed!" : "Success!"));
totalResult += result;
fprintf(stderr, "hdfsExists: %s\n", ((result = hdfsExists(fs, newDirectory)) ? "Success!" : "Failed!"));
totalResult += (result ? 0 : 1);
}
totalResult += (hdfsDisconnect(fs) != 0);
{
//
// Now test as connecting as a specific user
// This is only meant to test that we connected as that user, not to test
// the actual fs user capabilities. Thus just create a file and read
// the owner is correct.
const char *tuser = "******";
const char* writePath = "/tmp/usertestfile.txt";
const char **groups = (const char**)malloc(sizeof(char*)* 2);
groups[0] = "users";
groups[1] = "nobody";
fs = hdfsConnectAsUser("default", 0, tuser, groups, 2);
if(!fs) {
fprintf(stderr, "Oops! Failed to connect to hdfs as user %s!\n",tuser);
exit(-1);
}
hdfsFile writeFile = hdfsOpenFile(fs, writePath, O_WRONLY|O_CREAT, 0, 0, 0);
if(!writeFile) {
fprintf(stderr, "Failed to open %s for writing!\n", writePath);
exit(-1);
}
fprintf(stderr, "Opened %s for writing successfully...\n", writePath);
char* buffer = "Hello, World!";
tSize num_written_bytes = hdfsWrite(fs, writeFile, (void*)buffer, strlen(buffer)+1);
fprintf(stderr, "Wrote %d bytes\n", num_written_bytes);
if (hdfsFlush(fs, writeFile)) {
fprintf(stderr, "Failed to 'flush' %s\n", writePath);
exit(-1);
}
fprintf(stderr, "Flushed %s successfully!\n", writePath);
hdfsCloseFile(fs, writeFile);
hdfsFileInfo *finfo = hdfsGetPathInfo(fs, writePath);
fprintf(stderr, "hdfs new file user is correct: %s\n", ((result = (strcmp(finfo->mOwner, tuser) != 0)) ? "Failed!" : "Success!"));
totalResult += result;
}
totalResult += (hdfsDisconnect(fs) != 0);
if (totalResult != 0) {
return -1;
} else {
return 0;
}
}
