mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-09-20 08:40:50 +00:00
Refactor ReadBufferFromHDFS
This commit is contained in:
parent
78671d15bd
commit
2d3e08fc74
@ -299,8 +299,8 @@ target_include_directories (dbms SYSTEM BEFORE PRIVATE ${DIVIDE_INCLUDE_DIR})
|
||||
target_include_directories (dbms SYSTEM BEFORE PRIVATE ${SPARCEHASH_INCLUDE_DIR})
|
||||
|
||||
if (USE_HDFS)
|
||||
target_link_libraries (dbms PRIVATE ${HDFS3_LIBRARY})
|
||||
target_include_directories (dbms SYSTEM BEFORE PRIVATE ${HDFS3_INCLUDE_DIR})
|
||||
target_link_libraries (clickhouse_common_io PRIVATE ${HDFS3_LIBRARY})
|
||||
target_include_directories (clickhouse_common_io SYSTEM BEFORE PRIVATE ${HDFS3_INCLUDE_DIR})
|
||||
endif()
|
||||
|
||||
if (USE_JEMALLOC)
|
||||
|
104
dbms/src/IO/ReadBufferFromHDFS.cpp
Normal file
104
dbms/src/IO/ReadBufferFromHDFS.cpp
Normal file
@ -0,0 +1,104 @@
|
||||
#include <IO/ReadBufferFromHDFS.h>
|
||||
|
||||
#if USE_HDFS
|
||||
#include <Poco/URI.h>
|
||||
#include <hdfs/hdfs.h>
|
||||
|
||||
namespace DB
|
||||
{
|
||||
namespace ErrorCodes
|
||||
{
|
||||
extern const int BAD_ARGUMENTS;
|
||||
extern const int NETWORK_ERROR;
|
||||
}
|
||||
|
||||
struct ReadBufferFromHDFS::ReadBufferFromHDFSImpl
|
||||
{
|
||||
std::string hdfs_uri;
|
||||
struct hdfsBuilder * builder;
|
||||
hdfsFS fs;
|
||||
hdfsFile fin;
|
||||
|
||||
ReadBufferFromHDFSImpl(const std::string & hdfs_name_)
|
||||
: hdfs_uri(hdfs_name_)
|
||||
, builder(hdfsNewBuilder())
|
||||
{
|
||||
builder = hdfsNewBuilder();
|
||||
hdfs_uri = hdfs_name_;
|
||||
Poco::URI uri(hdfs_name_);
|
||||
auto & host = uri.getHost();
|
||||
auto port = uri.getPort();
|
||||
auto & path = uri.getPath();
|
||||
if (host.empty() || port == 0 || path.empty())
|
||||
{
|
||||
throw Exception("Illegal HDFS URI: " + hdfs_uri, ErrorCodes::BAD_ARGUMENTS);
|
||||
}
|
||||
// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
|
||||
/// TODO Allow to tune from query Settings.
|
||||
hdfsBuilderConfSetStr(builder, "input.read.timeout", "60000"); // 1 min
|
||||
hdfsBuilderConfSetStr(builder, "input.connect.timeout", "60000"); // 1 min
|
||||
|
||||
hdfsBuilderSetNameNode(builder, host.c_str());
|
||||
hdfsBuilderSetNameNodePort(builder, port);
|
||||
fs = hdfsBuilderConnect(builder);
|
||||
|
||||
if (fs == nullptr)
|
||||
{
|
||||
throw Exception("Unable to connect to HDFS: " + std::string(hdfsGetLastError()), ErrorCodes::NETWORK_ERROR);
|
||||
}
|
||||
|
||||
fin = hdfsOpenFile(fs, path.c_str(), O_RDONLY, 0, 0, 0);
|
||||
}
|
||||
|
||||
~ReadBufferFromHDFSImpl()
|
||||
{
|
||||
close();
|
||||
hdfsFreeBuilder(builder);
|
||||
}
|
||||
|
||||
void close()
|
||||
{
|
||||
hdfsCloseFile(fs, fin);
|
||||
}
|
||||
|
||||
int read(char * start, size_t size)
|
||||
{
|
||||
int bytes_read = hdfsRead(fs, fin, start, size);
|
||||
if (bytes_read < 0)
|
||||
{
|
||||
throw Exception("Fail to read HDFS file: " + hdfs_uri + " " + std::string(hdfsGetLastError()), ErrorCodes::NETWORK_ERROR);
|
||||
}
|
||||
return bytes_read;
|
||||
}
|
||||
};
|
||||
|
||||
ReadBufferFromHDFS::ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size)
|
||||
: BufferWithOwnMemory<ReadBuffer>(buf_size)
|
||||
, impl(std::make_unique<ReadBufferFromHDFSImpl>(hdfs_name_))
|
||||
{
|
||||
}
|
||||
|
||||
|
||||
bool ReadBufferFromHDFS::nextImpl()
|
||||
{
|
||||
int bytes_read = impl->read(internal_buffer.begin(), internal_buffer.size());
|
||||
|
||||
if (bytes_read)
|
||||
working_buffer.resize(bytes_read);
|
||||
else
|
||||
return false;
|
||||
return true;
|
||||
}
|
||||
|
||||
const std::string & ReadBufferFromHDFS::getHDFSUri() const
|
||||
{
|
||||
return impl->hdfs_uri;
|
||||
}
|
||||
|
||||
ReadBufferFromHDFS::~ReadBufferFromHDFS()
|
||||
{
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
#endif
|
@ -4,94 +4,28 @@
|
||||
|
||||
#if USE_HDFS
|
||||
#include <IO/ReadBuffer.h>
|
||||
#include <Poco/URI.h>
|
||||
#include <hdfs/hdfs.h>
|
||||
#include <IO/BufferWithOwnMemory.h>
|
||||
#include <string>
|
||||
|
||||
#ifndef O_DIRECT
|
||||
#define O_DIRECT 00040000
|
||||
#endif
|
||||
#include <memory>
|
||||
|
||||
namespace DB
|
||||
{
|
||||
namespace ErrorCodes
|
||||
{
|
||||
extern const int BAD_ARGUMENTS;
|
||||
extern const int NETWORK_ERROR;
|
||||
}
|
||||
/** Accepts path to file and opens it, or pre-opened file descriptor.
|
||||
* Closes file by himself (thus "owns" a file descriptor).
|
||||
*/
|
||||
class ReadBufferFromHDFS : public BufferWithOwnMemory<ReadBuffer>
|
||||
{
|
||||
protected:
|
||||
std::string hdfs_uri;
|
||||
struct hdfsBuilder *builder;
|
||||
hdfsFS fs;
|
||||
hdfsFile fin;
|
||||
public:
|
||||
ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE)
|
||||
: BufferWithOwnMemory<ReadBuffer>(buf_size), hdfs_uri(hdfs_name_) , builder(hdfsNewBuilder())
|
||||
{
|
||||
Poco::URI uri(hdfs_name_);
|
||||
auto & host = uri.getHost();
|
||||
auto port = uri.getPort();
|
||||
auto & path = uri.getPath();
|
||||
if (host.empty() || port == 0 || path.empty())
|
||||
{
|
||||
throw Exception("Illegal HDFS URI: " + hdfs_uri, ErrorCodes::BAD_ARGUMENTS);
|
||||
}
|
||||
// set read/connect timeout, default value in libhdfs3 is about 1 hour, and too large
|
||||
/// TODO Allow to tune from query Settings.
|
||||
hdfsBuilderConfSetStr(builder, "input.read.timeout", "60000"); // 1 min
|
||||
hdfsBuilderConfSetStr(builder, "input.connect.timeout", "60000"); // 1 min
|
||||
/** Accepts path to file and opens it, or pre-opened file descriptor.
|
||||
* Closes file by himself (thus "owns" a file descriptor).
|
||||
*/
|
||||
class ReadBufferFromHDFS : public BufferWithOwnMemory<ReadBuffer>
|
||||
{
|
||||
struct ReadBufferFromHDFSImpl;
|
||||
std::unique_ptr<ReadBufferFromHDFSImpl> impl;
|
||||
public:
|
||||
ReadBufferFromHDFS(const std::string & hdfs_name_, size_t buf_size = DBMS_DEFAULT_BUFFER_SIZE);
|
||||
ReadBufferFromHDFS(ReadBufferFromHDFS &&) = default;
|
||||
|
||||
hdfsBuilderSetNameNode(builder, host.c_str());
|
||||
hdfsBuilderSetNameNodePort(builder, port);
|
||||
fs = hdfsBuilderConnect(builder);
|
||||
bool nextImpl() override;
|
||||
|
||||
if (fs == nullptr)
|
||||
{
|
||||
throw Exception("Unable to connect to HDFS: " + std::string(hdfsGetLastError()), ErrorCodes::NETWORK_ERROR);
|
||||
}
|
||||
~ReadBufferFromHDFS() override;
|
||||
|
||||
fin = hdfsOpenFile(fs, path.c_str(), O_RDONLY, 0, 0, 0);
|
||||
}
|
||||
|
||||
ReadBufferFromHDFS(ReadBufferFromHDFS &&) = default;
|
||||
|
||||
~ReadBufferFromHDFS() override
|
||||
{
|
||||
close();
|
||||
hdfsFreeBuilder(builder);
|
||||
}
|
||||
|
||||
/// Close HDFS connection before destruction of object.
|
||||
void close()
|
||||
{
|
||||
hdfsCloseFile(fs, fin);
|
||||
}
|
||||
|
||||
bool nextImpl() override
|
||||
{
|
||||
int bytes_read = hdfsRead(fs, fin, internal_buffer.begin(), internal_buffer.size());
|
||||
if (bytes_read < 0)
|
||||
{
|
||||
throw Exception("Fail to read HDFS file: " + hdfs_uri + " " + std::string(hdfsGetLastError()), ErrorCodes::NETWORK_ERROR);
|
||||
}
|
||||
|
||||
if (bytes_read)
|
||||
working_buffer.resize(bytes_read);
|
||||
else
|
||||
return false;
|
||||
return true;
|
||||
}
|
||||
|
||||
const std::string & getHDFSUri() const
|
||||
{
|
||||
return hdfs_uri;
|
||||
}
|
||||
};
|
||||
const std::string & getHDFSUri() const;
|
||||
};
|
||||
}
|
||||
#endif
|
||||
|
Loading…
Reference in New Issue
Block a user