2011-10-31 17:55:06 +00:00
|
|
|
#pragma once
|
2010-03-18 19:32:14 +00:00
|
|
|
|
2015-01-25 05:07:51 +00:00
|
|
|
#include <map>
|
2017-07-28 17:34:02 +00:00
|
|
|
#include <shared_mutex>
|
2017-06-06 17:18:32 +00:00
|
|
|
#include <ext/shared_ptr_helper.h>
|
2016-08-26 21:25:05 +00:00
|
|
|
|
2010-03-18 19:32:14 +00:00
|
|
|
#include <Poco/File.h>
|
|
|
|
|
2017-04-01 09:19:00 +00:00
|
|
|
#include <Storages/IStorage.h>
|
|
|
|
#include <Common/FileChecker.h>
|
|
|
|
#include <Common/escapeForFileName.h>
|
2010-03-18 19:32:14 +00:00
|
|
|
|
|
|
|
|
|
|
|
namespace DB
|
|
|
|
{
|
|
|
|
|
2017-08-07 07:31:16 +00:00
|
|
|
/** Implements simple table engine without support of indices.
|
2017-04-16 15:00:33 +00:00
|
|
|
* The data is stored in a compressed form.
|
2010-03-18 19:32:14 +00:00
|
|
|
*/
|
2017-06-06 18:36:13 +00:00
|
|
|
class StorageLog : public ext::shared_ptr_helper<StorageLog>, public IStorage
|
2010-03-18 19:32:14 +00:00
|
|
|
{
|
|
|
|
friend class LogBlockInputStream;
|
|
|
|
friend class LogBlockOutputStream;
|
|
|
|
|
|
|
|
public:
|
2017-04-01 07:20:54 +00:00
|
|
|
std::string getName() const override { return "Log"; }
|
|
|
|
std::string getTableName() const override { return name; }
|
|
|
|
|
2017-05-24 21:06:29 +00:00
|
|
|
BlockInputStreams read(
|
2017-04-01 07:20:54 +00:00
|
|
|
const Names & column_names,
|
2017-07-15 03:48:36 +00:00
|
|
|
const SelectQueryInfo & query_info,
|
2017-04-01 07:20:54 +00:00
|
|
|
const Context & context,
|
2018-04-19 14:47:09 +00:00
|
|
|
QueryProcessingStage::Enum processed_stage,
|
2017-06-02 15:54:39 +00:00
|
|
|
size_t max_block_size,
|
|
|
|
unsigned num_streams) override;
|
2017-04-01 07:20:54 +00:00
|
|
|
|
2017-05-21 22:25:25 +00:00
|
|
|
BlockOutputStreamPtr write(const ASTPtr & query, const Settings & settings) override;
|
2017-04-01 07:20:54 +00:00
|
|
|
|
|
|
|
void rename(const String & new_path_to_db, const String & new_database_name, const String & new_table_name) override;
|
|
|
|
|
|
|
|
bool checkData() const override;
|
2014-08-04 06:36:24 +00:00
|
|
|
|
2018-06-09 15:48:22 +00:00
|
|
|
void truncate(const ASTPtr &) override;
|
2018-04-21 00:35:20 +00:00
|
|
|
|
2018-02-21 19:26:59 +00:00
|
|
|
std::string full_path() const { return path + escapeForFileName(name) + '/';}
|
|
|
|
|
|
|
|
String getDataPath() const override { return full_path(); }
|
|
|
|
|
2013-02-07 13:03:19 +00:00
|
|
|
protected:
|
2017-06-06 18:36:13 +00:00
|
|
|
/** Attach the table with the appropriate name, along the appropriate path (with / at the end),
|
|
|
|
* (the correctness of names and paths is not verified)
|
|
|
|
* consisting of the specified columns; Create files if they do not exist.
|
|
|
|
*/
|
2017-04-01 07:20:54 +00:00
|
|
|
StorageLog(
|
|
|
|
const std::string & path_,
|
|
|
|
const std::string & name_,
|
2018-03-06 20:18:34 +00:00
|
|
|
const ColumnsDescription & columns_,
|
2017-04-01 07:20:54 +00:00
|
|
|
size_t max_compress_block_size_);
|
|
|
|
|
2017-11-04 03:20:18 +00:00
|
|
|
private:
|
|
|
|
String path;
|
|
|
|
String name;
|
2017-04-01 07:20:54 +00:00
|
|
|
|
2017-11-04 03:20:18 +00:00
|
|
|
mutable std::shared_mutex rwlock;
|
2017-04-01 07:20:54 +00:00
|
|
|
|
2017-11-26 19:22:33 +00:00
|
|
|
/** Offsets to some row number in a file for column in table.
|
|
|
|
* They are needed so that you can read the data in several threads.
|
|
|
|
*/
|
|
|
|
struct Mark
|
|
|
|
{
|
2017-11-28 02:13:46 +00:00
|
|
|
size_t rows; /// How many rows are before this offset including the block at this offset.
|
2017-11-26 19:22:33 +00:00
|
|
|
size_t offset; /// The offset in compressed file.
|
|
|
|
};
|
|
|
|
|
|
|
|
using Marks = std::vector<Mark>;
|
|
|
|
|
|
|
|
/// Column data
|
|
|
|
struct ColumnData
|
|
|
|
{
|
|
|
|
/// Specifies the column number in the marks file.
|
|
|
|
/// Does not necessarily match the column number among the columns of the table: columns with lengths of arrays are also numbered here.
|
|
|
|
size_t column_index;
|
|
|
|
|
|
|
|
Poco::File data_file;
|
|
|
|
Marks marks;
|
|
|
|
};
|
|
|
|
using Files_t = std::map<String, ColumnData>;
|
|
|
|
|
2017-04-01 07:20:54 +00:00
|
|
|
Files_t files; /// name -> data
|
2014-08-04 06:36:24 +00:00
|
|
|
|
2017-04-01 07:20:54 +00:00
|
|
|
Names column_names; /// column_index -> name
|
2014-09-30 03:08:47 +00:00
|
|
|
|
2017-04-01 07:20:54 +00:00
|
|
|
Poco::File marks_file;
|
2014-09-30 03:08:47 +00:00
|
|
|
|
2017-04-16 15:00:33 +00:00
|
|
|
/// The order of adding files should not change: it corresponds to the order of the columns in the marks file.
|
2017-08-07 07:31:16 +00:00
|
|
|
void addFiles(const String & column_name, const IDataType & type);
|
2012-08-29 20:07:24 +00:00
|
|
|
|
2017-11-28 02:13:46 +00:00
|
|
|
bool loaded_marks = false;
|
2013-12-12 22:55:47 +00:00
|
|
|
|
2017-04-01 07:20:54 +00:00
|
|
|
size_t max_compress_block_size;
|
|
|
|
size_t file_count = 0;
|
2014-03-28 14:36:24 +00:00
|
|
|
|
2017-04-01 07:20:54 +00:00
|
|
|
FileChecker file_checker;
|
2014-08-04 06:36:24 +00:00
|
|
|
|
2017-11-04 03:20:18 +00:00
|
|
|
/// Read marks files if they are not already read.
|
|
|
|
/// It is done lazily, so that with a large number of tables, the server starts quickly.
|
|
|
|
/// You can not call with a write locked `rwlock`.
|
|
|
|
void loadMarks();
|
|
|
|
|
|
|
|
/// The order of adding files should not change: it corresponds to the order of the columns in the marks file.
|
|
|
|
void addFile(const String & column_name, const IDataType & type, size_t level = 0);
|
|
|
|
|
2017-04-16 15:00:33 +00:00
|
|
|
/** For normal columns, the number of rows in the block is specified in the marks.
|
|
|
|
* For array columns and nested structures, there are more than one group of marks that correspond to different files
|
2017-11-26 19:22:33 +00:00
|
|
|
* - for elements (file name.bin) - the total number of array elements in the block is specified,
|
2017-04-16 15:00:33 +00:00
|
|
|
* - for array sizes (file name.size0.bin) - the number of rows (the whole arrays themselves) in the block is specified.
|
2017-04-01 07:20:54 +00:00
|
|
|
*
|
2017-04-16 15:00:33 +00:00
|
|
|
* Return the first group of marks that contain the number of rows, but not the internals of the arrays.
|
2017-04-01 07:20:54 +00:00
|
|
|
*/
|
|
|
|
const Marks & getMarksWithRealRowCount() const;
|
|
|
|
|
2017-08-07 07:31:16 +00:00
|
|
|
std::string getFullPath() const { return path + escapeForFileName(name) + '/'; }
|
2010-03-18 19:32:14 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
}
|