path: root/kexi/kexidb/cursor.h

/* This file is part of the KDE project
   Copyright (C) 2003-2006 Jaroslaw Staniek <[email protected]>

   This program is free software; you can redistribute it and/or
   modify it under the terms of the GNU Library General Public
   License as published by the Free Software Foundation; either
   version 2 of the License, or (at your option) any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Library General Public License for more details.

   You should have received a copy of the GNU Library General Public License
   along with this program; see the file COPYING.  If not, write to
   the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
 * Boston, MA 02110-1301, USA.
*/

#ifndef KEXIDB_CURSOR_H
#define KEXIDB_CURSOR_H

#include <tqstring.h>
#include <tqvariant.h>
#include <tqptrvector.h>
#include <tqvaluevector.h>

#include <kexidb/connection.h>
#include <kexidb/object.h>

namespace KexiDB {

class RowEditBuffer;

//! Provides database cursor functionality.
/*!
	Cursor can be defined in two ways:

	-# by passing QuerySchema object to Connection::executeQuery() or Connection::prepareQuery();
	   then query is defined for in engine-independent way -- this is recommended usage

	-# by passing raw query statement string to Connection::executeQuery() or Connection::prepareQuery();
	   then query may be defined for in engine-dependent way -- this is not recommended usage,
	   but convenient when we can't or do not want to allocate QuerySchema object, while we
	   know that the query statement is syntactically and logically ok in our context.

	You can move cursor to next record with moveNext() and move back with movePrev().
	The cursor is always positioned on record, not between records, with exception that 
	ofter open() it is positioned before first record (if any) -- then bof() equals true,
	and can be positioned after the last record (if any) with moveNext() -- then eof() equals true,
	For example, if you have four records 1, 2, 3, 4, then after calling moveNext(), 
	moveNext(), moveNext(), movePrev() you are going through records: 1, 2, 3, 2.

	Cursor can be buffered or unbuferred.
	Buffering in this class is not related to any SQL engine capatibilities for server-side cursors 
	(eg. like 'DECLARE CURSOR' statement) - buffered data is at client (application) side.
	Any record retrieved in buffered cursor will be stored inside an internal buffer
	and reused when needed. Unbuffered cursor always requires one record fetching from
	db connection at every step done with moveNext(), movePrev(), etc.

	Notes:
	- Do not use delete operator for Cursor objects - this will fail; use Connection::deleteCursor()
	instead.
	- QuerySchema object is not owned by Cursor object that uses it.
*/
class KEXI_DB_EXPORT Cursor: public TQObject, public Object
{
	Q_OBJECT
  TQ_OBJECT

	public:
		//! Cursor options that describes its behaviour
		enum Options {
			NoOptions = 0,
			Buffered = 1
		};

		virtual ~Cursor();

		/*! \return connection used for the cursor */
		inline Connection* connection() const { return m_conn; }

		/*! Opens the cursor using data provided on creation. 
		 The data might be either QuerySchema or raw sql statement. */
		bool open();

		/*! Closes and then opens again the same cursor. 
		 If the cursor is not opened it is just opened and result of this open is returned.
		 Otherwise, true is returned if cursor is successfully closed and then opened. */
		bool reopen();

//		/*! Opens the cursor using \a statement. 
//		 Omit \a statement if cursor is already initialized with statement 
//		 at creation time. If \a statement is not empty, existing statement
//		 (if any) is overwritten. */
//		bool open( const TQString& statement = TQString() );

		/*! Closes previously opened cursor. 
			If the cursor is closed, nothing happens. */
		virtual bool close();

		/*! \return query schema used to define this cursor
		 or NULL if the cursor is not defined by a query schema but by a raw statement. */
		inline QuerySchema *query() const { return m_query; }

		//! \return query parameters assigned to this cursor
		TQValueList<TQVariant> queryParameters() const;

		//! Sets query parameters \a params for this cursor.
		void setQueryParameters(const TQValueList<TQVariant>& params);

		/*! \return raw query statement used to define this cursor
		 or null string if raw statement instead (but QuerySchema is defined instead). */
		inline TQString rawStatement() const { return m_rawStatement; }

		/*! \return logically or'd cursor's options, 
			selected from Cursor::Options enum. */
		inline uint options() const { return m_options; }

		/*! \return true if the cursor is opened. */
		inline bool isOpened() const { return m_opened; }

		/*! \return true if the cursor is buffered. */
		bool isBuffered() const;

		/*! Sets this cursor to buffered type or not. See description 
			of buffered and nonbuffered cursors in class description.
			This method only works if cursor is not opened (isOpened()==false).
			You can close already opened cursor and then switch this option on/off.
		*/
		void setBuffered(bool buffered);

		/*! Moves current position to the first record and retrieves it.
			\return true if the first record was retrieved.
			False could mean that there was an error or there is no record available. */
		bool moveFirst();

		/*! Moves current position to the last record and retrieves it. 
			\return true if the last record was retrieved.
			False could mean that there was an error or there is no record available. */
		virtual bool moveLast();

		/*! Moves current position to the next record and retrieves it. */
		virtual bool moveNext();

		/*! Moves current position to the next record and retrieves it. 
		 Currently it's only supported for buffered cursors. */
		virtual bool movePrev();

		/*! \return true if current position is after last record. */
		bool eof() const;

		/*! \return true if current position is before first record. */
		bool bof() const;

		/*! \return current internal position of the cursor's query. 
		 We are counting records from 0.
		 Value -1 means that cursor does not point to any valid record
		 (this happens eg. after open(), close(), 
		 and after moving after last record or before first one. */
		TQ_LLONG at() const;

		/*! \return number of fields available for this cursor. 
		 This never includes ROWID column or other internal coluns (e.g. lookup). */
		inline uint fieldCount() const { return m_query ? m_logicalFieldCount : m_fieldCount; }

		/*! \return true if ROWID information is appended with every row.
		 ROWID information is available 
		 if DriverBehaviour::ROW_ID_FIELD_RETURNS_LAST_AUTOINCREMENTED_VALUE == false
		 for a KexiDB database driver and the master table has no primary key defined. 
		 Phisically, ROWID value is returned after last returned field,
		 so data vector's length is expanded by one. */
		inline bool containsROWIDInfo() const { return m_containsROWIDInfo; }

		/*! \return a value stored in column number \a i (counting from 0).
		 Is has unspecified behaviour if the cursor is not at valid record.
		 Note for driver developers: 
		 If \a i is >= than m_fieldCount, null TQVariant value should be returned. 
		 To return a value typically you can use a pointer to internal structure 
		 that contain current row data (buffered or unbuffered). */
		virtual TQVariant value(uint i) = 0;

		/*! [PROTOTYPE] \return current record data or NULL if there is no current records. */
		virtual const char ** rowData() const = 0;

		/*! Sets a list of columns for ORDER BY section of the query. 
		 Only works when the cursor has been created using QuerySchema object 
		 (i.e. when query()!=0; does not work with raw statements).
		 Each name on the list must be a field or alias present within the query 
		 and must not be covered by aliases. If one or more names cannot be found within 
		 the query, the method will have no effect. Any previous ORDER BY settings will be removed. 

		 The order list provided here has priority over a list defined in the QuerySchema
		 object itseld (using QuerySchema::setOrderByColumnList()).
		 The QuerySchema object itself is not modifed by this method: only order of records retrieved 
		 by this cursor is affected. 
		
		 Use this method before calling open(). You can also call reopen() after calling this method 
		 to see effects of applying records order. */
		void setOrderByColumnList(const TQStringList& columnNames);

		/*! Convenience method, similar to setOrderByColumnList(const TQStringList&). */
		void setOrderByColumnList(const TQString& column1, const TQString& column2 = TQString(), 
			const TQString& column3 = TQString(), const TQString& column4 = TQString(), 
			const TQString& column5 = TQString());

		/*! \return a list of fields contained in ORDER BY section of the query. 
		 @see setOrderBy(const TQStringList&) */
		QueryColumnInfo::Vector orderByColumnList() const;

		/*! Puts current record's data into \a data (makes a deep copy).
		 This have unspecified behaviour if the cursor is not at valid record.
		 Note: For reimplementation in driver's code. Shortly, this method translates
		 a row data from internal representation (probably also used in buffer)
		 to simple public RecordData representation. */
		virtual void storeCurrentRow(RowData &data) const = 0;

		bool updateRow(RowData& data, RowEditBuffer& buf, bool useROWID = false);

		bool insertRow(RowData& data, RowEditBuffer& buf, bool getROWID = false);

		bool deleteRow(RowData& data, bool useROWID = false);

		bool deleteAllRows();

		/*! \return a code of last executed operation's result at the server side.
		 This code is engine dependent and may be even engine-version dependent.
		 It can be visible in applications mainly after clicking a "Details>>" button
		 or something like that -- this just can be useful for advanced users and 
		 for testing. 
		 Note for driver developers: Return here the value you usually store as result 
		 of most lower-level operations. By default this method returns 0. */
		virtual int serverResult() { return 0; }
		
		/*! \return (not i18n'd) name of last executed operation's result at the server side.
		 Sometimes engines have predefined its result names that can be used e.g. 
		 to refer a documentation. SQLite is one of such engines. 
		 Note for driver developers: Leave the default implementation (null 
		 string is returned ) if your engine has no such capability. */
		virtual TQString serverResultName() { return TQString(); }

		/*! \return (not i18n'd) description text (message) of last operation's error/result.
		 In most cases engines do return such a messages, any user can then use this
		 to refer a documentation.
		 Note for driver developers: Leave the default implementation (null 
		 string is returned ) if your engine has no such capability. */
		virtual TQString serverErrorMsg() { return TQString(); }
	
		/*! \return Debug information. */
		TQString debugString() const;
		
		//! Outputs debug information.
		void debug() const;

	protected:
		//! possible results of row fetching, used for m_result
		typedef enum FetchResult { FetchError=0, FetchOK=1, FetchEnd=2 };

		/*! Cursor will operate on \a conn, raw \a statement will be used to execute query. */
		Cursor(Connection* conn, const TQString& statement, uint options = NoOptions );

		/*! Cursor will operate on \a conn, \a query schema will be used to execute query. */
		Cursor(Connection* conn, QuerySchema& query, uint options = NoOptions );

		void init();

		/*! Internal: cares about proper flag setting depending on result of drv_getNextRecord()
		 and depending on wherher a cursor is buffered. */
		bool getNextRecord();

		/* Note for driver developers: this method should initialize engine-specific cursor's
		 resources using m_sql statement. It is not required to store \a statement somewhere
		 in your Cursor subclass (it is already stored in m_query or m_rawStatement, 
		 depending query type) - only pass it to proper engine's function. */
		virtual bool drv_open() = 0;

		virtual bool drv_close() = 0;
//		virtual bool drv_moveFirst() = 0;
		virtual void drv_getNextRecord() = 0;
//unused		virtual bool drv_getPrevRecord() = 0;

		/*! Stores currently fetched record's values in appropriate place of the buffer.
		 Note for driver developers: 
		 This place can be computed using m_at. Do not change value of m_at or any other
		 Cursor members, only change your internal structures like pointer to current 
		 row, etc. If your database engine's API function (for record fetching) 
		 do not allocates such a space, you want to allocate a space for current
		 record. Otherwise, reuse existing structure, what could be more efficient.
		 All functions like drv_appendCurrentRecordToBuffer() operates on the buffer,
		 i.e. array of stored rows. You are not forced to have any particular
		 fixed structure for buffer item or buffer itself - the structure is internal and
		 only methods like storeCurrentRecord() visible to public.
		*/
		virtual void drv_appendCurrentRecordToBuffer() = 0;
		/*! Moves pointer (that points to the buffer) -- to next item in this buffer.
		 Note for driver developers: probably just execute "your_pointer++" is enough.
		*/
		virtual void drv_bufferMovePointerNext() = 0;
		/*! Like drv_bufferMovePointerNext() but execute "your_pointer--". */
		virtual void drv_bufferMovePointerPrev() = 0;
		/*! Moves pointer (that points to the buffer) to a new place: \a at.
		*/
		virtual void drv_bufferMovePointerTo(TQ_LLONG at) = 0;

		/*DISABLED: ! This is called only once in open(), after successful drv_open().
			Reimplement this if you need (or not) to do get the first record after drv_open(),
			eg. to know if there are any records in table. Value returned by this method
			will be assigned to m_readAhead.
			Default implementation just calls drv_getNextRecord(). */

		/*! Clears cursor's buffer if this was allocated (only for buffered cursor type).
			Otherwise do nothing. For reimplementing. Default implementation does nothing. */
		virtual void drv_clearBuffer() {}

		//! @internal clears buffer with reimplemented drv_clearBuffer(). */
		void clearBuffer();

		/*! Clears an internal member that is used to storing last result code, 
		 the same that is returend by serverResult(). */
		virtual void drv_clearServerResult() = 0;

		TQGuardedPtr<Connection> m_conn;
		QuerySchema *m_query;
//		CursorData *m_data;
		TQString m_rawStatement;
		bool m_opened : 1;
//js (m_at==0 is enough)		bool m_beforeFirst : 1;
		bool m_atLast : 1;
		bool m_afterLast : 1;
//		bool m_atLast;
		bool m_validRecord : 1; //!< true if valid record is currently retrieved @ current position
		bool m_containsROWIDInfo : 1;
		TQ_LLONG m_at;
		uint m_fieldCount; //!< cached field count information
		uint m_logicalFieldCount;  //!< logical field count, i.e. without intrernal values like ROWID or lookup
		uint m_options; //!< cursor options that describes its behaviour
		char m_result; //!< result of a row fetching
		
		//<members related to buffering>
		int m_records_in_buf;          //!< number of records currently stored in the buffer
		bool m_buffering_completed : 1;   //!< true if we already have all records stored in the buffer
		//</members related to buffering>

		//! Useful e.g. for value(int) method when we need access to schema def.
		QueryColumnInfo::Vector* m_fieldsExpanded;

		//! Used by setOrderByColumnList()
		QueryColumnInfo::Vector* m_orderByColumnList;

		TQValueList<TQVariant>* m_queryParameters;

	private:
		bool m_readAhead : 1;
		
		//<members related to buffering>
		bool m_at_buffer : 1;             //!< true if we already point to the buffer with curr_coldata
		//</members related to buffering>


		class Private;
		Private *d;
};

} //namespace KexiDB

#endif