You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
koffice/kexi/kexidb/cursor.cpp

572 lines
15 KiB

/* This file is part of the KDE project
Copyright (C) 2003-2006 Jaroslaw Staniek <js@iidea.pl>
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.
*/
#include <kexidb/cursor.h>
#include <kexidb/driver.h>
#include <kexidb/driver_p.h>
#include <kexidb/error.h>
#include <kexidb/roweditbuffer.h>
#include <kexiutils/utils.h>
#include <kdebug.h>
#include <klocale.h>
#include <assert.h>
#include <stdlib.h>
using namespace KexiDB;
#ifdef KEXI_DEBUG_GUI
#endif
Cursor::Cursor(Connection* conn, const TQString& statement, uint options)
: TQObject()
, m_conn(conn)
, m_query(0)
, m_rawStatement(statement)
, m_options(options)
{
#ifdef KEXI_DEBUG_GUI
KexiUtils::addKexiDBDebug(TQString("Create cursor: ")+statement);
#endif
init();
}
Cursor::Cursor(Connection* conn, QuerySchema& query, uint options )
: TQObject()
, m_conn(conn)
, m_query(&query)
, m_options(options)
{
#ifdef KEXI_DEBUG_GUI
KexiUtils::addKexiDBDebug(TQString("Create cursor for query \"%1\": ").arg(query.name())+query.debugString());
#endif
init();
}
void Cursor::init()
{
assert(m_conn);
m_conn->m_cursors.insert(this,this);
m_opened = false;
// , m_atFirst(false)
// , m_atLast(false)
// , m_beforeFirst(false)
m_atLast = false;
m_afterLast = false;
m_readAhead = false;
m_at = 0;
//js:todo: if (m_query)
// m_fieldCount = m_query->fieldsCount();
// m_fieldCount = m_query ? m_query->fieldCount() : 0; //do not know
//<members related to buffering>
// m_cols_pointers_mem_size = 0;
m_records_in_buf = 0;
m_buffering_completed = false;
m_at_buffer = false;
m_result = -1;
m_containsROWIDInfo = (m_query && m_query->masterTable())
&& m_conn->driver()->beh->ROW_ID_FIELD_RETURNS_LAST_AUTOINCREMENTED_VALUE == false;
if (m_query) {
//get list of all fields
m_fieldsExpanded = new QueryColumnInfo::Vector();
*m_fieldsExpanded = m_query->fieldsExpanded(
m_containsROWIDInfo ? QuerySchema::WithInternalFieldsAndRowID : QuerySchema::WithInternalFields);
m_logicalFieldCount = m_fieldsExpanded->count()
- m_query->internalFields().count() - (m_containsROWIDInfo?1:0);
m_fieldCount = m_fieldsExpanded->count();
} else {
m_fieldsExpanded = 0;
m_logicalFieldCount = 0;
m_fieldCount = 0;
}
m_orderByColumnList = 0;
m_queryParameters = 0;
}
Cursor::~Cursor()
{
#ifdef KEXI_DEBUG_GUI
if (m_query)
KexiUtils::addKexiDBDebug(TQString("~ Delete cursor for query"));
else
KexiUtils::addKexiDBDebug(TQString("~ Delete cursor: ")+m_rawStatement);
#endif
/* if (!m_query)
KexiDBDbg << "Cursor::~Cursor() '" << m_rawStatement.latin1() << "'" << endl;
else
KexiDBDbg << "Cursor::~Cursor() " << endl;*/
//take me if delete was
if (!m_conn->m_destructor_started)
m_conn->m_cursors.take(this);
else {
KexiDBDbg << "Cursor::~Cursor() can be destroyed with Conenction::deleteCursor(), not with delete operator !"<< endl;
exit(1);
}
delete m_fieldsExpanded;
delete m_queryParameters;
}
bool Cursor::open()
{
if (m_opened) {
if (!close())
return false;
}
if (!m_rawStatement.isEmpty())
m_conn->m_sql = m_rawStatement;
else {
if (!m_query) {
KexiDBDbg << "Cursor::open(): no query statement (or schema) defined!" << endl;
setError(ERR_SQL_EXECUTION_ERROR, i18n("No query statement or schema defined."));
return false;
}
Connection::SelectStatementOptions options;
options.alsoRetrieveROWID = m_containsROWIDInfo; /*get ROWID if needed*/
m_conn->m_sql = m_queryParameters
? m_conn->selectStatement( *m_query, *m_queryParameters, options )
: m_conn->selectStatement( *m_query, options );
if (m_conn->m_sql.isEmpty()) {
KexiDBDbg << "Cursor::open(): empty statement!" << endl;
setError(ERR_SQL_EXECUTION_ERROR, i18n("Query statement is empty."));
return false;
}
}
m_sql = m_conn->m_sql;
m_opened = drv_open();
// m_beforeFirst = true;
m_afterLast = false; //we are not @ the end
m_at = 0; //we are before 1st rec
if (!m_opened) {
setError(ERR_SQL_EXECUTION_ERROR, i18n("Error opening database cursor."));
return false;
}
m_validRecord = false;
//luci: WHAT_EXACTLY_SHOULD_THAT_BE?
// if (!m_readAhead) // jowenn: to ensure before first state, without cluttering implementation code
if (m_conn->driver()->beh->_1ST_ROW_READ_AHEAD_REQUIRED_TO_KNOW_IF_THE_RESULT_IS_EMPTY) {
// KexiDBDbg << "READ AHEAD:" << endl;
m_readAhead = getNextRecord(); //true if any record in this query
// KexiDBDbg << "READ AHEAD = " << m_readAhead << endl;
}
m_at = 0; //we are still before 1st rec
return !error();
}
bool Cursor::close()
{
if (!m_opened)
return true;
bool ret = drv_close();
clearBuffer();
m_opened = false;
// m_beforeFirst = false;
m_afterLast = false;
m_readAhead = false;
m_fieldCount = 0;
m_logicalFieldCount = 0;
m_at = -1;
// KexiDBDbg<<"Cursor::close() == "<<ret<<endl;
return ret;
}
bool Cursor::reopen()
{
if (!m_opened)
return open();
return close() && open();
}
bool Cursor::moveFirst()
{
if (!m_opened)
return false;
// if (!m_beforeFirst) { //cursor isn't @ first record now: reopen
if (!m_readAhead) {
if (m_options & Buffered) {
if (m_records_in_buf==0 && m_buffering_completed) {
//eof and bof should now return true:
m_afterLast = true;
m_at = 0;
return false; //buffering completed and there is no records!
}
if (m_records_in_buf>0) {
//set state as we would be before first rec:
m_at_buffer = false;
m_at = 0;
//..and move to next, ie. 1st record
// m_afterLast = m_afterLast = !getNextRecord();
m_afterLast = !getNextRecord();
return !m_afterLast;
}
}
if (m_afterLast && m_at==0) //failure if already no records
return false;
if (!reopen()) //try reopen
return false;
if (m_afterLast) //eof
return false;
}
else {
//we have a record already read-ahead: we now point @ that:
m_at = 1;
}
// if (!m_atFirst) { //cursor isn't @ first record now: reopen
// reopen();
// }
// if (m_validRecord) {
// return true; //there is already valid record retrieved
// }
//get first record
// if (drv_moveFirst() && drv_getRecord()) {
// m_beforeFirst = false;
m_afterLast = false;
m_readAhead = false; //1st record had been read
// }
return m_validRecord;
}
bool Cursor::moveLast()
{
if (!m_opened)
return false;
if (m_afterLast || m_atLast) {
return m_validRecord; //we already have valid last record retrieved
}
if (!getNextRecord()) { //at least next record must be retrieved
// m_beforeFirst = false;
m_afterLast = true;
m_validRecord = false;
m_atLast = false;
return false; //no records
}
while (getNextRecord()) //move after last rec.
;
// m_beforeFirst = false;
m_afterLast = false;
//cursor shows last record data
m_atLast = true;
// m_validRecord = true;
/*
//we are before or @ last record:
// if (m_atLast && m_validRecord) //we're already @ last rec.
// return true;
if (m_validRecord) {
if (drv_getRecord())
}
if (!m_validRecord) {
if (drv_getRecord() && m_validRecord)
return true;
reopen();
}
*/
return true;
}
bool Cursor::moveNext()
{
if (!m_opened || m_afterLast)
return false;
if (getNextRecord()) {
// m_validRecord = true;
return true;
}
return false;
}
bool Cursor::movePrev()
{
if (!m_opened /*|| m_beforeFirst*/ || !(m_options & Buffered))
return false;
//we're after last record and there are records in the buffer
//--let's move to last record
if (m_afterLast && (m_records_in_buf>0)) {
drv_bufferMovePointerTo(m_records_in_buf-1);
m_at=m_records_in_buf;
m_at_buffer = true; //now current record is stored in the buffer
m_validRecord=true;
m_afterLast=false;
return true;
}
//we're at first record: go BOF
if ((m_at <= 1) || (m_records_in_buf <= 1/*sanity*/)) {
m_at=0;
m_at_buffer = false;
m_validRecord=false;
return false;
}
m_at--;
if (m_at_buffer) {//we already have got a pointer to buffer
drv_bufferMovePointerPrev(); //just move to prev record in the buffer
} else {//we have no pointer
//compute a place in the buffer that contain next record's data
drv_bufferMovePointerTo(m_at-1);
m_at_buffer = true; //now current record is stored in the buffer
}
m_validRecord=true;
m_afterLast=false;
return true;
}
bool Cursor::eof() const
{
return m_afterLast;
}
bool Cursor::bof() const
{
return m_at==0;
}
TQ_LLONG Cursor::at() const
{
if (m_readAhead)
return 0;
return m_at - 1;
}
bool Cursor::isBuffered() const
{
return m_options & Buffered;
}
void Cursor::setBuffered(bool buffered)
{
if (!m_opened)
return;
if (isBuffered()==buffered)
return;
m_options ^= Buffered;
}
void Cursor::clearBuffer()
{
if ( !isBuffered() || m_fieldCount==0)
return;
drv_clearBuffer();
m_records_in_buf=0;
m_at_buffer=false;
}
bool Cursor::getNextRecord()
{
m_result = -1; //by default: invalid result of row fetching
if ((m_options & Buffered)) {//this cursor is buffered:
// KexiDBDbg << "m_at < m_records_in_buf :: " << (long)m_at << " < " << m_records_in_buf << endl;
//js if (m_at==-1) m_at=0;
if (m_at < m_records_in_buf) {//we have next record already buffered:
/// if (m_at < (m_records_in_buf-1)) {//we have next record already buffered:
//js if (m_at_buffer && (m_at!=0)) {//we already have got a pointer to buffer
if (m_at_buffer) {//we already have got a pointer to buffer
drv_bufferMovePointerNext(); //just move to next record in the buffer
} else {//we have no pointer
//compute a place in the buffer that contain next record's data
drv_bufferMovePointerTo(m_at-1+1);
// drv_bufferMovePointerTo(m_at+1);
m_at_buffer = true; //now current record is stored in the buffer
}
}
else {//we are after last retrieved record: we need to physically fetch next record:
if (!m_readAhead) {//we have no record that was read ahead
if (!m_buffering_completed) {
//retrieve record only if we are not after
//the last buffer's item (i.e. when buffer is not fully filled):
// KexiDBDbg<<"==== buffering: drv_getNextRecord() ===="<<endl;
drv_getNextRecord();
}
if ((FetchResult) m_result != FetchOK) {//there is no record
m_buffering_completed = true; //no more records for buffer
// KexiDBDbg<<"m_result != FetchOK ********"<<endl;
m_validRecord = false;
m_afterLast = true;
//js m_at = m_records_in_buf;
m_at = -1; //position is invalid now and will not be used
if ((FetchResult) m_result == FetchEnd) {
return false;
}
setError(ERR_CURSOR_RECORD_FETCHING, i18n("Cannot fetch next record."));
return false;
}
//we have a record: store this record's values in the buffer
drv_appendCurrentRecordToBuffer();
m_records_in_buf++;
}
else //we have a record that was read ahead: eat this
m_readAhead = false;
}
}
else {//we are after last retrieved record: we need to physically fetch next record:
if (!m_readAhead) {//we have no record that was read ahead
// KexiDBDbg<<"==== no prefetched record ===="<<endl;
drv_getNextRecord();
if ((FetchResult)m_result != FetchOK) {//there is no record
// KexiDBDbg<<"m_result != FetchOK ********"<<endl;
m_validRecord = false;
m_afterLast = true;
m_at = -1;
if ((FetchResult) m_result == FetchEnd) {
return false;
}
setError(ERR_CURSOR_RECORD_FETCHING, i18n("Cannot fetch next record."));
return false;
}
}
else //we have a record that was read ahead: eat this
m_readAhead = false;
}
m_at++;
// if (m_data->curr_colname && m_data->curr_coldata)
// for (int i=0;i<m_data->curr_cols;i++) {
// KexiDBDbg<<i<<": "<< m_data->curr_colname[i]<<" == "<< m_data->curr_coldata[i]<<endl;
// }
// KexiDBDbg<<"m_at == "<<(long)m_at<<endl;
m_validRecord = true;
return true;
}
bool Cursor::updateRow(RowData& data, RowEditBuffer& buf, bool useROWID)
{
//! @todo doesn't update cursor's buffer YET!
clearError();
if (!m_query)
return false;
return m_conn->updateRow(*m_query, data, buf, useROWID);
}
bool Cursor::insertRow(RowData& data, RowEditBuffer& buf, bool getROWID)
{
//! @todo doesn't update cursor's buffer YET!
clearError();
if (!m_query)
return false;
return m_conn->insertRow(*m_query, data, buf, getROWID);
}
bool Cursor::deleteRow(RowData& data, bool useROWID)
{
//! @todo doesn't update cursor's buffer YET!
clearError();
if (!m_query)
return false;
return m_conn->deleteRow(*m_query, data, useROWID);
}
bool Cursor::deleteAllRows()
{
//! @todo doesn't update cursor's buffer YET!
clearError();
if (!m_query)
return false;
return m_conn->deleteAllRows(*m_query);
}
TQString Cursor::debugString() const
{
TQString dbg = "CURSOR( ";
if (!m_query) {
dbg += "RAW STATEMENT: '";
dbg += m_rawStatement;
dbg += "'\n";
}
else {
dbg += "QuerySchema: '";
dbg += m_conn->selectStatement( *m_query );
dbg += "'\n";
}
if (isOpened())
dbg += " OPENED";
else
dbg += " NOT_OPENED";
if (isBuffered())
dbg += " BUFFERED";
else
dbg += " NOT_BUFFERED";
dbg += " AT=";
dbg += TQString::number((unsigned long)at());
dbg += " )";
return dbg;
}
void Cursor::debug() const
{
KexiDBDbg << debugString() << endl;
}
void Cursor::setOrderByColumnList(const TQStringList& columnNames)
{
Q_UNUSED(columnNames);
//! @todo implement this:
// all field names should be fooun, exit otherwise ..........
// OK
//TODO if (!m_orderByColumnList)
//TODO
}
/*! Convenience method, similar to setOrderBy(const TQStringList&). */
void Cursor::setOrderByColumnList(const TQString& column1, const TQString& column2,
const TQString& column3, const TQString& column4, const TQString& column5)
{
Q_UNUSED(column1);
Q_UNUSED(column2);
Q_UNUSED(column3);
Q_UNUSED(column4);
Q_UNUSED(column5);
//! @todo implement this, like above
//! @todo add ORDER BY info to debugString()
}
QueryColumnInfo::Vector Cursor::orderByColumnList() const
{
return m_orderByColumnList ? *m_orderByColumnList: QueryColumnInfo::Vector();
}
TQValueList<TQVariant> Cursor::queryParameters() const
{
return m_queryParameters ? *m_queryParameters : TQValueList<TQVariant>();
}
void Cursor::setQueryParameters(const TQValueList<TQVariant>& params)
{
if (!m_queryParameters)
m_queryParameters = new TQValueList<TQVariant>(params);
else
*m_queryParameters = params;
}
#include "cursor.moc"