未验证 提交 8fce74f0 编写于 作者: S Shengliang Guan 提交者: GitHub

Merge pull request #2654 from taosdata/feature/query

Feature/query
......@@ -45,6 +45,7 @@ void doAsyncQuery(STscObj* pObj, SSqlObj* pSql, void (*fp)(), void* param, const
pSql->pTscObj = pObj;
pSql->maxRetry = TSDB_MAX_REPLICA;
pSql->fp = fp;
pSql->fetchFp = fp;
pSql->sqlstr = calloc(1, sqlLen + 1);
if (pSql->sqlstr == NULL) {
......@@ -159,7 +160,7 @@ static void tscProcessAsyncRetrieveImpl(void *param, TAOS_RES *tres, int numOfRo
pRes->code = numOfRows;
}
tscQueueAsyncError(pSql->fetchFp, param, pRes->code);
tscQueueAsyncRes(pSql);
return;
}
......@@ -167,6 +168,7 @@ static void tscProcessAsyncRetrieveImpl(void *param, TAOS_RES *tres, int numOfRo
if (pCmd->command != TSDB_SQL_RETRIEVE_LOCALMERGE && pCmd->command < TSDB_SQL_LOCAL) {
pCmd->command = (pCmd->command > TSDB_SQL_MGMT) ? TSDB_SQL_RETRIEVE : TSDB_SQL_FETCH;
}
tscProcessSql(pSql);
}
......@@ -346,31 +348,32 @@ void tscProcessFetchRow(SSchedMsg *pMsg) {
void tscProcessAsyncRes(SSchedMsg *pMsg) {
SSqlObj *pSql = (SSqlObj *)pMsg->ahandle;
SSqlCmd *pCmd = &pSql->cmd;
// SSqlCmd *pCmd = &pSql->cmd;
SSqlRes *pRes = &pSql->res;
void *taosres = pSql;
// void *taosres = pSql;
// pCmd may be released, so cache pCmd->command
int cmd = pCmd->command;
int code = pRes->code;
// int cmd = pCmd->command;
// int code = pRes->code;
// in case of async insert, restore the user specified callback function
bool shouldFree = tscShouldBeFreed(pSql);
if (cmd == TSDB_SQL_INSERT) {
assert(pSql->fp != NULL);
pSql->fp = pSql->fetchFp;
}
if (pSql->fp) {
(*pSql->fp)(pSql->param, taosres, code);
}
if (shouldFree) {
tscDebug("%p sqlObj is automatically freed in async res", pSql);
tscFreeSqlObj(pSql);
}
// bool shouldFree = tscShouldBeFreed(pSql);
// if (pCmd->command == TSDB_SQL_INSERT) {
// assert(pSql->fp != NULL);
assert(pSql->fp != NULL && pSql->fetchFp != NULL);
// }
// if (pSql->fp) {
pSql->fp = pSql->fetchFp;
(*pSql->fp)(pSql->param, pSql, pRes->code);
// }
// if (shouldFree) {
// tscDebug("%p sqlObj is automatically freed in async res", pSql);
// tscFreeSqlObj(pSql);
// }
}
static void tscProcessAsyncError(SSchedMsg *pMsg) {
......@@ -420,15 +423,15 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
SSqlCmd *pCmd = &pSql->cmd;
SSqlRes *pRes = &pSql->res;
pRes->code = code;
if (code != TSDB_CODE_SUCCESS) {
pRes->code = code;
tscQueueAsyncRes(pSql);
return;
tscError("%p ge tableMeta failed, code:%s", pSql, tstrerror(code));
goto _error;
} else {
tscDebug("%p get tableMeta successfully", pSql);
}
tscDebug("%p get tableMeta successfully", pSql);
if (pSql->pStream == NULL) {
SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
......@@ -453,11 +456,9 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
assert(pParObj->signature == pParObj && trs->subqueryIndex == pTableMetaInfo->vgroupIndex &&
pTableMetaInfo->vgroupIndex >= 0 && pTableMetaInfo->vgroupList != NULL);
if ((code = tscProcessSql(pSql)) == TSDB_CODE_SUCCESS) {
return;
}
goto _error;
// tscProcessSql can add error into async res
tscProcessSql(pSql);
return;
} else { // continue to process normal async query
if (pCmd->parseFinished) {
tscDebug("%p update table meta in local cache, continue to process sql and send corresponding query", pSql);
......@@ -481,26 +482,21 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
return;
} else if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
if (code == TSDB_CODE_SUCCESS) {
/*
* Discard previous built submit blocks, and then parse the sql string again and build up all submit blocks,
* and send the required submit block according to index value in supporter to server.
*/
pSql->fp = pSql->fetchFp; // restore the fp
if ((code = tscHandleInsertRetry(pSql)) == TSDB_CODE_SUCCESS) {
return;
}
}
/*
* Discard previous built submit blocks, and then parse the sql string again and build up all submit blocks,
* and send the required submit block according to index value in supporter to server.
*/
pSql->fp = pSql->fetchFp; // restore the fp
tscHandleInsertRetry(pSql);
} else {// in case of other query type, continue
if ((code = tscProcessSql(pSql)) == TSDB_CODE_SUCCESS) {
return;
}
tscProcessSql(pSql);
}
goto _error;
return;
} else {
tscDebug("%p continue parse sql after get table meta", pSql);
......@@ -538,7 +534,7 @@ void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) {
goto _error;
}
if (code == TSDB_CODE_SUCCESS && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) {
code = tscGetSTableVgroupInfo(pSql, pCmd->clauseIndex);
if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
return;
......
此差异已折叠。
......@@ -2471,7 +2471,7 @@ static bool functionCompatibleCheck(SQueryInfo* pQueryInfo) {
startIdx++;
}
int32_t factor = funcCompatDefList[tscSqlExprGet(pQueryInfo, startIdx)->functionId];
int32_t factor = functionCompatList[tscSqlExprGet(pQueryInfo, startIdx)->functionId];
// diff function cannot be executed with other function
// arithmetic function can be executed with other arithmetic functions
......@@ -2489,7 +2489,7 @@ static bool functionCompatibleCheck(SQueryInfo* pQueryInfo) {
continue;
}
if (funcCompatDefList[functionId] != factor) {
if (functionCompatList[functionId] != factor) {
return false;
}
}
......
......@@ -339,7 +339,7 @@ void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcIpSet *pIpSet) {
}
if (rpcMsg->code != TSDB_CODE_TSC_ACTION_IN_PROGRESS) {
rpcMsg->code = (pRes->code == TSDB_CODE_SUCCESS) ? pRes->numOfRows: pRes->code;
rpcMsg->code = (pRes->code == TSDB_CODE_SUCCESS)? pRes->numOfRows: pRes->code;
bool shouldFree = tscShouldBeFreed(pSql);
(*pSql->fp)(pSql->param, pSql, rpcMsg->code);
......@@ -412,7 +412,7 @@ int tscProcessSql(SSqlObj *pSql) {
return pSql->res.code;
}
} else if (pCmd->command < TSDB_SQL_LOCAL) {
pSql->ipList = tscMgmtIpSet; //?
pSql->ipList = tscMgmtIpSet;
} else { // local handler
return (*tscProcessMsgRsp[pCmd->command])(pSql);
}
......@@ -476,6 +476,8 @@ int tscBuildFetchMsg(SSqlObj *pSql, SSqlInfo *pInfo) {
int32_t vgIndex = pTableMetaInfo->vgroupIndex;
SVgroupsInfo* pVgroupInfo = pTableMetaInfo->vgroupList;
assert(pVgroupInfo->vgroups[vgIndex].vgId > 0 && vgIndex < pTableMetaInfo->vgroupList->numOfVgroups);
pRetrieveMsg->header.vgId = htonl(pVgroupInfo->vgroups[vgIndex].vgId);
} else {
STableMeta* pTableMeta = pTableMetaInfo->pTableMeta;
......@@ -549,6 +551,7 @@ static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char
assert(index >= 0);
if (pTableMetaInfo->vgroupList->numOfVgroups > 0) {
assert(index < pTableMetaInfo->vgroupList->numOfVgroups);
pVgroupInfo = &pTableMetaInfo->vgroupList->vgroups[index];
}
tscDebug("%p query on stable, vgIndex:%d, numOfVgroups:%d", pSql, index, pTableMetaInfo->vgroupList->numOfVgroups);
......@@ -1372,7 +1375,6 @@ static int tscLocalResultCommonBuilder(SSqlObj *pSql, int32_t numOfRes) {
SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex);
pRes->code = TSDB_CODE_SUCCESS;
if (pRes->rspType == 0) {
pRes->numOfRows = numOfRes;
pRes->row = 0;
......
......@@ -481,25 +481,14 @@ static bool tscFreeQhandleInVnode(SSqlObj* pSql) {
if (pRes->code == TSDB_CODE_SUCCESS && pRes->completed == false && !tscIsTwoStageSTableQuery(pQueryInfo, 0) &&
(pCmd->command == TSDB_SQL_SELECT ||
pCmd->command == TSDB_SQL_SHOW ||
pCmd->command == TSDB_SQL_RETRIEVE ||
pCmd->command == TSDB_SQL_FETCH) &&
(pCmd->command == TSDB_SQL_SELECT && pSql->pStream == NULL && pTableMetaInfo->pTableMeta != NULL)) {
pCmd->command == TSDB_SQL_SHOW ||
pCmd->command == TSDB_SQL_RETRIEVE ||
pCmd->command == TSDB_SQL_FETCH) &&
(pSql->pStream == NULL && pTableMetaInfo->pTableMeta != NULL)) {
pCmd->command = (pCmd->command > TSDB_SQL_MGMT) ? TSDB_SQL_RETRIEVE : TSDB_SQL_FETCH;
tscDebug("%p send msg to dnode to free qhandle ASAP, command:%s", pSql, sqlCmd[pCmd->command]);
tscDebug("%p send msg to dnode to free qhandle ASAP, command:%s, ", pSql, sqlCmd[pCmd->command]);
tscProcessSql(pSql);
// in case of sync model query, waits for response and then goes on
// if (pSql->fp == waitForQueryRsp || pSql->fp == waitForRetrieveRsp) {
// sem_wait(&pSql->rspSem);
// tscFreeSqlObj(pSql);
// tscDebug("%p sqlObj is freed by app", pSql);
// } else {
tscDebug("%p sqlObj will be freed while rsp received", pSql);
// }
return true;
}
......
......@@ -1895,9 +1895,11 @@ int32_t tscHandleInsertRetry(SSqlObj* pSql) {
assert(pSupporter->index < pSupporter->pState->numOfTotal);
STableDataBlocks* pTableDataBlock = taosArrayGetP(pCmd->pDataBlocks, pSupporter->index);
pRes->code = tscCopyDataBlockToPayload(pSql, pTableDataBlock);
if (pRes->code != TSDB_CODE_SUCCESS) {
return pRes->code;
int32_t code = tscCopyDataBlockToPayload(pSql, pTableDataBlock);
if ((pRes->code = code)!= TSDB_CODE_SUCCESS) {
tscQueueAsyncRes(pSql);
return code; // here the pSql may have been released already.
}
return tscProcessSql(pSql);
......
......@@ -1648,6 +1648,7 @@ SSqlObj* createSimpleSubObj(SSqlObj* pSql, void (*fp)(), void* param, int32_t cm
}
pNew->fp = fp;
pNew->fetchFp = fp;
pNew->param = param;
pNew->maxRetry = TSDB_MAX_REPLICA;
......@@ -1803,6 +1804,8 @@ SSqlObj* createSubqueryObj(SSqlObj* pSql, int16_t tableIndex, void (*fp)(), void
}
pNew->fp = fp;
pNew->fetchFp = fp;
pNew->param = param;
pNew->maxRetry = TSDB_MAX_REPLICA;
......@@ -2005,7 +2008,7 @@ void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0);
int32_t totalVgroups = pTableMetaInfo->vgroupList->numOfVgroups;
while (++pTableMetaInfo->vgroupIndex < totalVgroups) {
if (++pTableMetaInfo->vgroupIndex < totalVgroups) {
tscDebug("%p results from vgroup index:%d completed, try next:%d. total vgroups:%d. current numOfRes:%" PRId64, pSql,
pTableMetaInfo->vgroupIndex - 1, pTableMetaInfo->vgroupIndex, totalVgroups, pRes->numOfClauseTotal);
......@@ -2041,11 +2044,9 @@ void tscTryQueryNextVnode(SSqlObj* pSql, __async_cb_func_t fp) {
// set the callback function
pSql->fp = fp;
int32_t ret = tscProcessSql(pSql);
if (ret == TSDB_CODE_SUCCESS) {
return;
} else {// todo check for failure
}
tscProcessSql(pSql);
} else {
tscDebug("%p try all %d vnodes, query complete. current numOfRes:%" PRId64, pSql, totalVgroups, pRes->numOfClauseTotal);
}
}
......
......@@ -121,6 +121,7 @@ typedef struct SQueryCostInfo {
uint32_t loadBlockStatis;
uint32_t discardBlocks;
uint64_t elapsedTime;
uint64_t ioTime;
uint64_t computTime;
} SQueryCostInfo;
......
......@@ -125,7 +125,8 @@ typedef struct SArithmeticSupport {
} SArithmeticSupport;
typedef struct SQLPreAggVal {
bool isSet;
bool isSet; // statistics info set or not
bool dataBlockLoaded; // data block is loaded or not
SDataStatis statis;
} SQLPreAggVal;
......@@ -224,25 +225,14 @@ int32_t getResultDataInfo(int32_t dataType, int32_t dataBytes, int32_t functionI
#define IS_SINGLEOUTPUT(x) (((x)&TSDB_FUNCSTATE_SO) != 0)
#define IS_OUTER_FORWARD(x) (((x)&TSDB_FUNCSTATE_OF) != 0)
/*
* the status of one block, used in metric query. all blocks are mixed together,
* we need the status to decide if one block is a first/end/inter block of one meter
*/
enum {
BLK_FILE_BLOCK = 0x1,
BLK_BLOCK_LOADED = 0x2,
BLK_CACHE_BLOCK = 0x4, // in case of cache block, block must be loaded
};
/* determine the real data need to calculated the result */
enum {
BLK_DATA_NO_NEEDED = 0x0,
BLK_DATA_NO_NEEDED = 0x0,
BLK_DATA_STATIS_NEEDED = 0x1,
BLK_DATA_ALL_NEEDED = 0x3,
BLK_DATA_ALL_NEEDED = 0x3,
BLK_DATA_DISCARD = 0x4, // discard current data block since it is not qualified for filter
};
#define SET_DATA_BLOCK_NOT_LOADED(x) ((x) &= (~BLK_BLOCK_LOADED));
typedef struct STwaInfo {
TSKEY lastKey;
int8_t hasResult; // flag to denote has value
......@@ -264,12 +254,9 @@ typedef struct STwaInfo {
/* global sql function array */
extern struct SQLAggFuncElem aAggs[];
/* compatible check array list */
extern int32_t funcCompatDefList[];
bool top_bot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, char *minval, char *maxval);
extern int32_t functionCompatList[]; // compatible check array list
bool stableQueryFunctChanged(int32_t funcId);
bool topbot_datablock_filter(SQLFunctionCtx *pCtx, int32_t functionId, const char *minval, const char *maxval);
void resetResultInfo(SResultInfo *pResInfo);
void setResultInfoBuf(SResultInfo *pResInfo, int32_t size, bool superTable, char* buf);
......
......@@ -1358,6 +1358,8 @@ void setExecParams(SQuery *pQuery, SQLFunctionCtx *pCtx, void* inputData, TSKEY
pCtx->preAggVals.isSet = false;
}
pCtx->preAggVals.dataBlockLoaded = (inputData != NULL);
// limit/offset query will affect this value
pCtx->startOffset = QUERY_IS_ASC_QUERY(pQuery) ? pQuery->pos:0;
pCtx->size = QUERY_IS_ASC_QUERY(pQuery) ? pBlockInfo->rows - pQuery->pos : pQuery->pos + 1;
......@@ -1928,73 +1930,46 @@ char *getPosInResultPage(SQueryRuntimeEnv *pRuntimeEnv, int32_t columnIndex, SWi
pQuery->pSelectExpr[columnIndex].bytes * realRowId;
}
/**
* decrease the refcount for each table involved in this query
* @param pQInfo
*/
UNUSED_FUNC void vnodeDecMeterRefcnt(SQInfo *pQInfo) {
if (pQInfo != NULL) {
// assert(taosHashGetSize(pQInfo->tableqinfoGroupInfo) >= 1);
}
#if 0
if (pQInfo == NULL || pQInfo->tableqinfoGroupInfo.numOfTables == 1) {
atomic_fetch_sub_32(&pQInfo->pObj->numOfQueries, 1);
qDebug("QInfo:%p vid:%d sid:%d meterId:%s, query is over, numOfQueries:%d", pQInfo, pQInfo->pObj->vnode,
pQInfo->pObj->sid, pQInfo->pObj->meterId, pQInfo->pObj->numOfQueries);
} else {
int32_t num = 0;
for (int32_t i = 0; i < pQInfo->tableqinfoGroupInfo.numOfTables; ++i) {
SMeterObj *pMeter = getMeterObj(pQInfo->tableqinfoGroupInfo, pQInfo->pSidSet->pTableIdList[i]->sid);
atomic_fetch_sub_32(&(pMeter->numOfQueries), 1);
if (pMeter->numOfQueries > 0) {
qDebug("QInfo:%p vid:%d sid:%d meterId:%s, query is over, numOfQueries:%d", pQInfo, pMeter->vnode, pMeter->sid,
pMeter->meterId, pMeter->numOfQueries);
num++;
}
}
/*
* in order to reduce log output, for all meters of which numOfQueries count are 0,
* we do not output corresponding information
*/
num = pQInfo->tableqinfoGroupInfo.numOfTables - num;
qDebug("QInfo:%p metric query is over, dec query ref for %d meters, numOfQueries on %d meters are 0", pQInfo,
pQInfo->tableqinfoGroupInfo.numOfTables, num);
}
#endif
}
#define IS_PREFILTER_TYPE(_t) ((_t) != TSDB_DATA_TYPE_BINARY && (_t) != TSDB_DATA_TYPE_NCHAR)
static bool needToLoadDataBlock(SQuery *pQuery, SDataStatis *pDataStatis, SQLFunctionCtx *pCtx,
int32_t numOfTotalPoints) {
if (pDataStatis == NULL) {
static bool needToLoadDataBlock(SQueryRuntimeEnv* pRuntimeEnv, SDataStatis *pDataStatis, SQLFunctionCtx *pCtx,
int32_t numOfRows) {
SQuery* pQuery = pRuntimeEnv->pQuery;
if (pDataStatis == NULL || (pQuery->numOfFilterCols == 0 && (!pRuntimeEnv->topBotQuery))) {
return true;
}
#if 0
for (int32_t k = 0; k < pQuery->numOfFilterCols; ++k) {
SSingleColumnFilterInfo *pFilterInfo = &pQuery->pFilterInfo[k];
int32_t colIndex = pFilterInfo->info.colIndex;
// this column not valid in current data block
if (colIndex < 0 || pDataStatis[colIndex].colId != pFilterInfo->info.data.colId) {
continue;
int32_t index = -1;
for(int32_t i = 0; i < pQuery->numOfCols; ++i) {
if (pDataStatis[i].colId == pFilterInfo->info.colId) {
index = i;
break;
}
}
// no statistics data
if (index == -1) {
return true;
}
// not support pre-filter operation on binary/nchar data type
if (!vnodeSupportPrefilter(pFilterInfo->info.data.type)) {
continue;
if (!IS_PREFILTER_TYPE(pFilterInfo->info.type)) {
return true;
}
// all points in current column are NULL, no need to check its boundary value
if (pDataStatis[colIndex].numOfNull == numOfTotalPoints) {
if (pDataStatis[index].numOfNull == numOfRows) {
continue;
}
if (pFilterInfo->info.info.type == TSDB_DATA_TYPE_FLOAT) {
float minval = *(double *)(&pDataStatis[colIndex].min);
float maxval = *(double *)(&pDataStatis[colIndex].max);
SDataStatis* pDataBlockst = &pDataStatis[index];
if (pFilterInfo->info.type == TSDB_DATA_TYPE_FLOAT) {
float minval = *(double *)(&pDataBlockst->min);
float maxval = *(double *)(&pDataBlockst->max);
for (int32_t i = 0; i < pFilterInfo->numOfFilters; ++i) {
if (pFilterInfo->pFilters[i].fp(&pFilterInfo->pFilters[i], (char *)&minval, (char *)&maxval)) {
......@@ -2003,53 +1978,50 @@ static bool needToLoadDataBlock(SQuery *pQuery, SDataStatis *pDataStatis, SQLFun
}
} else {
for (int32_t i = 0; i < pFilterInfo->numOfFilters; ++i) {
if (pFilterInfo->pFilters[i].fp(&pFilterInfo->pFilters[i], (char *)&pDataStatis[colIndex].min,
(char *)&pDataStatis[colIndex].max)) {
if (pFilterInfo->pFilters[i].fp(&pFilterInfo->pFilters[i], (char *)&pDataBlockst->min, (char *)&pDataBlockst->max)) {
return true;
}
}
}
}
// todo disable this opt code block temporarily
// for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
// int32_t functId = pQuery->pSelectExpr[i].base.functionId;
// if (functId == TSDB_FUNC_TOP || functId == TSDB_FUNC_BOTTOM) {
// return top_bot_datablock_filter(&pCtx[i], functId, (char *)&pField[i].min, (char *)&pField[i].max);
// }
// }
if (pRuntimeEnv->topBotQuery) {
for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
int32_t functionId = pQuery->pSelectExpr[i].base.functionId;
if (functionId == TSDB_FUNC_TOP || functionId == TSDB_FUNC_BOTTOM) {
return topbot_datablock_filter(&pCtx[i], functionId, (char *)&pDataStatis[i].min, (char *)&pDataStatis[i].max);
}
}
}
#endif
return true;
return false;
}
SArray *loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, void* pQueryHandle, SDataBlockInfo* pBlockInfo, SDataStatis **pStatis) {
int32_t loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, void* pQueryHandle, SDataBlockInfo* pBlockInfo, SDataStatis **pStatis, SArray** pDataBlock) {
SQuery *pQuery = pRuntimeEnv->pQuery;
uint32_t r = 0;
SArray * pDataBlock = NULL;
uint32_t status = 0;
if (pQuery->numOfFilterCols > 0) {
r = BLK_DATA_ALL_NEEDED;
} else {
// check if this data block is required to load
status = BLK_DATA_ALL_NEEDED;
} else { // check if this data block is required to load
for (int32_t i = 0; i < pQuery->numOfOutput; ++i) {
SSqlFuncMsg* pSqlFunc = &pQuery->pSelectExpr[i].base;
int32_t functionId = pSqlFunc->functionId;
int32_t colId = pSqlFunc->colInfo.colId;
r |= aAggs[functionId].dataReqFunc(&pRuntimeEnv->pCtx[i], pQuery->window.skey, pQuery->window.ekey, colId);
status |= aAggs[functionId].dataReqFunc(&pRuntimeEnv->pCtx[i], pBlockInfo->window.skey, pBlockInfo->window.ekey, colId);
}
if (pRuntimeEnv->pTSBuf > 0 || QUERY_IS_INTERVAL_QUERY(pQuery)) {
r |= BLK_DATA_ALL_NEEDED;
status |= BLK_DATA_ALL_NEEDED;
}
}
if (r == BLK_DATA_NO_NEEDED) {
qDebug("QInfo:%p data block discard, rows:%d", GET_QINFO_ADDR(pRuntimeEnv), pBlockInfo->rows);
if (status == BLK_DATA_NO_NEEDED) {
qDebug("QInfo:%p data block discard, brange:%"PRId64 "-%"PRId64", rows:%d", GET_QINFO_ADDR(pRuntimeEnv),
pBlockInfo->window.skey, pBlockInfo->window.ekey, pBlockInfo->rows);
pRuntimeEnv->summary.discardBlocks += 1;
} else if (r == BLK_DATA_STATIS_NEEDED) {
} else if (status == BLK_DATA_STATIS_NEEDED) {
if (tsdbRetrieveDataBlockStatisInfo(pQueryHandle, pStatis) != TSDB_CODE_SUCCESS) {
// return DISK_DATA_LOAD_FAILED;
}
......@@ -2057,32 +2029,34 @@ SArray *loadDataBlockOnDemand(SQueryRuntimeEnv *pRuntimeEnv, void* pQueryHandle,
pRuntimeEnv->summary.loadBlockStatis += 1;
if (*pStatis == NULL) { // data block statistics does not exist, load data block
pDataBlock = tsdbRetrieveDataBlock(pQueryHandle, NULL);
*pDataBlock = tsdbRetrieveDataBlock(pQueryHandle, NULL);
pRuntimeEnv->summary.totalCheckedRows += pBlockInfo->rows;
}
} else {
assert(r == BLK_DATA_ALL_NEEDED);
assert(status == BLK_DATA_ALL_NEEDED);
// load the data block statistics to perform further filter
pRuntimeEnv->summary.loadBlockStatis +=1;
pRuntimeEnv->summary.loadBlockStatis += 1;
if (tsdbRetrieveDataBlockStatisInfo(pQueryHandle, pStatis) != TSDB_CODE_SUCCESS) {
}
if (!needToLoadDataBlock(pQuery,*pStatis, pRuntimeEnv->pCtx, pBlockInfo->rows)) {
if (!needToLoadDataBlock(pRuntimeEnv, *pStatis, pRuntimeEnv->pCtx, pBlockInfo->rows)) {
#if defined(_DEBUG_VIEW)
qDebug("QInfo:%p block discarded by per-filter", GET_QINFO_ADDR(pRuntimeEnv));
#endif
// current block has been discard due to filter applied
pRuntimeEnv->summary.discardBlocks += 1;
// return DISK_DATA_DISCARDED;
qDebug("QInfo:%p data block discard, brange:%"PRId64 "-%"PRId64", rows:%d", GET_QINFO_ADDR(pRuntimeEnv),
pBlockInfo->window.skey, pBlockInfo->window.ekey, pBlockInfo->rows);
return BLK_DATA_DISCARD;
}
pRuntimeEnv->summary.totalCheckedRows += pBlockInfo->rows;
pRuntimeEnv->summary.loadBlocks += 1;
pDataBlock = tsdbRetrieveDataBlock(pQueryHandle, NULL);
*pDataBlock = tsdbRetrieveDataBlock(pQueryHandle, NULL);
}
return pDataBlock;
return TSDB_CODE_SUCCESS;
}
int32_t binarySearchForKey(char *pValue, int num, TSKEY key, int order) {
......@@ -2225,13 +2199,13 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
pQuery->order.order);
TsdbQueryHandleT pQueryHandle = IS_MASTER_SCAN(pRuntimeEnv)? pRuntimeEnv->pQueryHandle : pRuntimeEnv->pSecQueryHandle;
int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order);
SDataBlockInfo blockInfo = SDATA_BLOCK_INITIALIZER;
while (tsdbNextDataBlock(pQueryHandle)) {
summary->totalBlocks += 1;
if (IS_QUERY_KILLED(GET_QINFO_ADDR(pRuntimeEnv))) {
finalizeQueryResult(pRuntimeEnv); // clean up allocated resource during query
longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED);
}
......@@ -2259,7 +2233,11 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
ensureOutputBuffer(pRuntimeEnv, &blockInfo);
SDataStatis *pStatis = NULL;
SArray *pDataBlock = loadDataBlockOnDemand(pRuntimeEnv, pQueryHandle, &blockInfo, &pStatis);
SArray *pDataBlock = NULL;
if (loadDataBlockOnDemand(pRuntimeEnv, pQueryHandle, &blockInfo, &pStatis, &pDataBlock) == BLK_DATA_DISCARD) {
pQuery->current->lastKey = QUERY_IS_ASC_QUERY(pQuery)? blockInfo.window.ekey + step:blockInfo.window.skey + step;
continue;
}
// query start position can not move into tableApplyFunctionsOnBlock due to limit/offset condition
pQuery->pos = QUERY_IS_ASC_QUERY(pQuery)? 0 : blockInfo.rows - 1;
......@@ -2282,8 +2260,6 @@ static int64_t doScanAllDataBlocks(SQueryRuntimeEnv *pRuntimeEnv) {
if (QUERY_IS_INTERVAL_QUERY(pQuery) && IS_MASTER_SCAN(pRuntimeEnv)) {
if (Q_STATUS_EQUAL(pQuery->status, QUERY_COMPLETED)) {
// int32_t step = QUERY_IS_ASC_QUERY(pQuery) ? QUERY_ASC_FORWARD_STEP : QUERY_DESC_FORWARD_STEP;
closeAllTimeWindow(&pRuntimeEnv->windowResInfo);
// removeRedundantWindow(&pRuntimeEnv->windowResInfo, pTableQueryInfo->lastKey - step, step);
pRuntimeEnv->windowResInfo.curIndex = pRuntimeEnv->windowResInfo.size - 1; // point to the last time window
......@@ -3393,12 +3369,10 @@ void destroyTableQueryInfo(STableQueryInfo *pTableQueryInfo, int32_t numOfCols)
cleanupTimeWindowInfo(&pTableQueryInfo->windowResInfo, numOfCols);
}
#define SET_CURRENT_QUERY_TABLE_INFO(_runtime, _tableInfo) \
do { \
SQuery *_query = (_runtime)->pQuery; \
_query->current = _tableInfo; \
assert((((_tableInfo)->lastKey >= (_tableInfo)->win.skey) && QUERY_IS_ASC_QUERY(_query)) || \
(((_tableInfo)->lastKey <= (_tableInfo)->win.skey) && !QUERY_IS_ASC_QUERY(_query))); \
#define CHECK_QUERY_TIME_RANGE(_q, _tableInfo) \
do { \
assert((((_tableInfo)->lastKey >= (_tableInfo)->win.skey) && QUERY_IS_ASC_QUERY(_q)) || \
(((_tableInfo)->lastKey <= (_tableInfo)->win.skey) && !QUERY_IS_ASC_QUERY(_q))); \
} while (0)
/**
......@@ -3710,7 +3684,7 @@ static void updateWindowResNumOfRes(SQueryRuntimeEnv *pRuntimeEnv) {
}
}
void stableApplyFunctionsOnBlock(SQueryRuntimeEnv *pRuntimeEnv, SDataBlockInfo *pDataBlockInfo, SDataStatis *pStatis,
static void stableApplyFunctionsOnBlock(SQueryRuntimeEnv *pRuntimeEnv, SDataBlockInfo *pDataBlockInfo, SDataStatis *pStatis,
SArray *pDataBlock, __block_search_fn_t searchFn) {
SQuery * pQuery = pRuntimeEnv->pQuery;
STableQueryInfo* pTableQueryInfo = pQuery->current;
......@@ -3869,9 +3843,10 @@ static void queryCostStatis(SQInfo *pQInfo) {
// pQInfo, pSummary->readDiskBlocks, pSummary->totalBlockSize, pSummary->loadBlocksUs / 1000.0,
// pSummary->skippedFileBlocks, pSummary->totalGenData);
qDebug("QInfo:%p :cost summary: elpased time:%"PRId64" us, total blocks:%d, use block statis:%d, use block data:%d, "
"total rows:%"PRId64 ", check rows:%"PRId64, pQInfo, pSummary->elapsedTime, pSummary->totalBlocks,
pSummary->loadBlockStatis, pSummary->loadBlocks, pSummary->totalRows, pSummary->totalCheckedRows);
qDebug("QInfo:%p :cost summary: elapsed time:%"PRId64" us, io time:%"PRId64" us, total blocks:%d, load block statis:%d,"
" load data block:%d, total rows:%"PRId64 ", check rows:%"PRId64,
pQInfo, pSummary->elapsedTime, pSummary->ioTime, pSummary->totalBlocks, pSummary->loadBlockStatis,
pSummary->loadBlocks, pSummary->totalRows, pSummary->totalCheckedRows);
// qDebug("QInfo:%p cost: temp file:%d Bytes", pQInfo, pSummary->tmpBufferInDisk);
//
......@@ -4253,6 +4228,23 @@ static void enableExecutionForNextTable(SQueryRuntimeEnv *pRuntimeEnv) {
}
}
static FORCE_INLINE void setEnvForEachBlock(SQInfo* pQInfo, STableQueryInfo* pTableQueryInfo, SDataBlockInfo* pBlockInfo) {
SQueryRuntimeEnv* pRuntimeEnv = &pQInfo->runtimeEnv;
SQuery* pQuery = pQInfo->runtimeEnv.pQuery;
int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order);
if (!QUERY_IS_INTERVAL_QUERY(pQuery)) {
setExecutionContext(pQInfo, pTableQueryInfo->groupIndex, pBlockInfo->window.ekey + step);
} else { // interval query
TSKEY nextKey = pBlockInfo->window.skey;
setIntervalQueryRange(pQInfo, nextKey);
if (pRuntimeEnv->hasTagResults || pRuntimeEnv->pTSBuf != NULL) {
setAdditionalInfo(pQInfo, pTableQueryInfo->pTable, pTableQueryInfo);
}
}
}
static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) {
SQueryRuntimeEnv *pRuntimeEnv = &pQInfo->runtimeEnv;
SQuery* pQuery = pRuntimeEnv->pQuery;
......@@ -4263,10 +4255,12 @@ static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) {
TsdbQueryHandleT pQueryHandle = IS_MASTER_SCAN(pRuntimeEnv)? pRuntimeEnv->pQueryHandle : pRuntimeEnv->pSecQueryHandle;
SDataBlockInfo blockInfo = SDATA_BLOCK_INITIALIZER;
int32_t step = GET_FORWARD_DIRECTION_FACTOR(pQuery->order.order);
while (tsdbNextDataBlock(pQueryHandle)) {
summary->totalBlocks += 1;
if (IS_QUERY_KILLED(pQInfo)) {
finalizeQueryResult(pRuntimeEnv); // clean up allocated resource during query
longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED);
}
......@@ -4276,24 +4270,19 @@ static int64_t scanMultiTableDataBlocks(SQInfo *pQInfo) {
break;
}
assert(*pTableQueryInfo != NULL);
SET_CURRENT_QUERY_TABLE_INFO(pRuntimeEnv, *pTableQueryInfo);
pQuery->current = *pTableQueryInfo;
CHECK_QUERY_TIME_RANGE(pQuery, *pTableQueryInfo);
if (!pRuntimeEnv->groupbyNormalCol) {
setEnvForEachBlock(pQInfo, *pTableQueryInfo, &blockInfo);
}
SDataStatis *pStatis = NULL;
SArray *pDataBlock = loadDataBlockOnDemand(pRuntimeEnv, pQueryHandle, &blockInfo, &pStatis);
SArray *pDataBlock = NULL;
if (!pRuntimeEnv->groupbyNormalCol) {
if (!QUERY_IS_INTERVAL_QUERY(pQuery)) {
int32_t step = QUERY_IS_ASC_QUERY(pQuery)? 1:-1;
setExecutionContext(pQInfo, (*pTableQueryInfo)->groupIndex, blockInfo.window.ekey + step);
} else { // interval query
TSKEY nextKey = blockInfo.window.skey;
setIntervalQueryRange(pQInfo, nextKey);
if (pRuntimeEnv->hasTagResults || pRuntimeEnv->pTSBuf != NULL) {
setAdditionalInfo(pQInfo, (*pTableQueryInfo)->pTable, *pTableQueryInfo);
}
}
if (loadDataBlockOnDemand(pRuntimeEnv, pQueryHandle, &blockInfo, &pStatis, &pDataBlock) == BLK_DATA_DISCARD) {
pQuery->current->lastKey = QUERY_IS_ASC_QUERY(pQuery)? blockInfo.window.ekey + step:blockInfo.window.skey + step;
continue;
}
summary->totalRows += blockInfo.rows;
......@@ -4553,7 +4542,6 @@ static void sequentialTableProcess(SQInfo *pQInfo) {
while (pQInfo->tableIndex < pQInfo->tableqinfoGroupInfo.numOfTables) {
if (IS_QUERY_KILLED(pQInfo)) {
finalizeQueryResult(pRuntimeEnv); // clean up allocated resource during query
longjmp(pRuntimeEnv->env, TSDB_CODE_TSC_QUERY_CANCELLED);
}
......@@ -5051,6 +5039,7 @@ static void stableQueryImpl(SQInfo *pQInfo) {
isFirstLastRowQuery(pQuery) || pRuntimeEnv->groupbyNormalCol);
sequentialTableProcess(pQInfo);
}
// record the total elapsed time
......@@ -6176,11 +6165,6 @@ _over:
//pQInfo already freed in initQInfo, but *pQInfo may not pointer to null;
if (code != TSDB_CODE_SUCCESS) {
*pQInfo = NULL;
} else {
// SQInfo* pq = (SQInfo*) (*pQInfo);
// T_REF_INC(pq);
// T_REF_INC(pq);
}
// if failed to add ref for all meters in this query, abort current query
......
......@@ -1801,7 +1801,8 @@ int32_t tsdbRetrieveDataBlockStatisInfo(TsdbQueryHandleT* pQueryHandle, SDataSta
}
tsdbLoadCompData(&pHandle->rhelper, pBlockInfo->compBlock, NULL);
// todo opt perf
size_t numOfCols = QH_GET_NUM_OF_COLS(pHandle);
for(int32_t i = 0; i < numOfCols; ++i) {
SDataStatis* st = &pHandle->statis[i];
......@@ -1820,6 +1821,13 @@ int32_t tsdbRetrieveDataBlockStatisInfo(TsdbQueryHandleT* pQueryHandle, SDataSta
if (pHandle->statis[i].numOfNull == -1) { // set the column data are all NULL
pHandle->statis[i].numOfNull = pBlockInfo->compBlock->numOfRows;
}
// todo opt perf
SColumnInfo* pColInfo = taosArrayGet(pHandle->pColumns, i);
if (pColInfo->type == TSDB_DATA_TYPE_TIMESTAMP) {
pHandle->statis[i].min = pBlockInfo->compBlock->keyFirst;
pHandle->statis[i].max = pBlockInfo->compBlock->keyLast;
}
}
return TSDB_CODE_SUCCESS;
......
......@@ -413,57 +413,90 @@ void taosCacheRelease(SCacheObj *pCacheObj, void **data, bool _remove) {
*data = NULL;
// note: extend lifespan before dec ref count
if (pCacheObj->extendLifespan) {
bool inTrashCan = pNode->inTrashCan;
if (pCacheObj->extendLifespan && (!inTrashCan)) {
atomic_store_64(&pNode->expireTime, pNode->lifespan + taosGetTimestampMs());
uDebug("cache:%s data:%p extend life time to %"PRId64 " before release", pCacheObj->name, pNode->data, pNode->expireTime);
}
bool inTrashCan = pNode->inTrashCan;
uDebug("cache:%s, key:%p, %p is released, refcnt:%d", pCacheObj->name, pNode->key, pNode->data, T_REF_VAL_GET(pNode) - 1);
// NOTE: once refcount is decrease, pNode may be free by other thread immediately.
int32_t ref = T_REF_DEC(pNode);
if (inTrashCan) {
// Remove it if the ref count is 0.
// The ref count does not need to load and check again after lock acquired, since ref count can not be increased when
// the node is in trashcan.
if (ref == 0) {
__cache_wr_lock(pCacheObj);
assert(pNode->pTNodeHeader->pData == pNode);
taosRemoveFromTrashCan(pCacheObj, pNode->pTNodeHeader);
__cache_unlock(pCacheObj);
if (_remove) {
__cache_wr_lock(pCacheObj);
// NOTE: once refcount is decrease, pNode may be freed by other thread immediately.
int32_t ref = T_REF_DEC(pNode);
uDebug("cache:%s, key:%p, %p is released, refcnt:%d", pCacheObj->name, pNode->key, pNode->data, ref);
/*
* If it is not referenced by other users, remove it immediately. Otherwise move this node to trashcan wait for all users
* releasing this resources.
*
* NOTE: previous ref is 0, and current ref is still 0, remove it. If previous is not 0, there is another thread
* that tries to do the same thing.
*/
if (pNode->inTrashCan) {
if (ref == 0) {
assert(pNode->pTNodeHeader->pData == pNode);
taosRemoveFromTrashCan(pCacheObj, pNode->pTNodeHeader);
}
} else {
if (ref > 0) {
assert(pNode->pTNodeHeader == NULL);
taosCacheMoveToTrash(pCacheObj, pNode);
} else {
taosCacheReleaseNode(pCacheObj, pNode);
}
}
__cache_unlock(pCacheObj);
} else {
assert(pNode->pTNodeHeader == NULL);
if (_remove) { // not in trash can, but need to remove it
__cache_wr_lock(pCacheObj);
/*
* If not referenced by other users. Otherwise move this node to trashcan wait for all users
* releasing this resources.
*
* NOTE: previous ref is 0, and current ref is still 0, remove it. If previous is not 0, there is another thread
* that tries to do the same thing.
*/
uDebug("cache:%s, key:%p, %p is released, refcnt:%d", pCacheObj->name, pNode->key, pNode->data, T_REF_VAL_GET(pNode) - 1);
// NOTE: once refcount is decrease, pNode may be freed by other thread immediately.
int32_t ref = T_REF_DEC(pNode);
if (inTrashCan) {
// Remove it if the ref count is 0.
// The ref count does not need to load and check again after lock acquired, since ref count can not be increased when
// the node is in trashcan.
if (ref == 0) {
if (T_REF_VAL_GET(pNode) == 0) {
taosCacheReleaseNode(pCacheObj, pNode);
} else {
taosCacheMoveToTrash(pCacheObj, pNode);
}
__cache_wr_lock(pCacheObj);
assert(pNode->pTNodeHeader->pData == pNode);
taosRemoveFromTrashCan(pCacheObj, pNode->pTNodeHeader);
__cache_unlock(pCacheObj);
}
__cache_unlock(pCacheObj);
// } else { // extend its life time
// if (pCacheObj->extendLifespan) {
// atomic_store_64(&pNode->expireTime, pNode->lifespan + taosGetTimestampMs());
// uDebug("cache:%s data:%p extend life time to %"PRId64 " after release", pCacheObj->name, pNode->data, pNode->expireTime);
// }
}
}
// else {
// if (_remove) { // not in trash can, but need to remove it
// __cache_wr_lock(pCacheObj);
//
// /*
// * If not referenced by other users. Otherwise move this node to trashcan wait for all users
// * releasing this resources.
// *
// * NOTE: previous ref is 0, and current ref is still 0, remove it. If previous is not 0, there is another thread
// * that tries to do the same thing.
// */
// if (ref == 0) {
// if (T_REF_VAL_GET(pNode) == 0) {
// taosCacheReleaseNode(pCacheObj, pNode);
// } else {
// taosCacheMoveToTrash(pCacheObj, pNode);
// }
// } else if (ref > 0) {
// if (!pNode->inTrashCan) {
// assert(pNode->pTNodeHeader == NULL);
// taosCacheMoveToTrash(pCacheObj, pNode);
// }
// }
//
// __cache_unlock(pCacheObj);
// }
// }
}
void taosCacheEmpty(SCacheObj *pCacheObj) {
......
......@@ -110,6 +110,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SReadMsg *pReadMsg) {
if (handle == NULL) { // failed to register qhandle
pRsp->code = TSDB_CODE_QRY_INVALID_QHANDLE;
qDestroyQueryInfo(pQInfo); // destroy it directly
vError("vgId:%d QInfo:%p register qhandle failed, return to app, code:%s", pVnode->vgId, (void*) pQInfo, tstrerror(pRsp->code));
} else {
assert(*handle == pQInfo);
pRsp->qhandle = htobe64((uint64_t) pQInfo);
......@@ -125,12 +126,14 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SReadMsg *pReadMsg) {
} else {
assert(pQInfo == NULL);
}
if (handle != NULL) {
vDebug("vgId:%d, QInfo:%p, dnode query msg disposed, register qhandle and return to app", vgId, *handle);
dnodePutItemIntoReadQueue(pVnode, *handle);
qReleaseQInfo(pVnode->qMgmt, (void**) &handle, false);
}
vDebug("vgId:%d, QInfo:%p, dnode query msg disposed", vgId, pQInfo);
} else {
assert(pCont != NULL);
handle = qAcquireQInfo(pVnode->qMgmt, (uint64_t) pCont);
......@@ -138,12 +141,13 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SReadMsg *pReadMsg) {
vWarn("QInfo:%p invalid qhandle in continuing exec query, conn:%p", (void*) pCont, pReadMsg->rpcMsg.handle);
code = TSDB_CODE_QRY_INVALID_QHANDLE;
} else {
vDebug("vgId:%d, QInfo:%p, dnode query msg in progress", pVnode->vgId, (void*) pCont);
vDebug("vgId:%d, QInfo:%p, dnode continue exec query", pVnode->vgId, (void*) pCont);
code = TSDB_CODE_VND_ACTION_IN_PROGRESS;
qTableQuery(*handle); // do execute query
}
qReleaseQInfo(pVnode->qMgmt, (void**) &handle, false);
}
return code;
}
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册