/* * Copyright (c) 2019 TAOS Data, Inc. * * This program is free software: you can use, redistribute, and/or modify * it under the terms of the GNU Affero General Public License, version 3 * or later ("AGPL"), as published by the Free Software Foundation. * * 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. * * You should have received a copy of the GNU Affero General Public License * along with this program. If not, see . */ #include "os.h" #include "tcache.h" #include "tcmdtype.h" #include "trpc.h" #include "tscLocalMerge.h" #include "tscLog.h" #include "tscProfile.h" #include "tscUtil.h" #include "tschemautil.h" #include "tsclient.h" #include "ttimer.h" #include "tutil.h" #include "tlockfree.h" #define TSC_MGMT_VNODE 999 SRpcCorEpSet tscMgmtEpSet; SRpcEpSet tscDnodeEpSet; int (*tscBuildMsg[TSDB_SQL_MAX])(SSqlObj *pSql, SSqlInfo *pInfo) = {0}; int (*tscProcessMsgRsp[TSDB_SQL_MAX])(SSqlObj *pSql); void tscProcessActivityTimer(void *handle, void *tmrId); int tscKeepConn[TSDB_SQL_MAX] = {0}; TSKEY tscGetSubscriptionProgress(void* sub, int64_t uid, TSKEY dflt); void tscUpdateSubscriptionProgress(void* sub, int64_t uid, TSKEY ts); void tscSaveSubscriptionProgress(void* sub); static int32_t minMsgSize() { return tsRpcHeadSize + 100; } static int32_t getWaitingTimeInterval(int32_t count) { int32_t initial = 100; // 100 ms by default if (count <= 1) { return 0; } return initial * (2<<(count - 2)); } static void tscSetDnodeEpSet(SSqlObj* pSql, SCMVgroupInfo* pVgroupInfo) { assert(pSql != NULL && pVgroupInfo != NULL && pVgroupInfo->numOfEps > 0); SRpcEpSet* pEpSet = &pSql->epSet; pEpSet->inUse = 0; // apply the FQDN string length check here bool hasFqdn = false; pEpSet->numOfEps = pVgroupInfo->numOfEps; for(int32_t i = 0; i < pVgroupInfo->numOfEps; ++i) { strcpy(pEpSet->fqdn[i], pVgroupInfo->epAddr[i].fqdn); pEpSet->port[i] = pVgroupInfo->epAddr[i].port; if (!hasFqdn) { hasFqdn = (strlen(pEpSet->fqdn[i]) > 0); } } assert(hasFqdn); } static void tscDumpMgmtEpSet(SRpcEpSet *epSet) { taosCorBeginRead(&tscMgmtEpSet.version); *epSet = tscMgmtEpSet.epSet; taosCorEndRead(&tscMgmtEpSet.version); } static void tscEpSetHtons(SRpcEpSet *s) { for (int32_t i = 0; i < s->numOfEps; i++) { s->port[i] = htons(s->port[i]); } } bool tscEpSetIsEqual(SRpcEpSet *s1, SRpcEpSet *s2) { if (s1->numOfEps != s2->numOfEps || s1->inUse != s2->inUse) { return false; } for (int32_t i = 0; i < s1->numOfEps; i++) { if (s1->port[i] != s2->port[i] || strncmp(s1->fqdn[i], s2->fqdn[i], TSDB_FQDN_LEN) != 0) return false; } return true; } void tscUpdateMgmtEpSet(SRpcEpSet *pEpSet) { // no need to update if equal taosCorBeginWrite(&tscMgmtEpSet.version); tscMgmtEpSet.epSet = *pEpSet; taosCorEndWrite(&tscMgmtEpSet.version); } static void tscDumpEpSetFromVgroupInfo(SCMCorVgroupInfo *pVgroupInfo, SRpcEpSet *pEpSet) { if (pVgroupInfo == NULL) { return;} taosCorBeginRead(&pVgroupInfo->version); int8_t inUse = pVgroupInfo->inUse; pEpSet->inUse = (inUse >= 0 && inUse < TSDB_MAX_REPLICA) ? inUse: 0; pEpSet->numOfEps = pVgroupInfo->numOfEps; for (int32_t i = 0; i < pVgroupInfo->numOfEps; ++i) { tstrncpy(pEpSet->fqdn[i], pVgroupInfo->epAddr[i].fqdn, sizeof(pEpSet->fqdn[i])); pEpSet->port[i] = pVgroupInfo->epAddr[i].port; } taosCorEndRead(&pVgroupInfo->version); } static void tscUpdateVgroupInfo(SSqlObj *pObj, SRpcEpSet *pEpSet) { SSqlCmd *pCmd = &pObj->cmd; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); if (pTableMetaInfo == NULL || pTableMetaInfo->pTableMeta == NULL) { return;} SCMCorVgroupInfo *pVgroupInfo = &pTableMetaInfo->pTableMeta->corVgroupInfo; taosCorBeginWrite(&pVgroupInfo->version); tscDebug("before: Endpoint in use: %d", pVgroupInfo->inUse); pVgroupInfo->inUse = pEpSet->inUse; pVgroupInfo->numOfEps = pEpSet->numOfEps; for (int32_t i = 0; i < pVgroupInfo->numOfEps; i++) { tstrncpy(pVgroupInfo->epAddr[i].fqdn, pEpSet->fqdn[i], TSDB_FQDN_LEN); pVgroupInfo->epAddr[i].port = pEpSet->port[i]; } tscDebug("after: EndPoint in use: %d", pVgroupInfo->inUse); taosCorEndWrite(&pVgroupInfo->version); } void tscPrintMgmtEp() { SRpcEpSet dump; tscDumpMgmtEpSet(&dump); if (dump.numOfEps <= 0) { tscError("invalid mnode EP list:%d", dump.numOfEps); } else { for (int i = 0; i < dump.numOfEps; ++i) { tscDebug("mnode index:%d %s:%d", i, dump.fqdn[i], dump.port[i]); } } } void tscProcessHeartBeatRsp(void *param, TAOS_RES *tres, int code) { STscObj *pObj = (STscObj *)param; if (pObj == NULL) return; if (pObj != pObj->signature) { tscError("heart beat msg, pObj:%p, signature:%p invalid", pObj, pObj->signature); return; } SSqlObj *pSql = pObj->pHb; SSqlRes *pRes = &pSql->res; if (code == 0) { SCMHeartBeatRsp *pRsp = (SCMHeartBeatRsp *)pRes->pRsp; SRpcEpSet * epSet = &pRsp->epSet; if (epSet->numOfEps > 0) { tscEpSetHtons(epSet); tscUpdateMgmtEpSet(epSet); } pSql->pTscObj->connId = htonl(pRsp->connId); if (pRsp->killConnection) { tscKillConnection(pObj); return; } else { if (pRsp->queryId) tscKillQuery(pObj, htonl(pRsp->queryId)); if (pRsp->streamId) tscKillStream(pObj, htonl(pRsp->streamId)); } } else { tscDebug("heart beat failed, code:%s", tstrerror(code)); } taosTmrReset(tscProcessActivityTimer, tsShellActivityTimer * 500, pObj, tscTmr, &pObj->pTimer); } void tscProcessActivityTimer(void *handle, void *tmrId) { STscObj *pObj = (STscObj *)handle; if (pObj == NULL || pObj->signature != pObj) { return; } SSqlObj* pHB = pObj->pHb; if (pObj->pTimer != tmrId || pHB == NULL) { return; } if (tscShouldFreeHeartBeat(pHB)) { tscDebug("%p free HB object and release connection", pHB); tscFreeSqlObj(pHB); tscCloseTscObj(pObj); } else { int32_t code = tscProcessSql(pHB); if (code != TSDB_CODE_SUCCESS) { tscError("%p failed to sent HB to server, reason:%s", pHB, tstrerror(code)); } } } int tscSendMsgToServer(SSqlObj *pSql) { STscObj* pObj = pSql->pTscObj; SSqlCmd* pCmd = &pSql->cmd; char *pMsg = rpcMallocCont(pCmd->payloadLen); if (NULL == pMsg) { tscError("%p msg:%s malloc failed", pSql, taosMsg[pSql->cmd.msgType]); return TSDB_CODE_TSC_OUT_OF_MEMORY; } // set the mgmt ip list if (pSql->cmd.command >= TSDB_SQL_MGMT) { tscDumpMgmtEpSet(&pSql->epSet); } memcpy(pMsg, pSql->cmd.payload, pSql->cmd.payloadLen); SRpcMsg rpcMsg = { .msgType = pSql->cmd.msgType, .pCont = pMsg, .contLen = pSql->cmd.payloadLen, .ahandle = pSql, .handle = &pSql->pRpcCtx, .code = 0 }; // NOTE: the rpc context should be acquired before sending data to server. // Otherwise, the pSql object may have been released already during the response function, which is // processMsgFromServer function. In the meanwhile, the assignment of the rpc context to sql object will absolutely // cause crash. rpcSendRequest(pObj->pDnodeConn, &pSql->epSet, &rpcMsg); return TSDB_CODE_SUCCESS; } void tscProcessMsgFromServer(SRpcMsg *rpcMsg, SRpcEpSet *pEpSet) { SSqlObj *pSql = (SSqlObj *)rpcMsg->ahandle; if (pSql == NULL || pSql->signature != pSql) { tscError("%p sql is already released", pSql); return; } STscObj *pObj = pSql->pTscObj; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; if (pObj->signature != pObj) { tscDebug("%p DB connection is closed, cmd:%d pObj:%p signature:%p", pSql, pCmd->command, pObj, pObj->signature); tscFreeSqlObj(pSql); rpcFreeCont(rpcMsg->pCont); return; } pSql->pRpcCtx = NULL; // clear the rpcCtx SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); if (pQueryInfo != NULL && pQueryInfo->type == TSDB_QUERY_TYPE_FREE_RESOURCE) { tscDebug("%p sqlObj needs to be released or DB connection is closed, cmd:%d type:%d, pObj:%p signature:%p", pSql, pCmd->command, pQueryInfo->type, pObj, pObj->signature); tscFreeSqlObj(pSql); rpcFreeCont(rpcMsg->pCont); return; } if (pEpSet) { if (!tscEpSetIsEqual(&pSql->epSet, pEpSet)) { if(pCmd->command < TSDB_SQL_MGMT) { tscUpdateVgroupInfo(pSql, pEpSet); } else { tscUpdateMgmtEpSet(pEpSet); } } } int32_t cmd = pCmd->command; if ((cmd == TSDB_SQL_SELECT || cmd == TSDB_SQL_FETCH || cmd == TSDB_SQL_INSERT || cmd == TSDB_SQL_UPDATE_TAGS_VAL) && (rpcMsg->code == TSDB_CODE_TDB_INVALID_TABLE_ID || rpcMsg->code == TSDB_CODE_VND_INVALID_VGROUP_ID || rpcMsg->code == TSDB_CODE_RPC_NETWORK_UNAVAIL || rpcMsg->code == TSDB_CODE_APP_NOT_READY || rpcMsg->code == TSDB_CODE_TDB_TABLE_RECONFIGURE)) { tscWarn("%p it shall renew table meta, code:%s, retry:%d", pSql, tstrerror(rpcMsg->code), ++pSql->retry); // set the flag to denote that sql string needs to be re-parsed and build submit block with table schema if (rpcMsg->code == TSDB_CODE_TDB_TABLE_RECONFIGURE) { pSql->cmd.submitSchema = 1; } pSql->res.code = rpcMsg->code; // keep the previous error code if (pSql->retry > pSql->maxRetry) { tscError("%p max retry %d reached, give up", pSql, pSql->maxRetry); } else { // wait for a little bit moment and then retry if (rpcMsg->code == TSDB_CODE_APP_NOT_READY || rpcMsg->code == TSDB_CODE_VND_INVALID_VGROUP_ID) { int32_t duration = getWaitingTimeInterval(pSql->retry); taosMsleep(duration); } rpcMsg->code = tscRenewTableMeta(pSql, 0); // if there is an error occurring, proceed to the following error handling procedure. if (rpcMsg->code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) { rpcFreeCont(rpcMsg->pCont); return; } } } pRes->rspLen = 0; if (pRes->code == TSDB_CODE_TSC_QUERY_CANCELLED) { tscDebug("%p query is cancelled, code:%s", pSql, tstrerror(pRes->code)); } else { pRes->code = rpcMsg->code; } if (pRes->code == TSDB_CODE_SUCCESS) { tscDebug("%p reset retry counter to be 0 due to success rsp, old:%d", pSql, pSql->retry); pSql->retry = 0; } if (pRes->code != TSDB_CODE_TSC_QUERY_CANCELLED) { assert(rpcMsg->msgType == pCmd->msgType + 1); pRes->code = rpcMsg->code; pRes->rspType = rpcMsg->msgType; pRes->rspLen = rpcMsg->contLen; if (pRes->rspLen > 0 && rpcMsg->pCont) { char *tmp = (char *)realloc(pRes->pRsp, pRes->rspLen); if (tmp == NULL) { pRes->code = TSDB_CODE_TSC_OUT_OF_MEMORY; } else { pRes->pRsp = tmp; memcpy(pRes->pRsp, rpcMsg->pCont, pRes->rspLen); } } else { pRes->pRsp = NULL; } /* * There is not response callback function for submit response. * The actual inserted number of points is the first number. */ if (rpcMsg->msgType == TSDB_MSG_TYPE_SUBMIT_RSP && pRes->pRsp != NULL) { SShellSubmitRspMsg *pMsg = (SShellSubmitRspMsg*)pRes->pRsp; pMsg->code = htonl(pMsg->code); pMsg->numOfRows = htonl(pMsg->numOfRows); pMsg->affectedRows = htonl(pMsg->affectedRows); pMsg->failedRows = htonl(pMsg->failedRows); pMsg->numOfFailedBlocks = htonl(pMsg->numOfFailedBlocks); pRes->numOfRows += pMsg->affectedRows; tscDebug("%p SQL cmd:%s, code:%s inserted rows:%d rspLen:%d", pSql, sqlCmd[pCmd->command], tstrerror(pRes->code), pMsg->affectedRows, pRes->rspLen); } else { tscDebug("%p SQL cmd:%s, code:%s rspLen:%d", pSql, sqlCmd[pCmd->command], tstrerror(pRes->code), pRes->rspLen); } } if (pRes->code == TSDB_CODE_SUCCESS && tscProcessMsgRsp[pCmd->command]) { rpcMsg->code = (*tscProcessMsgRsp[pCmd->command])(pSql); } if (rpcMsg->code != TSDB_CODE_TSC_ACTION_IN_PROGRESS) { rpcMsg->code = (pRes->code == TSDB_CODE_SUCCESS) ? (int32_t)pRes->numOfRows : pRes->code; bool shouldFree = tscShouldBeFreed(pSql); (*pSql->fp)(pSql->param, pSql, rpcMsg->code); if (shouldFree) { tscDebug("%p sqlObj is automatically freed", pSql); tscFreeSqlObj(pSql); } } rpcFreeCont(rpcMsg->pCont); } int doProcessSql(SSqlObj *pSql) { SSqlCmd *pCmd = &pSql->cmd; SSqlRes *pRes = &pSql->res; if (pCmd->command == TSDB_SQL_SELECT || pCmd->command == TSDB_SQL_FETCH || pCmd->command == TSDB_SQL_RETRIEVE || pCmd->command == TSDB_SQL_INSERT || pCmd->command == TSDB_SQL_CONNECT || pCmd->command == TSDB_SQL_HB || pCmd->command == TSDB_SQL_META || pCmd->command == TSDB_SQL_STABLEVGROUP) { pRes->code = tscBuildMsg[pCmd->command](pSql, NULL); } if (pRes->code != TSDB_CODE_SUCCESS) { tscQueueAsyncRes(pSql); return pRes->code; } int32_t code = tscSendMsgToServer(pSql); // NOTE: if code is TSDB_CODE_SUCCESS, pSql may have been released here already by other threads. if (code != TSDB_CODE_SUCCESS) { pRes->code = code; tscQueueAsyncRes(pSql); return pRes->code; } return TSDB_CODE_SUCCESS; } int tscProcessSql(SSqlObj *pSql) { char *name = NULL; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); STableMetaInfo *pTableMetaInfo = NULL; uint32_t type = 0; if (pQueryInfo != NULL) { pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); name = (pTableMetaInfo != NULL)? pTableMetaInfo->name:NULL; type = pQueryInfo->type; // while numOfTables equals to 0, it must be Heartbeat assert((pQueryInfo->numOfTables == 0 && pQueryInfo->command == TSDB_SQL_HB) || pQueryInfo->numOfTables > 0); } tscDebug("%p SQL cmd:%s will be processed, name:%s, type:%d", pSql, sqlCmd[pCmd->command], name, type); if (pCmd->command < TSDB_SQL_MGMT) { // the pTableMetaInfo cannot be NULL if (pTableMetaInfo == NULL) { pSql->res.code = TSDB_CODE_TSC_APP_ERROR; return pSql->res.code; } } else if (pCmd->command < TSDB_SQL_LOCAL) { //pSql->epSet = tscMgmtEpSet; } else { // local handler return (*tscProcessMsgRsp[pCmd->command])(pSql); } return doProcessSql(pSql); } void tscKillSTableQuery(SSqlObj *pSql) { SSqlCmd* pCmd = &pSql->cmd; SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); if (!tscIsTwoStageSTableQuery(pQueryInfo, 0)) { return; } pSql->res.code = TSDB_CODE_TSC_QUERY_CANCELLED; for (int i = 0; i < pSql->numOfSubs; ++i) { // NOTE: pSub may have been released already here SSqlObj *pSub = pSql->pSubs[i]; if (pSub == NULL) { continue; } pSub->res.code = TSDB_CODE_TSC_QUERY_CANCELLED; if (pSub->pRpcCtx != NULL) { rpcCancelRequest(pSub->pRpcCtx); } tscQueueAsyncRes(pSub); // async res? not other functions? } tscDebug("%p super table query cancelled", pSql); } int tscBuildFetchMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SRetrieveTableMsg *pRetrieveMsg = (SRetrieveTableMsg *) pSql->cmd.payload; pRetrieveMsg->qhandle = htobe64(pSql->res.qhandle); SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, pSql->cmd.clauseIndex); pRetrieveMsg->free = htons(pQueryInfo->type); // todo valid the vgroupId at the client side STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { 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; pRetrieveMsg->header.vgId = htonl(pTableMeta->vgroupInfo.vgId); } pSql->cmd.payloadLen = sizeof(SRetrieveTableMsg); pSql->cmd.msgType = TSDB_MSG_TYPE_FETCH; pRetrieveMsg->header.contLen = htonl(sizeof(SRetrieveTableMsg)); return TSDB_CODE_SUCCESS; } int tscBuildSubmitMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0); STableMeta* pTableMeta = tscGetMetaInfo(pQueryInfo, 0)->pTableMeta; char* pMsg = pSql->cmd.payload; // NOTE: shell message size should not include SMsgDesc int32_t size = pSql->cmd.payloadLen - sizeof(SMsgDesc); int32_t vgId = pTableMeta->vgroupInfo.vgId; SMsgDesc* pMsgDesc = (SMsgDesc*) pMsg; pMsgDesc->numOfVnodes = htonl(1); // always one vnode pMsg += sizeof(SMsgDesc); SSubmitMsg *pShellMsg = (SSubmitMsg *)pMsg; pShellMsg->header.vgId = htonl(vgId); pShellMsg->header.contLen = htonl(size); // the length not includes the size of SMsgDesc pShellMsg->length = pShellMsg->header.contLen; pShellMsg->numOfBlocks = htonl(pSql->cmd.numOfTablesInSubmit); // number of tables to be inserted // pSql->cmd.payloadLen is set during copying data into payload pSql->cmd.msgType = TSDB_MSG_TYPE_SUBMIT; tscDumpEpSetFromVgroupInfo(&pTableMeta->corVgroupInfo, &pSql->epSet); tscDebug("%p build submit msg, vgId:%d numOfTables:%d numberOfEP:%d", pSql, vgId, pSql->cmd.numOfTablesInSubmit, pSql->epSet.numOfEps); return TSDB_CODE_SUCCESS; } /* * for table query, simply return the size <= 1k */ static int32_t tscEstimateQueryMsgSize(SSqlCmd *pCmd, int32_t clauseIndex) { const static int32_t MIN_QUERY_MSG_PKT_SIZE = TSDB_MAX_BYTES_PER_ROW * 5; SQueryInfo * pQueryInfo = tscGetQueryInfoDetail(pCmd, clauseIndex); int32_t srcColListSize = (int32_t)(taosArrayGetSize(pQueryInfo->colList) * sizeof(SColumnInfo)); size_t numOfExprs = tscSqlExprNumOfExprs(pQueryInfo); int32_t exprSize = (int32_t)(sizeof(SSqlFuncMsg) * numOfExprs); return MIN_QUERY_MSG_PKT_SIZE + minMsgSize() + sizeof(SQueryTableMsg) + srcColListSize + exprSize + 4096; } static char *doSerializeTableInfo(SQueryTableMsg* pQueryMsg, SSqlObj *pSql, char *pMsg) { STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, pSql->cmd.clauseIndex, 0); TSKEY dfltKey = htobe64(pQueryMsg->window.skey); STableMeta * pTableMeta = pTableMetaInfo->pTableMeta; if (UTIL_TABLE_IS_NORMAL_TABLE(pTableMetaInfo) || pTableMetaInfo->pVgroupTables == NULL) { SCMVgroupInfo* pVgroupInfo = NULL; if (UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { int32_t index = pTableMetaInfo->vgroupIndex; 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); } else { pVgroupInfo = &pTableMeta->vgroupInfo; } assert(pVgroupInfo != NULL); tscSetDnodeEpSet(pSql, pVgroupInfo); pQueryMsg->head.vgId = htonl(pVgroupInfo->vgId); STableIdInfo *pTableIdInfo = (STableIdInfo *)pMsg; pTableIdInfo->tid = htonl(pTableMeta->id.tid); pTableIdInfo->uid = htobe64(pTableMeta->id.uid); pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pTableMeta->id.uid, dfltKey)); pQueryMsg->numOfTables = htonl(1); // set the number of tables pMsg += sizeof(STableIdInfo); } else { // it is a subquery of the super table query, this EP info is acquired from vgroupInfo int32_t index = pTableMetaInfo->vgroupIndex; int32_t numOfVgroups = (int32_t)taosArrayGetSize(pTableMetaInfo->pVgroupTables); assert(index >= 0 && index < numOfVgroups); tscDebug("%p query on stable, vgIndex:%d, numOfVgroups:%d", pSql, index, numOfVgroups); SVgroupTableInfo* pTableIdList = taosArrayGet(pTableMetaInfo->pVgroupTables, index); // set the vgroup info tscSetDnodeEpSet(pSql, &pTableIdList->vgInfo); pQueryMsg->head.vgId = htonl(pTableIdList->vgInfo.vgId); int32_t numOfTables = (int32_t)taosArrayGetSize(pTableIdList->itemList); pQueryMsg->numOfTables = htonl(numOfTables); // set the number of tables // serialize each table id info for(int32_t i = 0; i < numOfTables; ++i) { STableIdInfo* pItem = taosArrayGet(pTableIdList->itemList, i); STableIdInfo *pTableIdInfo = (STableIdInfo *)pMsg; pTableIdInfo->tid = htonl(pItem->tid); pTableIdInfo->uid = htobe64(pItem->uid); pTableIdInfo->key = htobe64(tscGetSubscriptionProgress(pSql->pSubscription, pItem->uid, dfltKey)); pMsg += sizeof(STableIdInfo); } } tscDebug("%p vgId:%d, query on table:%s, tid:%d, uid:%" PRIu64, pSql, htonl(pQueryMsg->head.vgId), pTableMetaInfo->name, pTableMeta->id.tid, pTableMeta->id.uid); return pMsg; } int tscBuildQueryMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; int32_t size = tscEstimateQueryMsgSize(pCmd, pCmd->clauseIndex); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, size)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_INVALID_SQL; // todo add test for this } SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); STableMeta * pTableMeta = pTableMetaInfo->pTableMeta; size_t numOfSrcCols = taosArrayGetSize(pQueryInfo->colList); if (numOfSrcCols <= 0 && !tscQueryTags(pQueryInfo)) { tscError("%p illegal value of numOfCols in query msg: %"PRIu64", table cols:%d", pSql, numOfSrcCols, tscGetNumOfColumns(pTableMeta)); return TSDB_CODE_TSC_INVALID_SQL; } if (pQueryInfo->intervalTime < 0) { tscError("%p illegal value of aggregation time interval in query msg: %ld", pSql, pQueryInfo->intervalTime); return TSDB_CODE_TSC_INVALID_SQL; } if (pQueryInfo->groupbyExpr.numOfGroupCols < 0) { tscError("%p illegal value of numOfGroupCols in query msg: %d", pSql, pQueryInfo->groupbyExpr.numOfGroupCols); return TSDB_CODE_TSC_INVALID_SQL; } SQueryTableMsg *pQueryMsg = (SQueryTableMsg *)pCmd->payload; int32_t numOfTags = (int32_t)taosArrayGetSize(pTableMetaInfo->tagColList); if (pQueryInfo->order.order == TSDB_ORDER_ASC) { pQueryMsg->window.skey = htobe64(pQueryInfo->window.skey); pQueryMsg->window.ekey = htobe64(pQueryInfo->window.ekey); } else { pQueryMsg->window.skey = htobe64(pQueryInfo->window.ekey); pQueryMsg->window.ekey = htobe64(pQueryInfo->window.skey); } pQueryMsg->order = htons(pQueryInfo->order.order); pQueryMsg->orderColId = htons(pQueryInfo->order.orderColId); pQueryMsg->fillType = htons(pQueryInfo->fillType); pQueryMsg->limit = htobe64(pQueryInfo->limit.limit); pQueryMsg->offset = htobe64(pQueryInfo->limit.offset); pQueryMsg->numOfCols = htons((int16_t)taosArrayGetSize(pQueryInfo->colList)); pQueryMsg->intervalTime = htobe64(pQueryInfo->intervalTime); pQueryMsg->slidingTime = htobe64(pQueryInfo->slidingTime); pQueryMsg->intervalTimeUnit = pQueryInfo->intervalTimeUnit; pQueryMsg->slidingTimeUnit = pQueryInfo->slidingTimeUnit; pQueryMsg->numOfGroupCols = htons(pQueryInfo->groupbyExpr.numOfGroupCols); pQueryMsg->numOfTags = htonl(numOfTags); pQueryMsg->tagNameRelType = htons(pQueryInfo->tagCond.relType); pQueryMsg->queryType = htonl(pQueryInfo->type); size_t numOfOutput = tscSqlExprNumOfExprs(pQueryInfo); pQueryMsg->numOfOutput = htons((int16_t)numOfOutput); // set column list ids size_t numOfCols = taosArrayGetSize(pQueryInfo->colList); char *pMsg = (char *)(pQueryMsg->colList) + numOfCols * sizeof(SColumnInfo); SSchema *pSchema = tscGetTableSchema(pTableMeta); for (int32_t i = 0; i < numOfCols; ++i) { SColumn *pCol = taosArrayGetP(pQueryInfo->colList, i); SSchema *pColSchema = &pSchema[pCol->colIndex.columnIndex]; if (pCol->colIndex.columnIndex >= tscGetNumOfColumns(pTableMeta) || pColSchema->type < TSDB_DATA_TYPE_BOOL || pColSchema->type > TSDB_DATA_TYPE_NCHAR) { tscError("%p tid:%d uid:%" PRIu64" id:%s, column index out of range, numOfColumns:%d, index:%d, column name:%s", pSql, pTableMeta->id.tid, pTableMeta->id.uid, pTableMetaInfo->name, tscGetNumOfColumns(pTableMeta), pCol->colIndex.columnIndex, pColSchema->name); return TSDB_CODE_TSC_INVALID_SQL; } pQueryMsg->colList[i].colId = htons(pColSchema->colId); pQueryMsg->colList[i].bytes = htons(pColSchema->bytes); pQueryMsg->colList[i].type = htons(pColSchema->type); pQueryMsg->colList[i].numOfFilters = htons(pCol->numOfFilters); // append the filter information after the basic column information for (int32_t f = 0; f < pCol->numOfFilters; ++f) { SColumnFilterInfo *pColFilter = &pCol->filterInfo[f]; SColumnFilterInfo *pFilterMsg = (SColumnFilterInfo *)pMsg; pFilterMsg->filterstr = htons(pColFilter->filterstr); pMsg += sizeof(SColumnFilterInfo); if (pColFilter->filterstr) { pFilterMsg->len = htobe64(pColFilter->len); memcpy(pMsg, (void *)pColFilter->pz, (size_t)(pColFilter->len + 1)); pMsg += (pColFilter->len + 1); // append the additional filter binary info } else { pFilterMsg->lowerBndi = htobe64(pColFilter->lowerBndi); pFilterMsg->upperBndi = htobe64(pColFilter->upperBndi); } pFilterMsg->lowerRelOptr = htons(pColFilter->lowerRelOptr); pFilterMsg->upperRelOptr = htons(pColFilter->upperRelOptr); if (pColFilter->lowerRelOptr == TSDB_RELATION_INVALID && pColFilter->upperRelOptr == TSDB_RELATION_INVALID) { tscError("invalid filter info"); return TSDB_CODE_TSC_INVALID_SQL; } } } SSqlFuncMsg *pSqlFuncExpr = (SSqlFuncMsg *)pMsg; for (int32_t i = 0; i < tscSqlExprNumOfExprs(pQueryInfo); ++i) { SSqlExpr *pExpr = tscSqlExprGet(pQueryInfo, i); if (!tscValidateColumnId(pTableMetaInfo, pExpr->colInfo.colId, pExpr->numOfParams)) { /* column id is not valid according to the cached table meta, the table meta is expired */ tscError("%p table schema is not matched with parsed sql", pSql); return TSDB_CODE_TSC_INVALID_SQL; } pSqlFuncExpr->colInfo.colId = htons(pExpr->colInfo.colId); pSqlFuncExpr->colInfo.colIndex = htons(pExpr->colInfo.colIndex); pSqlFuncExpr->colInfo.flag = htons(pExpr->colInfo.flag); pSqlFuncExpr->functionId = htons(pExpr->functionId); pSqlFuncExpr->numOfParams = htons(pExpr->numOfParams); pMsg += sizeof(SSqlFuncMsg); for (int32_t j = 0; j < pExpr->numOfParams; ++j) { // todo add log pSqlFuncExpr->arg[j].argType = htons((uint16_t)pExpr->param[j].nType); pSqlFuncExpr->arg[j].argBytes = htons(pExpr->param[j].nLen); if (pExpr->param[j].nType == TSDB_DATA_TYPE_BINARY) { memcpy(pMsg, pExpr->param[j].pz, pExpr->param[j].nLen); pMsg += pExpr->param[j].nLen; } else { pSqlFuncExpr->arg[j].argValue.i64 = htobe64(pExpr->param[j].i64Key); } } pSqlFuncExpr = (SSqlFuncMsg *)pMsg; } // serialize the table info (sid, uid, tags) pMsg = doSerializeTableInfo(pQueryMsg, pSql, pMsg); SSqlGroupbyExpr *pGroupbyExpr = &pQueryInfo->groupbyExpr; if (pGroupbyExpr->numOfGroupCols > 0) { pQueryMsg->orderByIdx = htons(pGroupbyExpr->orderIndex); pQueryMsg->orderType = htons(pGroupbyExpr->orderType); for (int32_t j = 0; j < pGroupbyExpr->numOfGroupCols; ++j) { SColIndex* pCol = taosArrayGet(pGroupbyExpr->columnInfo, j); *((int16_t *)pMsg) = pCol->colId; pMsg += sizeof(pCol->colId); *((int16_t *)pMsg) += pCol->colIndex; pMsg += sizeof(pCol->colIndex); *((int16_t *)pMsg) += pCol->flag; pMsg += sizeof(pCol->flag); memcpy(pMsg, pCol->name, tListLen(pCol->name)); pMsg += tListLen(pCol->name); } } if (pQueryInfo->fillType != TSDB_FILL_NONE) { for (int32_t i = 0; i < pQueryInfo->fieldsInfo.numOfOutput; ++i) { *((int64_t *)pMsg) = htobe64(pQueryInfo->fillVal[i]); pMsg += sizeof(pQueryInfo->fillVal[0]); } } if (numOfTags != 0) { int32_t numOfColumns = tscGetNumOfColumns(pTableMeta); int32_t numOfTagColumns = tscGetNumOfTags(pTableMeta); int32_t total = numOfTagColumns + numOfColumns; pSchema = tscGetTableTagSchema(pTableMeta); for (int32_t i = 0; i < numOfTags; ++i) { SColumn *pCol = taosArrayGetP(pTableMetaInfo->tagColList, i); SSchema *pColSchema = &pSchema[pCol->colIndex.columnIndex]; if ((pCol->colIndex.columnIndex >= numOfTagColumns || pCol->colIndex.columnIndex < -1) || (pColSchema->type < TSDB_DATA_TYPE_BOOL || pColSchema->type > TSDB_DATA_TYPE_NCHAR)) { tscError("%p tid:%d uid:%" PRIu64 " id:%s, tag index out of range, totalCols:%d, numOfTags:%d, index:%d, column name:%s", pSql, pTableMeta->id.tid, pTableMeta->id.uid, pTableMetaInfo->name, total, numOfTagColumns, pCol->colIndex.columnIndex, pColSchema->name); return TSDB_CODE_TSC_INVALID_SQL; } SColumnInfo* pTagCol = (SColumnInfo*) pMsg; pTagCol->colId = htons(pColSchema->colId); pTagCol->bytes = htons(pColSchema->bytes); pTagCol->type = htons(pColSchema->type); pTagCol->numOfFilters = 0; pMsg += sizeof(SColumnInfo); } } // serialize tag column query condition if (pQueryInfo->tagCond.pCond != NULL && taosArrayGetSize(pQueryInfo->tagCond.pCond) > 0) { STagCond* pTagCond = &pQueryInfo->tagCond; SCond *pCond = tsGetSTableQueryCond(pTagCond, pTableMeta->id.uid); if (pCond != NULL && pCond->cond != NULL) { pQueryMsg->tagCondLen = htons(pCond->len); memcpy(pMsg, pCond->cond, pCond->len); pMsg += pCond->len; } } if (pQueryInfo->tagCond.tbnameCond.cond == NULL) { *pMsg = 0; pMsg++; } else { strcpy(pMsg, pQueryInfo->tagCond.tbnameCond.cond); pMsg += strlen(pQueryInfo->tagCond.tbnameCond.cond) + 1; } // compressed ts block pQueryMsg->tsOffset = htonl((int32_t)(pMsg - pCmd->payload)); int32_t tsLen = 0; int32_t numOfBlocks = 0; if (pQueryInfo->tsBuf != NULL) { STSVnodeBlockInfo *pBlockInfo = tsBufGetVnodeBlockInfo(pQueryInfo->tsBuf, pTableMetaInfo->vgroupIndex); assert(QUERY_IS_JOIN_QUERY(pQueryInfo->type) && pBlockInfo != NULL); // this query should not be sent // todo refactor if (fseek(pQueryInfo->tsBuf->f, pBlockInfo->offset, SEEK_SET) != 0) { int code = TAOS_SYSTEM_ERROR(ferror(pQueryInfo->tsBuf->f)); tscError("%p: fseek failed: %s", pSql, tstrerror(code)); return code; } size_t s = fread(pMsg, 1, pBlockInfo->compLen, pQueryInfo->tsBuf->f); if (s != pBlockInfo->compLen) { int code = TAOS_SYSTEM_ERROR(ferror(pQueryInfo->tsBuf->f)); tscError("%p: fread didn't return expected data: %s", pSql, tstrerror(code)); return code; } pMsg += pBlockInfo->compLen; tsLen = pBlockInfo->compLen; numOfBlocks = pBlockInfo->numOfBlocks; } pQueryMsg->tsLen = htonl(tsLen); pQueryMsg->tsNumOfBlocks = htonl(numOfBlocks); if (pQueryInfo->tsBuf != NULL) { pQueryMsg->tsOrder = htonl(pQueryInfo->tsBuf->tsOrder); } int32_t msgLen = (int32_t)(pMsg - pCmd->payload); tscDebug("%p msg built success,len:%d bytes", pSql, msgLen); pCmd->payloadLen = msgLen; pSql->cmd.msgType = TSDB_MSG_TYPE_QUERY; pQueryMsg->head.contLen = htonl(msgLen); assert(msgLen + minMsgSize() <= (int32_t)pCmd->allocSize); return TSDB_CODE_SUCCESS; } int32_t tscBuildCreateDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMCreateDbMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_DB; SCMCreateDbMsg *pCreateDbMsg = (SCMCreateDbMsg*)pCmd->payload; assert(pCmd->numOfClause == 1); STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pCreateDbMsg->db, pTableMetaInfo->name, sizeof(pCreateDbMsg->db)); return TSDB_CODE_SUCCESS; } int32_t tscBuildCreateDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMCreateDnodeMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMCreateDnodeMsg *pCreate = (SCMCreateDnodeMsg *)pCmd->payload; strncpy(pCreate->ep, pInfo->pDCLInfo->a[0].z, pInfo->pDCLInfo->a[0].n); pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_DNODE; return TSDB_CODE_SUCCESS; } int32_t tscBuildAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMCreateAcctMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMCreateAcctMsg *pAlterMsg = (SCMCreateAcctMsg *)pCmd->payload; SStrToken *pName = &pInfo->pDCLInfo->user.user; SStrToken *pPwd = &pInfo->pDCLInfo->user.passwd; strncpy(pAlterMsg->user, pName->z, pName->n); strncpy(pAlterMsg->pass, pPwd->z, pPwd->n); SCreateAcctSQL *pAcctOpt = &pInfo->pDCLInfo->acctOpt; pAlterMsg->cfg.maxUsers = htonl(pAcctOpt->maxUsers); pAlterMsg->cfg.maxDbs = htonl(pAcctOpt->maxDbs); pAlterMsg->cfg.maxTimeSeries = htonl(pAcctOpt->maxTimeSeries); pAlterMsg->cfg.maxStreams = htonl(pAcctOpt->maxStreams); pAlterMsg->cfg.maxPointsPerSecond = htonl(pAcctOpt->maxPointsPerSecond); pAlterMsg->cfg.maxStorage = htobe64(pAcctOpt->maxStorage); pAlterMsg->cfg.maxQueryTime = htobe64(pAcctOpt->maxQueryTime); pAlterMsg->cfg.maxConnections = htonl(pAcctOpt->maxConnections); if (pAcctOpt->stat.n == 0) { pAlterMsg->cfg.accessState = -1; } else { if (pAcctOpt->stat.z[0] == 'r' && pAcctOpt->stat.n == 1) { pAlterMsg->cfg.accessState = TSDB_VN_READ_ACCCESS; } else if (pAcctOpt->stat.z[0] == 'w' && pAcctOpt->stat.n == 1) { pAlterMsg->cfg.accessState = TSDB_VN_WRITE_ACCCESS; } else if (strncmp(pAcctOpt->stat.z, "all", 3) == 0 && pAcctOpt->stat.n == 3) { pAlterMsg->cfg.accessState = TSDB_VN_ALL_ACCCESS; } else if (strncmp(pAcctOpt->stat.z, "no", 2) == 0 && pAcctOpt->stat.n == 2) { pAlterMsg->cfg.accessState = 0; } } pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_ACCT; return TSDB_CODE_SUCCESS; } int32_t tscBuildUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMCreateUserMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMCreateUserMsg *pAlterMsg = (SCMCreateUserMsg*)pCmd->payload; SUserInfo *pUser = &pInfo->pDCLInfo->user; strncpy(pAlterMsg->user, pUser->user.z, pUser->user.n); pAlterMsg->flag = (int8_t)pUser->type; if (pUser->type == TSDB_ALTER_USER_PRIVILEGES) { pAlterMsg->privilege = (char)pCmd->count; } else if (pUser->type == TSDB_ALTER_USER_PASSWD) { strncpy(pAlterMsg->pass, pUser->passwd.z, pUser->passwd.n); } else { // create user password info strncpy(pAlterMsg->pass, pUser->passwd.z, pUser->passwd.n); } if (pUser->type == TSDB_ALTER_USER_PASSWD || pUser->type == TSDB_ALTER_USER_PRIVILEGES) { pCmd->msgType = TSDB_MSG_TYPE_CM_ALTER_USER; } else { pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_USER; } return TSDB_CODE_SUCCESS; } int32_t tscBuildCfgDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMCfgDnodeMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_CONFIG_DNODE; return TSDB_CODE_SUCCESS; } int32_t tscBuildDropDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMDropDbMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMDropDbMsg *pDropDbMsg = (SCMDropDbMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropDbMsg->db, pTableMetaInfo->name, sizeof(pDropDbMsg->db)); pDropDbMsg->ignoreNotExists = pInfo->pDCLInfo->existsCheck ? 1 : 0; pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_DB; return TSDB_CODE_SUCCESS; } int32_t tscBuildDropTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMDropTableMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMDropTableMsg *pDropTableMsg = (SCMDropTableMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); strcpy(pDropTableMsg->tableId, pTableMetaInfo->name); pDropTableMsg->igNotExists = pInfo->pDCLInfo->existsCheck ? 1 : 0; pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_TABLE; return TSDB_CODE_SUCCESS; } int32_t tscBuildDropDnodeMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMDropDnodeMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMDropDnodeMsg *pDrop = (SCMDropDnodeMsg *)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDrop->ep, pTableMetaInfo->name, sizeof(pDrop->ep)); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_DNODE; return TSDB_CODE_SUCCESS; } int32_t tscBuildDropUserMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMDropUserMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_USER; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMDropUserMsg *pDropMsg = (SCMDropUserMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropMsg->user, pTableMetaInfo->name, sizeof(pDropMsg->user)); return TSDB_CODE_SUCCESS; } int32_t tscBuildDropAcctMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMDropUserMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_DROP_ACCT; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMDropUserMsg *pDropMsg = (SCMDropUserMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pDropMsg->user, pTableMetaInfo->name, sizeof(pDropMsg->user)); return TSDB_CODE_SUCCESS; } int32_t tscBuildUseDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMUseDbMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMUseDbMsg *pUseDbMsg = (SCMUseDbMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); strcpy(pUseDbMsg->db, pTableMetaInfo->name); pCmd->msgType = TSDB_MSG_TYPE_CM_USE_DB; return TSDB_CODE_SUCCESS; } int32_t tscBuildShowMsg(SSqlObj *pSql, SSqlInfo *pInfo) { STscObj *pObj = pSql->pTscObj; SSqlCmd *pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_CM_SHOW; pCmd->payloadLen = sizeof(SCMShowMsg) + 100; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMShowMsg *pShowMsg = (SCMShowMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); size_t nameLen = strlen(pTableMetaInfo->name); if (nameLen > 0) { tstrncpy(pShowMsg->db, pTableMetaInfo->name, sizeof(pShowMsg->db)); // prefix is set here } else { tstrncpy(pShowMsg->db, pObj->db, sizeof(pShowMsg->db)); } SShowInfo *pShowInfo = &pInfo->pDCLInfo->showOpt; pShowMsg->type = pShowInfo->showType; if (pShowInfo->showType != TSDB_MGMT_TABLE_VNODES) { SStrToken *pPattern = &pShowInfo->pattern; if (pPattern->type > 0) { // only show tables support wildcard query strncpy(pShowMsg->payload, pPattern->z, pPattern->n); pShowMsg->payloadLen = htons(pPattern->n); } } else { SStrToken *pEpAddr = &pShowInfo->prefix; assert(pEpAddr->n > 0 && pEpAddr->type > 0); strncpy(pShowMsg->payload, pEpAddr->z, pEpAddr->n); pShowMsg->payloadLen = htons(pEpAddr->n); } pCmd->payloadLen = sizeof(SCMShowMsg) + pShowMsg->payloadLen; return TSDB_CODE_SUCCESS; } int32_t tscBuildKillMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMKillQueryMsg); switch (pCmd->command) { case TSDB_SQL_KILL_QUERY: pCmd->msgType = TSDB_MSG_TYPE_CM_KILL_QUERY; break; case TSDB_SQL_KILL_CONNECTION: pCmd->msgType = TSDB_MSG_TYPE_CM_KILL_CONN; break; case TSDB_SQL_KILL_STREAM: pCmd->msgType = TSDB_MSG_TYPE_CM_KILL_STREAM; break; } return TSDB_CODE_SUCCESS; } int tscEstimateCreateTableMsgLength(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &(pSql->cmd); int32_t size = minMsgSize() + sizeof(SCMCreateTableMsg); SCreateTableSQL *pCreateTableInfo = pInfo->pCreateTableInfo; if (pCreateTableInfo->type == TSQL_CREATE_TABLE_FROM_STABLE) { size += sizeof(STagData); } else { size += sizeof(SSchema) * (pCmd->numOfCols + pCmd->count); } if (pCreateTableInfo->pSelect != NULL) { size += (pCreateTableInfo->pSelect->selectToken.n + 1); } return size + TSDB_EXTRA_PAYLOAD_SIZE; } int tscBuildCreateTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { int msgLen = 0; SSchema * pSchema; int size = 0; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo * pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); // Reallocate the payload size size = tscEstimateCreateTableMsgLength(pSql, pInfo); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, size)) { tscError("%p failed to malloc for create table msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMCreateTableMsg *pCreateTableMsg = (SCMCreateTableMsg *)pCmd->payload; strcpy(pCreateTableMsg->tableId, pTableMetaInfo->name); // use dbinfo from table id without modifying current db info tscGetDBInfoFromTableFullName(pTableMetaInfo->name, pCreateTableMsg->db); SCreateTableSQL *pCreateTable = pInfo->pCreateTableInfo; pCreateTableMsg->igExists = pCreateTable->existCheck ? 1 : 0; pCreateTableMsg->numOfColumns = htons(pCmd->numOfCols); pCreateTableMsg->numOfTags = htons(pCmd->count); pCreateTableMsg->sqlLen = 0; char *pMsg = (char *)pCreateTableMsg->schema; int8_t type = pInfo->pCreateTableInfo->type; if (type == TSQL_CREATE_TABLE_FROM_STABLE) { // create by using super table, tags value STagData* pTag = &pInfo->pCreateTableInfo->usingInfo.tagdata; *(int32_t*)pMsg = htonl(pTag->dataLen); pMsg += sizeof(int32_t); memcpy(pMsg, pTag->name, sizeof(pTag->name)); pMsg += sizeof(pTag->name); memcpy(pMsg, pTag->data, pTag->dataLen); pMsg += pTag->dataLen; } else { // create (super) table pSchema = (SSchema *)pCreateTableMsg->schema; for (int i = 0; i < pCmd->numOfCols + pCmd->count; ++i) { TAOS_FIELD *pField = tscFieldInfoGetField(&pQueryInfo->fieldsInfo, i); pSchema->type = pField->type; strcpy(pSchema->name, pField->name); pSchema->bytes = htons(pField->bytes); pSchema++; } pMsg = (char *)pSchema; if (type == TSQL_CREATE_STREAM) { // check if it is a stream sql SQuerySQL *pQuerySql = pInfo->pCreateTableInfo->pSelect; strncpy(pMsg, pQuerySql->selectToken.z, pQuerySql->selectToken.n + 1); pCreateTableMsg->sqlLen = htons(pQuerySql->selectToken.n + 1); pMsg += pQuerySql->selectToken.n + 1; } } tscFieldInfoClear(&pQueryInfo->fieldsInfo); msgLen = (int32_t)(pMsg - (char*)pCreateTableMsg); pCreateTableMsg->contLen = htonl(msgLen); pCmd->payloadLen = msgLen; pCmd->msgType = TSDB_MSG_TYPE_CM_CREATE_TABLE; assert(msgLen + minMsgSize() <= size); return TSDB_CODE_SUCCESS; } int tscEstimateAlterTableMsgLength(SSqlCmd *pCmd) { SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); return minMsgSize() + sizeof(SCMAlterTableMsg) + sizeof(SSchema) * tscNumOfFields(pQueryInfo) + TSDB_EXTRA_PAYLOAD_SIZE; } int tscBuildAlterTableMsg(SSqlObj *pSql, SSqlInfo *pInfo) { char *pMsg; int msgLen = 0; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); SAlterTableSQL *pAlterInfo = pInfo->pAlterInfo; int size = tscEstimateAlterTableMsgLength(pCmd); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, size)) { tscError("%p failed to malloc for alter table msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMAlterTableMsg *pAlterTableMsg = (SCMAlterTableMsg *)pCmd->payload; tscGetDBInfoFromTableFullName(pTableMetaInfo->name, pAlterTableMsg->db); strcpy(pAlterTableMsg->tableId, pTableMetaInfo->name); pAlterTableMsg->type = htons(pAlterInfo->type); pAlterTableMsg->numOfCols = htons(tscNumOfFields(pQueryInfo)); SSchema *pSchema = pAlterTableMsg->schema; for (int i = 0; i < tscNumOfFields(pQueryInfo); ++i) { TAOS_FIELD *pField = tscFieldInfoGetField(&pQueryInfo->fieldsInfo, i); pSchema->type = pField->type; strcpy(pSchema->name, pField->name); pSchema->bytes = htons(pField->bytes); pSchema++; } pMsg = (char *)pSchema; pAlterTableMsg->tagValLen = htonl(pAlterInfo->tagData.dataLen); memcpy(pMsg, pAlterInfo->tagData.data, pAlterInfo->tagData.dataLen); pMsg += pAlterInfo->tagData.dataLen; msgLen = (int32_t)(pMsg - (char*)pAlterTableMsg); pCmd->payloadLen = msgLen; pCmd->msgType = TSDB_MSG_TYPE_CM_ALTER_TABLE; assert(msgLen + minMsgSize() <= size); return TSDB_CODE_SUCCESS; } int tscBuildUpdateTagMsg(SSqlObj* pSql, SSqlInfo *pInfo) { SSqlCmd* pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_UPDATE_TAG_VAL; SUpdateTableTagValMsg* pUpdateMsg = (SUpdateTableTagValMsg*) pCmd->payload; pCmd->payloadLen = htonl(pUpdateMsg->head.contLen); SQueryInfo * pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); tscDumpEpSetFromVgroupInfo(&pTableMetaInfo->pTableMeta->corVgroupInfo, &pSql->epSet); return TSDB_CODE_SUCCESS; } int tscAlterDbMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->payloadLen = sizeof(SCMAlterDbMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_ALTER_DB; SCMAlterDbMsg *pAlterDbMsg = (SCMAlterDbMsg*)pCmd->payload; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); tstrncpy(pAlterDbMsg->db, pTableMetaInfo->name, sizeof(pAlterDbMsg->db)); return TSDB_CODE_SUCCESS; } int tscBuildRetrieveFromMgmtMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_CM_RETRIEVE; pCmd->payloadLen = sizeof(SRetrieveTableMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); SRetrieveTableMsg *pRetrieveMsg = (SRetrieveTableMsg*)pCmd->payload; pRetrieveMsg->qhandle = htobe64(pSql->res.qhandle); pRetrieveMsg->free = htons(pQueryInfo->type); return TSDB_CODE_SUCCESS; } static int tscSetResultPointer(SQueryInfo *pQueryInfo, SSqlRes *pRes) { if (tscCreateResPointerInfo(pRes, pQueryInfo) != TSDB_CODE_SUCCESS) { return pRes->code; } for (int i = 0; i < pQueryInfo->fieldsInfo.numOfOutput; ++i) { int16_t offset = tscFieldInfoGetOffset(pQueryInfo, i); pRes->tsrow[i] = ((char*) pRes->data + offset * pRes->numOfRows); } return 0; } /* * this function can only be called once. * by using pRes->rspType to denote its status * * if pRes->rspType is 1, no more result */ static int tscLocalResultCommonBuilder(SSqlObj *pSql, int32_t numOfRes) { SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); pRes->code = TSDB_CODE_SUCCESS; if (pRes->rspType == 0) { pRes->numOfRows = numOfRes; pRes->row = 0; pRes->rspType = 1; tscSetResultPointer(pQueryInfo, pRes); } else { tscResetForNextRetrieve(pRes); } uint8_t code = pSql->res.code; if (pSql->fp) { if (code == TSDB_CODE_SUCCESS) { (*pSql->fp)(pSql->param, pSql, pSql->res.numOfRows); } else { tscQueueAsyncRes(pSql); } } return code; } int tscProcessDescribeTableRsp(SSqlObj *pSql) { SSqlCmd * pCmd = &pSql->cmd; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); STableComInfo tinfo = tscGetTableInfo(pTableMetaInfo->pTableMeta); int32_t numOfRes = tinfo.numOfColumns + tinfo.numOfTags; return tscLocalResultCommonBuilder(pSql, numOfRes); } int tscProcessLocalRetrieveRsp(SSqlObj *pSql) { int32_t numOfRes = 1; pSql->res.completed = true; return tscLocalResultCommonBuilder(pSql, numOfRes); } int tscProcessRetrieveLocalMergeRsp(SSqlObj *pSql) { SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; int32_t code = pRes->code; if (pRes->code != TSDB_CODE_SUCCESS) { tscQueueAsyncRes(pSql); return code; } pRes->code = tscDoLocalMerge(pSql); SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); if (pRes->code == TSDB_CODE_SUCCESS && pRes->numOfRows > 0) { tscCreateResPointerInfo(pRes, pQueryInfo); } pRes->row = 0; pRes->completed = (pRes->numOfRows == 0); code = pRes->code; if (pRes->code == TSDB_CODE_SUCCESS) { (*pSql->fp)(pSql->param, pSql, pRes->numOfRows); } else { tscQueueAsyncRes(pSql); } return code; } int tscProcessEmptyResultRsp(SSqlObj *pSql) { return tscLocalResultCommonBuilder(pSql, 0); } int tscBuildConnectMsg(SSqlObj *pSql, SSqlInfo *pInfo) { STscObj *pObj = pSql->pTscObj; SSqlCmd *pCmd = &pSql->cmd; pCmd->msgType = TSDB_MSG_TYPE_CM_CONNECT; pCmd->payloadLen = sizeof(SCMConnectMsg); if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, pCmd->payloadLen)) { tscError("%p failed to malloc for query msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMConnectMsg *pConnect = (SCMConnectMsg*)pCmd->payload; char *db; // ugly code to move the space db = strstr(pObj->db, TS_PATH_DELIMITER); db = (db == NULL) ? pObj->db : db + 1; tstrncpy(pConnect->db, db, sizeof(pConnect->db)); tstrncpy(pConnect->clientVersion, version, sizeof(pConnect->clientVersion)); tstrncpy(pConnect->msgVersion, "", sizeof(pConnect->msgVersion)); return TSDB_CODE_SUCCESS; } int tscBuildTableMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SCMTableInfoMsg *pInfoMsg; char * pMsg; int msgLen = 0; char *tmpData = NULL; uint32_t len = pSql->cmd.payloadLen; if (len > 0) { if ((tmpData = calloc(1, len)) == NULL) { return TSDB_CODE_TSC_OUT_OF_MEMORY; } // STagData is in binary format, strncpy is not available memcpy(tmpData, pSql->cmd.payload, len); } SSqlCmd * pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(&pSql->cmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); pInfoMsg = (SCMTableInfoMsg *)pCmd->payload; strcpy(pInfoMsg->tableId, pTableMetaInfo->name); pInfoMsg->createFlag = htons(pSql->cmd.autoCreated ? 1 : 0); pMsg = (char*)pInfoMsg + sizeof(SCMTableInfoMsg); if (pSql->cmd.autoCreated && len > 0) { memcpy(pInfoMsg->tags, tmpData, len); pMsg += len; } pCmd->payloadLen = (int32_t)(pMsg - (char*)pInfoMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_TABLE_META; taosTFree(tmpData); assert(msgLen + minMsgSize() <= (int32_t)pCmd->allocSize); return TSDB_CODE_SUCCESS; } /** * multi table meta req pkg format: * | SMgmtHead | SCMMultiTableInfoMsg | tableId0 | tableId1 | tableId2 | ...... * no used 4B **/ int tscBuildMultiMeterMetaMsg(SSqlObj *pSql, SSqlInfo *pInfo) { #if 0 SSqlCmd *pCmd = &pSql->cmd; // copy payload content to temp buff char *tmpData = 0; if (pCmd->payloadLen > 0) { if ((tmpData = calloc(1, pCmd->payloadLen + 1)) == NULL) return -1; memcpy(tmpData, pCmd->payload, pCmd->payloadLen); } // fill head info SMgmtHead *pMgmt = (SMgmtHead *)(pCmd->payload + tsRpcHeadSize); memset(pMgmt->db, 0, TSDB_TABLE_FNAME_LEN); // server don't need the db SCMMultiTableInfoMsg *pInfoMsg = (SCMMultiTableInfoMsg *)(pCmd->payload + tsRpcHeadSize + sizeof(SMgmtHead)); pInfoMsg->numOfTables = htonl((int32_t)pCmd->count); if (pCmd->payloadLen > 0) { memcpy(pInfoMsg->tableIds, tmpData, pCmd->payloadLen); } taosTFree(tmpData); pCmd->payloadLen += sizeof(SMgmtHead) + sizeof(SCMMultiTableInfoMsg); pCmd->msgType = TSDB_MSG_TYPE_CM_TABLES_META; assert(pCmd->payloadLen + minMsgSize() <= pCmd->allocSize); tscDebug("%p build load multi-metermeta msg completed, numOfTables:%d, msg size:%d", pSql, pCmd->count, pCmd->payloadLen); return pCmd->payloadLen; #endif return 0; } //static UNUSED_FUNC int32_t tscEstimateMetricMetaMsgSize(SSqlCmd *pCmd) { //// const int32_t defaultSize = //// minMsgSize() + sizeof(SSuperTableMetaMsg) + sizeof(SMgmtHead) + sizeof(int16_t) * TSDB_MAX_TAGS; //// SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); //// //// int32_t n = 0; //// size_t size = taosArrayGetSize(pQueryInfo->tagCond.pCond); //// for (int32_t i = 0; i < size; ++i) { //// assert(0); ////// n += strlen(pQueryInfo->tagCond.cond[i].cond); //// } //// //// int32_t tagLen = n * TSDB_NCHAR_SIZE; //// if (pQueryInfo->tagCond.tbnameCond.cond != NULL) { //// tagLen += strlen(pQueryInfo->tagCond.tbnameCond.cond) * TSDB_NCHAR_SIZE; //// } //// //// int32_t joinCondLen = (TSDB_TABLE_FNAME_LEN + sizeof(int16_t)) * 2; //// int32_t elemSize = sizeof(SSuperTableMetaElemMsg) * pQueryInfo->numOfTables; //// //// int32_t colSize = pQueryInfo->groupbyExpr.numOfGroupCols*sizeof(SColIndex); //// //// int32_t len = tagLen + joinCondLen + elemSize + colSize + defaultSize; //// //// return MAX(len, TSDB_DEFAULT_PAYLOAD_SIZE); //} int tscBuildSTableVgroupMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; char* pMsg = pCmd->payload; SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); SCMSTableVgroupMsg *pStableVgroupMsg = (SCMSTableVgroupMsg *) pMsg; pStableVgroupMsg->numOfTables = htonl(pQueryInfo->numOfTables); pMsg += sizeof(SCMSTableVgroupMsg); for(int32_t i = 0; i < pQueryInfo->numOfTables; ++i) { STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, i); size_t size = sizeof(pTableMetaInfo->name); tstrncpy(pMsg, pTableMetaInfo->name, size); pMsg += size; } pCmd->msgType = TSDB_MSG_TYPE_CM_STABLE_VGROUP; pCmd->payloadLen = (int32_t)(pMsg - pCmd->payload); return TSDB_CODE_SUCCESS; } int tscBuildHeartBeatMsg(SSqlObj *pSql, SSqlInfo *pInfo) { SSqlCmd *pCmd = &pSql->cmd; STscObj *pObj = pSql->pTscObj; pthread_mutex_lock(&pObj->mutex); int32_t numOfQueries = 2; SSqlObj *tpSql = pObj->sqlList; while (tpSql) { tpSql = tpSql->next; numOfQueries++; } int32_t numOfStreams = 2; SSqlStream *pStream = pObj->streamList; while (pStream) { pStream = pStream->next; numOfStreams++; } int size = numOfQueries * sizeof(SQueryDesc) + numOfStreams * sizeof(SStreamDesc) + sizeof(SCMHeartBeatMsg) + 100; if (TSDB_CODE_SUCCESS != tscAllocPayload(pCmd, size)) { pthread_mutex_unlock(&pObj->mutex); tscError("%p failed to malloc for heartbeat msg", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } SCMHeartBeatMsg *pHeartbeat = (SCMHeartBeatMsg *)pCmd->payload; pHeartbeat->numOfQueries = numOfQueries; pHeartbeat->numOfStreams = numOfStreams; int msgLen = tscBuildQueryStreamDesc(pHeartbeat, pObj); pthread_mutex_unlock(&pObj->mutex); pCmd->payloadLen = msgLen; pCmd->msgType = TSDB_MSG_TYPE_CM_HEARTBEAT; return TSDB_CODE_SUCCESS; } int tscProcessTableMetaRsp(SSqlObj *pSql) { STableMetaMsg *pMetaMsg = (STableMetaMsg *)pSql->res.pRsp; pMetaMsg->sid = htonl(pMetaMsg->sid); pMetaMsg->sversion = htons(pMetaMsg->sversion); pMetaMsg->tversion = htons(pMetaMsg->tversion); pMetaMsg->vgroup.vgId = htonl(pMetaMsg->vgroup.vgId); pMetaMsg->uid = htobe64(pMetaMsg->uid); pMetaMsg->contLen = htons(pMetaMsg->contLen); pMetaMsg->numOfColumns = htons(pMetaMsg->numOfColumns); if (pMetaMsg->sid < 0 || pMetaMsg->vgroup.numOfEps < 0) { tscError("invalid meter vgId:%d, sid%d", pMetaMsg->vgroup.numOfEps, pMetaMsg->sid); return TSDB_CODE_TSC_INVALID_VALUE; } if (pMetaMsg->numOfTags > TSDB_MAX_TAGS) { tscError("invalid numOfTags:%d", pMetaMsg->numOfTags); return TSDB_CODE_TSC_INVALID_VALUE; } if (pMetaMsg->numOfColumns > TSDB_MAX_COLUMNS || pMetaMsg->numOfColumns <= 0) { tscError("invalid numOfColumns:%d", pMetaMsg->numOfColumns); return TSDB_CODE_TSC_INVALID_VALUE; } for (int i = 0; i < pMetaMsg->vgroup.numOfEps; ++i) { pMetaMsg->vgroup.epAddr[i].port = htons(pMetaMsg->vgroup.epAddr[i].port); } SSchema* pSchema = pMetaMsg->schema; int32_t numOfTotalCols = pMetaMsg->numOfColumns + pMetaMsg->numOfTags; for (int i = 0; i < numOfTotalCols; ++i) { pSchema->bytes = htons(pSchema->bytes); pSchema->colId = htons(pSchema->colId); if (pSchema->colId == PRIMARYKEY_TIMESTAMP_COL_INDEX) { assert(i == 0); } assert(pSchema->type >= TSDB_DATA_TYPE_BOOL && pSchema->type <= TSDB_DATA_TYPE_NCHAR); pSchema++; } size_t size = 0; STableMeta* pTableMeta = tscCreateTableMetaFromMsg(pMetaMsg, &size); // todo add one more function: taosAddDataIfNotExists(); STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); assert(pTableMetaInfo->pTableMeta == NULL); pTableMetaInfo->pTableMeta = (STableMeta *) taosCachePut(tscCacheHandle, pTableMetaInfo->name, strlen(pTableMetaInfo->name), pTableMeta, size, tsTableMetaKeepTimer * 1000); // todo handle out of memory case if (pTableMetaInfo->pTableMeta == NULL) { free(pTableMeta); return TSDB_CODE_TSC_OUT_OF_MEMORY; } tscDebug("%p recv table meta, uid:%"PRId64 ", tid:%d, name:%s", pSql, pTableMeta->id.uid, pTableMeta->id.tid, pTableMetaInfo->name); free(pTableMeta); return TSDB_CODE_SUCCESS; } /** * multi table meta rsp pkg format: * | STaosRsp | ieType | SCMMultiTableInfoMsg | SMeterMeta0 | SSchema0 | SMeterMeta1 | SSchema1 | SMeterMeta2 | SSchema2 * |...... 1B 1B 4B **/ int tscProcessMultiMeterMetaRsp(SSqlObj *pSql) { #if 0 char *rsp = pSql->res.pRsp; ieType = *rsp; if (ieType != TSDB_IE_TYPE_META) { tscError("invalid ie type:%d", ieType); pSql->res.code = TSDB_CODE_TSC_INVALID_IE; pSql->res.numOfTotal = 0; return TSDB_CODE_TSC_APP_ERROR; } rsp++; SCMMultiTableInfoMsg *pInfo = (SCMMultiTableInfoMsg *)rsp; totalNum = htonl(pInfo->numOfTables); rsp += sizeof(SCMMultiTableInfoMsg); for (i = 0; i < totalNum; i++) { SMultiTableMeta *pMultiMeta = (SMultiTableMeta *)rsp; STableMeta * pMeta = pMultiMeta->metas; pMeta->sid = htonl(pMeta->sid); pMeta->sversion = htons(pMeta->sversion); pMeta->vgId = htonl(pMeta->vgId); pMeta->uid = htobe64(pMeta->uid); if (pMeta->sid <= 0 || pMeta->vgId < 0) { tscError("invalid meter vgId:%d, sid%d", pMeta->vgId, pMeta->sid); pSql->res.code = TSDB_CODE_TSC_INVALID_VALUE; pSql->res.numOfTotal = i; return TSDB_CODE_TSC_APP_ERROR; } // pMeta->numOfColumns = htons(pMeta->numOfColumns); // // if (pMeta->numOfTags > TSDB_MAX_TAGS || pMeta->numOfTags < 0) { // tscError("invalid tag value count:%d", pMeta->numOfTags); // pSql->res.code = TSDB_CODE_TSC_INVALID_VALUE; // pSql->res.numOfTotal = i; // return TSDB_CODE_TSC_APP_ERROR; // } // // if (pMeta->numOfTags > TSDB_MAX_TAGS || pMeta->numOfTags < 0) { // tscError("invalid numOfTags:%d", pMeta->numOfTags); // pSql->res.code = TSDB_CODE_TSC_INVALID_VALUE; // pSql->res.numOfTotal = i; // return TSDB_CODE_TSC_APP_ERROR; // } // // if (pMeta->numOfColumns > TSDB_MAX_COLUMNS || pMeta->numOfColumns < 0) { // tscError("invalid numOfColumns:%d", pMeta->numOfColumns); // pSql->res.code = TSDB_CODE_TSC_INVALID_VALUE; // pSql->res.numOfTotal = i; // return TSDB_CODE_TSC_APP_ERROR; // } // // for (int j = 0; j < TSDB_REPLICA_MAX_NUM; ++j) { // pMeta->vpeerDesc[j].vnode = htonl(pMeta->vpeerDesc[j].vnode); // } // // pMeta->rowSize = 0; // rsp += sizeof(SMultiTableMeta); // pSchema = (SSchema *)rsp; // // int32_t numOfTotalCols = pMeta->numOfColumns + pMeta->numOfTags; // for (int j = 0; j < numOfTotalCols; ++j) { // pSchema->bytes = htons(pSchema->bytes); // pSchema->colId = htons(pSchema->colId); // // // ignore the tags length // if (j < pMeta->numOfColumns) { // pMeta->rowSize += pSchema->bytes; // } // pSchema++; // } // // rsp += numOfTotalCols * sizeof(SSchema); // // int32_t tagLen = 0; // SSchema *pTagsSchema = tscGetTableTagSchema(pMeta); // // if (pMeta->tableType == TSDB_CHILD_TABLE) { // for (int32_t j = 0; j < pMeta->numOfTags; ++j) { // tagLen += pTagsSchema[j].bytes; // } // } // // rsp += tagLen; // int32_t size = (int32_t)(rsp - ((char *)pMeta)); // Consistent with STableMeta in cache // // pMeta->index = 0; // (void)taosCachePut(tscCacheHandle, pMeta->tableId, (char *)pMeta, size, tsTableMetaKeepTimer); // } } pSql->res.code = TSDB_CODE_SUCCESS; pSql->res.numOfTotal = i; tscDebug("%p load multi-metermeta resp from complete num:%d", pSql, pSql->res.numOfTotal); #endif return TSDB_CODE_SUCCESS; } int tscProcessSTableVgroupRsp(SSqlObj *pSql) { SSqlRes* pRes = &pSql->res; // NOTE: the order of several table must be preserved. SCMSTableVgroupRspMsg *pStableVgroup = (SCMSTableVgroupRspMsg *)pRes->pRsp; pStableVgroup->numOfTables = htonl(pStableVgroup->numOfTables); char* pMsg = pRes->pRsp + sizeof(SCMSTableVgroupRspMsg); // master sqlObj locates in param SSqlObj* parent = pSql->param; assert(parent != NULL); SSqlCmd* pCmd = &parent->cmd; for(int32_t i = 0; i < pStableVgroup->numOfTables; ++i) { STableMetaInfo *pInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, i); SVgroupsInfo * pVgroupInfo = (SVgroupsInfo *)pMsg; pVgroupInfo->numOfVgroups = htonl(pVgroupInfo->numOfVgroups); size_t size = sizeof(SCMVgroupInfo) * pVgroupInfo->numOfVgroups + sizeof(SVgroupsInfo); pInfo->vgroupList = calloc(1, size); assert(pInfo->vgroupList != NULL); memcpy(pInfo->vgroupList, pVgroupInfo, size); for (int32_t j = 0; j < pInfo->vgroupList->numOfVgroups; ++j) { //just init, no need to lock SCMVgroupInfo *pVgroups = &pInfo->vgroupList->vgroups[j]; pVgroups->vgId = htonl(pVgroups->vgId); assert(pVgroups->numOfEps >= 1); for (int32_t k = 0; k < pVgroups->numOfEps; ++k) { pVgroups->epAddr[k].port = htons(pVgroups->epAddr[k].port); } } pMsg += size; } return pSql->res.code; } /* * current process do not use the cache at all */ int tscProcessShowRsp(SSqlObj *pSql) { STableMetaMsg * pMetaMsg; SCMShowRsp *pShow; SSchema * pSchema; char key[20]; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); pShow = (SCMShowRsp *)pRes->pRsp; pShow->qhandle = htobe64(pShow->qhandle); pRes->qhandle = pShow->qhandle; tscResetForNextRetrieve(pRes); pMetaMsg = &(pShow->tableMeta); pMetaMsg->numOfColumns = ntohs(pMetaMsg->numOfColumns); pSchema = pMetaMsg->schema; pMetaMsg->sid = ntohs(pMetaMsg->sid); for (int i = 0; i < pMetaMsg->numOfColumns; ++i) { pSchema->bytes = htons(pSchema->bytes); pSchema++; } key[0] = pCmd->msgType + 'a'; strcpy(key + 1, "showlist"); taosCacheRelease(tscCacheHandle, (void *)&(pTableMetaInfo->pTableMeta), false); size_t size = 0; STableMeta* pTableMeta = tscCreateTableMetaFromMsg(pMetaMsg, &size); pTableMetaInfo->pTableMeta = taosCachePut(tscCacheHandle, key, strlen(key), (char *)pTableMeta, size, tsTableMetaKeepTimer * 1000); SSchema *pTableSchema = tscGetTableSchema(pTableMetaInfo->pTableMeta); if (pQueryInfo->colList == NULL) { pQueryInfo->colList = taosArrayInit(4, POINTER_BYTES); } SFieldInfo* pFieldInfo = &pQueryInfo->fieldsInfo; SColumnIndex index = {0}; pSchema = pMetaMsg->schema; for (int16_t i = 0; i < pMetaMsg->numOfColumns; ++i, ++pSchema) { index.columnIndex = i; tscColumnListInsert(pQueryInfo->colList, &index); TAOS_FIELD f = tscCreateField(pSchema->type, pSchema->name, pSchema->bytes); SFieldSupInfo* pInfo = tscFieldInfoAppend(pFieldInfo, &f); pInfo->pSqlExpr = tscSqlExprAppend(pQueryInfo, TSDB_FUNC_TS_DUMMY, &index, pTableSchema[i].type, pTableSchema[i].bytes, pTableSchema[i].bytes, false); } pCmd->numOfCols = pQueryInfo->fieldsInfo.numOfOutput; tscFieldInfoUpdateOffset(pQueryInfo); taosTFree(pTableMeta); return 0; } static void createHBObj(STscObj* pObj) { if (pObj->pHb != NULL) { return; } SSqlObj *pSql = (SSqlObj *)calloc(1, sizeof(SSqlObj)); if (NULL == pSql) return; pSql->fp = tscProcessHeartBeatRsp; SQueryInfo *pQueryInfo = tscGetQueryInfoDetailSafely(&pSql->cmd, 0); if (pQueryInfo == NULL) { pSql->res.code = terrno; return; } pQueryInfo->command = TSDB_SQL_HB; pSql->cmd.command = pQueryInfo->command; if (TSDB_CODE_SUCCESS != tscAllocPayload(&(pSql->cmd), TSDB_DEFAULT_PAYLOAD_SIZE)) { taosTFree(pSql); return; } pSql->param = pObj; pSql->pTscObj = pObj; pSql->signature = pSql; pObj->pHb = pSql; tscAddSubqueryInfo(&pObj->pHb->cmd); tscDebug("%p HB is allocated, pObj:%p", pObj->pHb, pObj); } int tscProcessConnectRsp(SSqlObj *pSql) { char temp[TSDB_TABLE_FNAME_LEN * 2]; STscObj *pObj = pSql->pTscObj; SSqlRes *pRes = &pSql->res; SCMConnectRsp *pConnect = (SCMConnectRsp *)pRes->pRsp; tstrncpy(pObj->acctId, pConnect->acctId, sizeof(pObj->acctId)); // copy acctId from response int32_t len = sprintf(temp, "%s%s%s", pObj->acctId, TS_PATH_DELIMITER, pObj->db); assert(len <= sizeof(pObj->db)); tstrncpy(pObj->db, temp, sizeof(pObj->db)); if (pConnect->epSet.numOfEps > 0) { tscEpSetHtons(&pConnect->epSet); tscUpdateMgmtEpSet(&pConnect->epSet); } strcpy(pObj->sversion, pConnect->serverVersion); pObj->writeAuth = pConnect->writeAuth; pObj->superAuth = pConnect->superAuth; pObj->connId = htonl(pConnect->connId); createHBObj(pObj); taosTmrReset(tscProcessActivityTimer, tsShellActivityTimer * 500, pObj, tscTmr, &pObj->pTimer); return 0; } int tscProcessUseDbRsp(SSqlObj *pSql) { STscObj * pObj = pSql->pTscObj; STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); tstrncpy(pObj->db, pTableMetaInfo->name, sizeof(pObj->db)); return 0; } int tscProcessDropDbRsp(SSqlObj *UNUSED_PARAM(pSql)) { taosCacheEmpty(tscCacheHandle); return 0; } int tscProcessDropTableRsp(SSqlObj *pSql) { STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); STableMeta *pTableMeta = taosCacheAcquireByKey(tscCacheHandle, pTableMetaInfo->name, strlen(pTableMetaInfo->name)); if (pTableMeta == NULL) { /* not in cache, abort */ return 0; } /* * 1. if a user drops one table, which is the only table in a vnode, remove operation will incur vnode to be removed. * 2. Then, a user creates a new metric followed by a table with identical name of removed table but different schema, * here the table will reside in a new vnode. * The cached information is expired, however, we may have lost the ref of original meter. So, clear whole cache * instead. */ tscDebug("%p force release table meta after drop table:%s", pSql, pTableMetaInfo->name); taosCacheRelease(tscCacheHandle, (void **)&pTableMeta, true); if (pTableMetaInfo->pTableMeta) { taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pTableMeta), true); } return 0; } int tscProcessAlterTableMsgRsp(SSqlObj *pSql) { STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(&pSql->cmd, 0, 0); STableMeta *pTableMeta = taosCacheAcquireByKey(tscCacheHandle, pTableMetaInfo->name, strlen(pTableMetaInfo->name)); if (pTableMeta == NULL) { /* not in cache, abort */ return 0; } tscDebug("%p force release metermeta in cache after alter-table: %s", pSql, pTableMetaInfo->name); taosCacheRelease(tscCacheHandle, (void **)&pTableMeta, true); if (pTableMetaInfo->pTableMeta) { bool isSuperTable = UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo); taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pTableMeta), true); if (isSuperTable) { // if it is a super table, reset whole query cache tscDebug("%p reset query cache since table:%s is stable", pSql, pTableMetaInfo->name); taosCacheEmpty(tscCacheHandle); } } return 0; } int tscProcessAlterDbMsgRsp(SSqlObj *pSql) { UNUSED(pSql); return 0; } int tscProcessQueryRsp(SSqlObj *pSql) { SSqlRes *pRes = &pSql->res; SQueryTableRsp *pQuery = (SQueryTableRsp *)pRes->pRsp; pQuery->qhandle = htobe64(pQuery->qhandle); pRes->qhandle = pQuery->qhandle; pRes->data = NULL; tscResetForNextRetrieve(pRes); return 0; } int tscProcessRetrieveRspFromNode(SSqlObj *pSql) { SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; SRetrieveTableRsp *pRetrieve = (SRetrieveTableRsp *)pRes->pRsp; pRes->numOfRows = htonl(pRetrieve->numOfRows); pRes->precision = htons(pRetrieve->precision); pRes->offset = htobe64(pRetrieve->offset); pRes->useconds = htobe64(pRetrieve->useconds); pRes->completed = (pRetrieve->completed == 1); pRes->data = pRetrieve->data; SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); if (tscCreateResPointerInfo(pRes, pQueryInfo) != TSDB_CODE_SUCCESS) { return pRes->code; } if (pSql->pSubscription != NULL) { int32_t numOfCols = pQueryInfo->fieldsInfo.numOfOutput; TAOS_FIELD *pField = tscFieldInfoGetField(&pQueryInfo->fieldsInfo, numOfCols - 1); int16_t offset = tscFieldInfoGetOffset(pQueryInfo, numOfCols - 1); char* p = pRes->data + (pField->bytes + offset) * pRes->numOfRows; int32_t numOfTables = htonl(*(int32_t*)p); p += sizeof(int32_t); for (int i = 0; i < numOfTables; i++) { int64_t uid = htobe64(*(int64_t*)p); p += sizeof(int64_t); p += sizeof(int32_t); // skip tid TSKEY key = htobe64(*(TSKEY*)p); p += sizeof(TSKEY); tscUpdateSubscriptionProgress(pSql->pSubscription, uid, key); } } pRes->row = 0; tscDebug("%p numOfRows:%" PRId64 ", offset:%" PRId64 ", complete:%d", pSql, pRes->numOfRows, pRes->offset, pRes->completed); return 0; } void tscTableMetaCallBack(void *param, TAOS_RES *res, int code); static int32_t getTableMetaFromMgmt(SSqlObj *pSql, STableMetaInfo *pTableMetaInfo) { SSqlObj *pNew = calloc(1, sizeof(SSqlObj)); if (NULL == pNew) { tscError("%p malloc failed for new sqlobj to get table meta", pSql); return TSDB_CODE_TSC_OUT_OF_MEMORY; } pNew->pTscObj = pSql->pTscObj; pNew->signature = pNew; pNew->cmd.command = TSDB_SQL_META; tscAddSubqueryInfo(&pNew->cmd); SQueryInfo *pNewQueryInfo = tscGetQueryInfoDetailSafely(&pNew->cmd, 0); pNew->cmd.autoCreated = pSql->cmd.autoCreated; // create table if not exists if (TSDB_CODE_SUCCESS != tscAllocPayload(&pNew->cmd, TSDB_DEFAULT_PAYLOAD_SIZE + pSql->cmd.payloadLen)) { tscError("%p malloc failed for payload to get table meta", pSql); free(pNew); return TSDB_CODE_TSC_OUT_OF_MEMORY; } STableMetaInfo *pNewMeterMetaInfo = tscAddEmptyMetaInfo(pNewQueryInfo); assert(pNew->cmd.numOfClause == 1 && pNewQueryInfo->numOfTables == 1); tstrncpy(pNewMeterMetaInfo->name, pTableMetaInfo->name, sizeof(pNewMeterMetaInfo->name)); memcpy(pNew->cmd.payload, pSql->cmd.payload, pSql->cmd.payloadLen); // tag information if table does not exists. pNew->cmd.payloadLen = pSql->cmd.payloadLen; tscDebug("%p new pSqlObj:%p to get tableMeta, auto create:%d", pSql, pNew, pNew->cmd.autoCreated); pNew->fp = tscTableMetaCallBack; pNew->param = pSql; int32_t code = tscProcessSql(pNew); if (code == TSDB_CODE_SUCCESS) { code = TSDB_CODE_TSC_ACTION_IN_PROGRESS; // notify upper application that current process need to be terminated } return code; } int32_t tscGetTableMeta(SSqlObj *pSql, STableMetaInfo *pTableMetaInfo) { assert(strlen(pTableMetaInfo->name) != 0); // If this STableMetaInfo owns a table meta, release it first if (pTableMetaInfo->pTableMeta != NULL) { taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pTableMeta), false); } pTableMetaInfo->pTableMeta = (STableMeta *)taosCacheAcquireByKey(tscCacheHandle, pTableMetaInfo->name, strlen(pTableMetaInfo->name)); if (pTableMetaInfo->pTableMeta != NULL) { STableComInfo tinfo = tscGetTableInfo(pTableMetaInfo->pTableMeta); tscDebug("%p retrieve table Meta from cache, the number of columns:%d, numOfTags:%d, %p", pSql, tinfo.numOfColumns, tinfo.numOfTags, pTableMetaInfo->pTableMeta); return TSDB_CODE_SUCCESS; } return getTableMetaFromMgmt(pSql, pTableMetaInfo); } int tscGetMeterMetaEx(SSqlObj *pSql, STableMetaInfo *pTableMetaInfo, bool createIfNotExists) { pSql->cmd.autoCreated = createIfNotExists; return tscGetTableMeta(pSql, pTableMetaInfo); } /** * retrieve table meta from mnode, and update the local table meta cache. * @param pSql sql object * @param tableIndex table index * @return status code */ int tscRenewTableMeta(SSqlObj *pSql, int32_t tableIndex) { SSqlCmd *pCmd = &pSql->cmd; SQueryInfo * pQueryInfo = tscGetQueryInfoDetail(pCmd, 0); STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, tableIndex); STableMeta* pTableMeta = pTableMetaInfo->pTableMeta; if (pTableMetaInfo->pTableMeta) { tscDebug("%p update table meta, old meta numOfTags:%d, numOfCols:%d, uid:%" PRId64 ", addr:%p", pSql, tscGetNumOfTags(pTableMeta), tscGetNumOfColumns(pTableMeta), pTableMeta->id.uid, pTableMeta); } taosCacheRelease(tscCacheHandle, (void **)&(pTableMetaInfo->pTableMeta), true); return getTableMetaFromMgmt(pSql, pTableMetaInfo); } static bool allVgroupInfoRetrieved(SSqlCmd* pCmd, int32_t clauseIndex) { SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, clauseIndex); for (int32_t i = 0; i < pQueryInfo->numOfTables; ++i) { STableMetaInfo *pTableMetaInfo = tscGetMetaInfo(pQueryInfo, i); if (pTableMetaInfo->vgroupList == NULL) { return false; } } // all super tables vgroupinfo are retrieved, no need to retrieve vgroup info anymore return true; } int tscGetSTableVgroupInfo(SSqlObj *pSql, int32_t clauseIndex) { int code = TSDB_CODE_RPC_NETWORK_UNAVAIL; SSqlCmd *pCmd = &pSql->cmd; if (allVgroupInfoRetrieved(pCmd, clauseIndex)) { return TSDB_CODE_SUCCESS; } SSqlObj *pNew = calloc(1, sizeof(SSqlObj)); pNew->pTscObj = pSql->pTscObj; pNew->signature = pNew; pNew->cmd.command = TSDB_SQL_STABLEVGROUP; SQueryInfo *pNewQueryInfo = tscGetQueryInfoDetailSafely(&pNew->cmd, 0); if (pNewQueryInfo == NULL) { tscFreeSqlObj(pNew); return code; } SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, clauseIndex); for (int32_t i = 0; i < pQueryInfo->numOfTables; ++i) { STableMetaInfo *pMInfo = tscGetMetaInfo(pQueryInfo, i); STableMeta *pTableMeta = taosCacheAcquireByData(tscCacheHandle, pMInfo->pTableMeta); tscAddTableMetaInfo(pNewQueryInfo, pMInfo->name, pTableMeta, NULL, pMInfo->tagColList); } if ((code = tscAllocPayload(&pNew->cmd, TSDB_DEFAULT_PAYLOAD_SIZE)) != TSDB_CODE_SUCCESS) { tscFreeSqlObj(pNew); return code; } pNewQueryInfo->numOfTables = pQueryInfo->numOfTables; tscDebug("%p new sqlObj:%p to get vgroupInfo, numOfTables:%d", pSql, pNew, pNewQueryInfo->numOfTables); pNew->fp = tscTableMetaCallBack; pNew->param = pSql; code = tscProcessSql(pNew); if (code == TSDB_CODE_SUCCESS) { code = TSDB_CODE_TSC_ACTION_IN_PROGRESS; } return code; } void tscInitMsgsFp() { tscBuildMsg[TSDB_SQL_SELECT] = tscBuildQueryMsg; tscBuildMsg[TSDB_SQL_INSERT] = tscBuildSubmitMsg; tscBuildMsg[TSDB_SQL_FETCH] = tscBuildFetchMsg; tscBuildMsg[TSDB_SQL_CREATE_DB] = tscBuildCreateDbMsg; tscBuildMsg[TSDB_SQL_CREATE_USER] = tscBuildUserMsg; tscBuildMsg[TSDB_SQL_CREATE_ACCT] = tscBuildAcctMsg; tscBuildMsg[TSDB_SQL_ALTER_ACCT] = tscBuildAcctMsg; tscBuildMsg[TSDB_SQL_CREATE_TABLE] = tscBuildCreateTableMsg; tscBuildMsg[TSDB_SQL_DROP_USER] = tscBuildDropUserMsg; tscBuildMsg[TSDB_SQL_DROP_ACCT] = tscBuildDropAcctMsg; tscBuildMsg[TSDB_SQL_DROP_DB] = tscBuildDropDbMsg; tscBuildMsg[TSDB_SQL_DROP_TABLE] = tscBuildDropTableMsg; tscBuildMsg[TSDB_SQL_ALTER_USER] = tscBuildUserMsg; tscBuildMsg[TSDB_SQL_CREATE_DNODE] = tscBuildCreateDnodeMsg; tscBuildMsg[TSDB_SQL_DROP_DNODE] = tscBuildDropDnodeMsg; tscBuildMsg[TSDB_SQL_CFG_DNODE] = tscBuildCfgDnodeMsg; tscBuildMsg[TSDB_SQL_ALTER_TABLE] = tscBuildAlterTableMsg; tscBuildMsg[TSDB_SQL_UPDATE_TAGS_VAL] = tscBuildUpdateTagMsg; tscBuildMsg[TSDB_SQL_ALTER_DB] = tscAlterDbMsg; tscBuildMsg[TSDB_SQL_CONNECT] = tscBuildConnectMsg; tscBuildMsg[TSDB_SQL_USE_DB] = tscBuildUseDbMsg; tscBuildMsg[TSDB_SQL_META] = tscBuildTableMetaMsg; tscBuildMsg[TSDB_SQL_STABLEVGROUP] = tscBuildSTableVgroupMsg; tscBuildMsg[TSDB_SQL_MULTI_META] = tscBuildMultiMeterMetaMsg; tscBuildMsg[TSDB_SQL_HB] = tscBuildHeartBeatMsg; tscBuildMsg[TSDB_SQL_SHOW] = tscBuildShowMsg; tscBuildMsg[TSDB_SQL_RETRIEVE] = tscBuildRetrieveFromMgmtMsg; tscBuildMsg[TSDB_SQL_KILL_QUERY] = tscBuildKillMsg; tscBuildMsg[TSDB_SQL_KILL_STREAM] = tscBuildKillMsg; tscBuildMsg[TSDB_SQL_KILL_CONNECTION] = tscBuildKillMsg; tscProcessMsgRsp[TSDB_SQL_SELECT] = tscProcessQueryRsp; tscProcessMsgRsp[TSDB_SQL_FETCH] = tscProcessRetrieveRspFromNode; tscProcessMsgRsp[TSDB_SQL_DROP_DB] = tscProcessDropDbRsp; tscProcessMsgRsp[TSDB_SQL_DROP_TABLE] = tscProcessDropTableRsp; tscProcessMsgRsp[TSDB_SQL_CONNECT] = tscProcessConnectRsp; tscProcessMsgRsp[TSDB_SQL_USE_DB] = tscProcessUseDbRsp; tscProcessMsgRsp[TSDB_SQL_META] = tscProcessTableMetaRsp; tscProcessMsgRsp[TSDB_SQL_STABLEVGROUP] = tscProcessSTableVgroupRsp; tscProcessMsgRsp[TSDB_SQL_MULTI_META] = tscProcessMultiMeterMetaRsp; tscProcessMsgRsp[TSDB_SQL_SHOW] = tscProcessShowRsp; tscProcessMsgRsp[TSDB_SQL_RETRIEVE] = tscProcessRetrieveRspFromNode; // rsp handled by same function. tscProcessMsgRsp[TSDB_SQL_DESCRIBE_TABLE] = tscProcessDescribeTableRsp; tscProcessMsgRsp[TSDB_SQL_CURRENT_DB] = tscProcessLocalRetrieveRsp; tscProcessMsgRsp[TSDB_SQL_CURRENT_USER] = tscProcessLocalRetrieveRsp; tscProcessMsgRsp[TSDB_SQL_SERV_VERSION] = tscProcessLocalRetrieveRsp; tscProcessMsgRsp[TSDB_SQL_CLI_VERSION] = tscProcessLocalRetrieveRsp; tscProcessMsgRsp[TSDB_SQL_SERV_STATUS] = tscProcessLocalRetrieveRsp; tscProcessMsgRsp[TSDB_SQL_RETRIEVE_EMPTY_RESULT] = tscProcessEmptyResultRsp; tscProcessMsgRsp[TSDB_SQL_RETRIEVE_LOCALMERGE] = tscProcessRetrieveLocalMergeRsp; tscProcessMsgRsp[TSDB_SQL_ALTER_TABLE] = tscProcessAlterTableMsgRsp; tscProcessMsgRsp[TSDB_SQL_ALTER_DB] = tscProcessAlterDbMsgRsp; tscKeepConn[TSDB_SQL_SHOW] = 1; tscKeepConn[TSDB_SQL_RETRIEVE] = 1; tscKeepConn[TSDB_SQL_SELECT] = 1; tscKeepConn[TSDB_SQL_FETCH] = 1; tscKeepConn[TSDB_SQL_HB] = 1; }