/* * 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 "tutil.h" #include "tnote.h" #include "trpc.h" #include "tscLog.h" #include "tscProfile.h" #include "tscSubquery.h" #include "tscSecondaryMerge.h" #include "tscUtil.h" #include "tsched.h" #include "tschemautil.h" #include "tsclient.h" static void tscProcessFetchRow(SSchedMsg *pMsg); static void tscAsyncQueryRowsForNextVnode(void *param, TAOS_RES *tres, int numOfRows); static void tscProcessAsyncRetrieveImpl(void *param, TAOS_RES *tres, int numOfRows, void (*fp)()); /* * Proxy function to perform sequentially query&retrieve operation. * If sql queries upon a super table and two-stage merge procedure is not involved (when employ the projection * query), it will sequentially query&retrieve data for all vnodes */ static void tscAsyncFetchRowsProxy(void *param, TAOS_RES *tres, int numOfRows); static void tscAsyncFetchSingleRowProxy(void *param, TAOS_RES *tres, int numOfRows); int doAsyncParseSql(SSqlObj* pSql) { SSqlCmd* pCmd = &pSql->cmd; SSqlRes* pRes = &pSql->res; int32_t code = tscAllocPayload(pCmd, TSDB_DEFAULT_PAYLOAD_SIZE); if (code != TSDB_CODE_SUCCESS) { tscError("failed to malloc payload"); tscQueueAsyncError(pSql->fp, pSql->param, TSDB_CODE_TSC_OUT_OF_MEMORY); return code; } pRes->qhandle = 0; pRes->numOfRows = 1; tscDump("%p SQL: %s", pSql, pSql->sqlstr); return tsParseSql(pSql, true); } void doAsyncQuery(STscObj* pObj, SSqlObj* pSql, void (*fp)(), void* param, const char* sqlstr, size_t sqlLen) { pSql->signature = pSql; pSql->param = param; pSql->pTscObj = pObj; pSql->maxRetry = TSDB_MAX_REPLICA_NUM; pSql->fp = fp; pSql->sqlstr = calloc(1, sqlLen + 1); if (pSql->sqlstr == NULL) { tscError("%p failed to malloc sql string buffer", pSql); tscQueueAsyncError(pSql->fp, pSql->param, TSDB_CODE_TSC_OUT_OF_MEMORY); return; } strtolower(pSql->sqlstr, sqlstr); int32_t code = doAsyncParseSql(pSql); if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return; if (code != TSDB_CODE_SUCCESS) { pSql->res.code = code; tscQueueAsyncRes(pSql); return; } tscDoQuery(pSql); } // TODO return the correct error code to client in tscQueueAsyncError void taos_query_a(TAOS *taos, const char *sqlstr, __async_cb_func_t fp, void *param) { STscObj *pObj = (STscObj *)taos; if (pObj == NULL || pObj->signature != pObj) { tscError("bug!!! pObj:%p", pObj); terrno = TSDB_CODE_TSC_DISCONNECTED; tscQueueAsyncError(fp, param, TSDB_CODE_TSC_DISCONNECTED); return; } int32_t sqlLen = strlen(sqlstr); if (sqlLen > tsMaxSQLStringLen) { tscError("sql string exceeds max length:%d", tsMaxSQLStringLen); terrno = TSDB_CODE_TSC_INVALID_SQL; tscQueueAsyncError(fp, param, TSDB_CODE_TSC_INVALID_SQL); return; } taosNotePrintTsc(sqlstr); SSqlObj *pSql = (SSqlObj *)calloc(1, sizeof(SSqlObj)); if (pSql == NULL) { tscError("failed to malloc sqlObj"); terrno = TSDB_CODE_TSC_OUT_OF_MEMORY; tscQueueAsyncError(fp, param, TSDB_CODE_TSC_OUT_OF_MEMORY); return; } doAsyncQuery(pObj, pSql, fp, param, sqlstr, sqlLen); } static void tscAsyncFetchRowsProxy(void *param, TAOS_RES *tres, int numOfRows) { if (tres == NULL) { return; } SSqlObj *pSql = (SSqlObj *)tres; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; if (numOfRows == 0) { if (hasMoreVnodesToTry(pSql)) { // sequentially retrieve data from remain vnodes. tscTryQueryNextVnode(pSql, tscAsyncQueryRowsForNextVnode); } else { /* * all available virtual node has been checked already, now we need to check * for the next subclause queries */ if (pCmd->clauseIndex < pCmd->numOfClause - 1) { tscTryQueryNextClause(pSql, tscAsyncQueryRowsForNextVnode); return; } /* * 1. has reach the limitation * 2. no remain virtual nodes to be retrieved anymore */ (*pSql->fetchFp)(param, pSql, 0); } return; } // local merge has handle this situation during super table non-projection query. if (pCmd->command != TSDB_SQL_RETRIEVE_LOCALMERGE) { pRes->numOfClauseTotal += pRes->numOfRows; } (*pSql->fetchFp)(param, tres, numOfRows); } // actual continue retrieve function with user-specified callback function static void tscProcessAsyncRetrieveImpl(void *param, TAOS_RES *tres, int numOfRows, void (*fp)()) { SSqlObj *pSql = (SSqlObj *)tres; if (pSql == NULL) { // error tscError("sql object is NULL"); return; } SSqlCmd *pCmd = &pSql->cmd; SSqlRes *pRes = &pSql->res; if ((pRes->qhandle == 0 || numOfRows != 0) && pCmd->command < TSDB_SQL_LOCAL) { if (pRes->qhandle == 0 && numOfRows != 0) { tscError("qhandle is NULL"); } else { pRes->code = numOfRows; } tscQueueAsyncError(pSql->fetchFp, param, pRes->code); return; } pSql->fp = fp; 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); } /* * retrieve callback for fetch rows proxy. * The below two functions both serve as the callback function of query virtual node. * query callback first, and then followed by retrieve callback */ static void tscAsyncQueryRowsForNextVnode(void *param, TAOS_RES *tres, int numOfRows) { // query completed, continue to retrieve tscProcessAsyncRetrieveImpl(param, tres, numOfRows, tscAsyncFetchRowsProxy); } void tscAsyncQuerySingleRowForNextVnode(void *param, TAOS_RES *tres, int numOfRows) { // query completed, continue to retrieve tscProcessAsyncRetrieveImpl(param, tres, numOfRows, tscAsyncFetchSingleRowProxy); } void taos_fetch_rows_a(TAOS_RES *taosa, void (*fp)(void *, TAOS_RES *, int), void *param) { SSqlObj *pSql = (SSqlObj *)taosa; if (pSql == NULL || pSql->signature != pSql) { tscError("sql object is NULL"); tscQueueAsyncError(fp, param, TSDB_CODE_TSC_DISCONNECTED); return; } SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; if (pRes->qhandle == 0) { tscError("qhandle is NULL"); tscQueueAsyncError(fp, param, TSDB_CODE_TSC_INVALID_QHANDLE); return; } // user-defined callback function is stored in fetchFp pSql->fetchFp = fp; pSql->fp = tscAsyncFetchRowsProxy; pSql->param = param; tscResetForNextRetrieve(pRes); // handle the sub queries of join query if (pCmd->command == TSDB_SQL_TABLE_JOIN_RETRIEVE) { tscFetchDatablockFromSubquery(pSql); } else if (pRes->completed && pCmd->command == TSDB_SQL_FETCH) { if (hasMoreVnodesToTry(pSql)) { // sequentially retrieve data from remain vnodes. tscTryQueryNextVnode(pSql, tscAsyncQueryRowsForNextVnode); return; } else { /* * all available virtual node has been checked already, now we need to check * for the next subclause queries */ if (pCmd->clauseIndex < pCmd->numOfClause - 1) { tscTryQueryNextClause(pSql, tscAsyncQueryRowsForNextVnode); return; } /* * 1. has reach the limitation * 2. no remain virtual nodes to be retrieved anymore */ (*pSql->fetchFp)(param, pSql, 0); } return; } else { // current query is not completed, continue retrieve from node 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); } } void taos_fetch_row_a(TAOS_RES *taosa, void (*fp)(void *, TAOS_RES *, TAOS_ROW), void *param) { SSqlObj *pSql = (SSqlObj *)taosa; if (pSql == NULL || pSql->signature != pSql) { tscError("sql object is NULL"); tscQueueAsyncError(fp, param, TSDB_CODE_TSC_DISCONNECTED); return; } SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; if (pRes->qhandle == 0) { tscError("qhandle is NULL"); tscQueueAsyncError(fp, param, TSDB_CODE_TSC_INVALID_QHANDLE); return; } pSql->fetchFp = fp; pSql->param = param; if (pRes->row >= pRes->numOfRows) { tscResetForNextRetrieve(pRes); pSql->fp = tscAsyncFetchSingleRowProxy; 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); } else { SSchedMsg schedMsg; schedMsg.fp = tscProcessFetchRow; schedMsg.ahandle = pSql; schedMsg.thandle = pRes->tsrow; schedMsg.msg = NULL; taosScheduleTask(tscQhandle, &schedMsg); } } void tscAsyncFetchSingleRowProxy(void *param, TAOS_RES *tres, int numOfRows) { SSqlObj *pSql = (SSqlObj *)tres; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); if (numOfRows == 0) { if (hasMoreVnodesToTry(pSql)) { // sequentially retrieve data from remain vnodes. tscTryQueryNextVnode(pSql, tscAsyncQuerySingleRowForNextVnode); } else { /* * 1. has reach the limitation * 2. no remain virtual nodes to be retrieved anymore */ (*pSql->fetchFp)(pSql->param, pSql, NULL); } return; } for (int i = 0; i < pCmd->numOfCols; ++i){ SFieldSupInfo* pSup = taosArrayGet(pQueryInfo->fieldsInfo.pSupportInfo, i); if (pSup->pSqlExpr != NULL) { // pRes->tsrow[i] = TSC_GET_RESPTR_BASE(pRes, pQueryInfo, i) + pSup->pSqlExpr->resBytes * pRes->row; } else { //todo add } } pRes->row++; (*pSql->fetchFp)(pSql->param, pSql, pSql->res.tsrow); } void tscProcessFetchRow(SSchedMsg *pMsg) { SSqlObj *pSql = (SSqlObj *)pMsg->ahandle; SSqlRes *pRes = &pSql->res; SSqlCmd *pCmd = &pSql->cmd; SQueryInfo *pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); for (int i = 0; i < pCmd->numOfCols; ++i) { SFieldSupInfo* pSup = taosArrayGet(pQueryInfo->fieldsInfo.pSupportInfo, i); if (pSup->pSqlExpr != NULL) { tscGetResultColumnChr(pRes, &pQueryInfo->fieldsInfo, i); } else { // todo add } } pRes->row++; (*pSql->fetchFp)(pSql->param, pSql, pRes->tsrow); } void tscProcessAsyncRes(SSchedMsg *pMsg) { SSqlObj *pSql = (SSqlObj *)pMsg->ahandle; SSqlCmd *pCmd = &pSql->cmd; SSqlRes *pRes = &pSql->res; void *taosres = pSql; // pCmd may be released, so cache pCmd->command 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; } (*pSql->fp)(pSql->param, taosres, code); if (shouldFree) { tscTrace("%p sqlObj is automatically freed in async res", pSql); tscFreeSqlObj(pSql); } } static void tscProcessAsyncError(SSchedMsg *pMsg) { void (*fp)() = pMsg->ahandle; (*fp)(pMsg->thandle, NULL, *(int32_t*)pMsg->msg); } void tscQueueAsyncError(void(*fp), void *param, int32_t code) { int32_t* c = malloc(sizeof(int32_t)); *c = code; SSchedMsg schedMsg; schedMsg.fp = tscProcessAsyncError; schedMsg.ahandle = fp; schedMsg.thandle = param; schedMsg.msg = c; taosScheduleTask(tscQhandle, &schedMsg); } void tscQueueAsyncRes(SSqlObj *pSql) { if (pSql == NULL || pSql->signature != pSql) { tscTrace("%p SqlObj is freed, not add into queue async res", pSql); return; } else { tscError("%p add into queued async res, code:%s", pSql, tstrerror(pSql->res.code)); } SSchedMsg schedMsg; schedMsg.fp = tscProcessAsyncRes; schedMsg.ahandle = pSql; schedMsg.thandle = (void *)1; schedMsg.msg = NULL; taosScheduleTask(tscQhandle, &schedMsg); } void tscProcessAsyncFree(SSchedMsg *pMsg) { SSqlObj *pSql = (SSqlObj *)pMsg->ahandle; tscTrace("%p sql is freed", pSql); taos_free_result(pSql); } void tscQueueAsyncFreeResult(SSqlObj *pSql) { tscTrace("%p sqlObj put in queue to async free", pSql); SSchedMsg schedMsg; schedMsg.fp = tscProcessAsyncFree; schedMsg.ahandle = pSql; schedMsg.thandle = (void *)1; schedMsg.msg = NULL; taosScheduleTask(tscQhandle, &schedMsg); } int tscSendMsgToServer(SSqlObj *pSql); void tscTableMetaCallBack(void *param, TAOS_RES *res, int code) { SSqlObj *pSql = (SSqlObj *)param; if (pSql == NULL || pSql->signature != pSql) return; SSqlCmd *pCmd = &pSql->cmd; SSqlRes *pRes = &pSql->res; if (code != TSDB_CODE_SUCCESS) { pRes->code = code; tscQueueAsyncRes(pSql); return; } if (pSql->pStream == NULL) { // check if it is a sub-query of super table query first, if true, enter another routine SQueryInfo* pQueryInfo = tscGetQueryInfoDetail(pCmd, pCmd->clauseIndex); if ((pQueryInfo->type & TSDB_QUERY_TYPE_STABLE_SUBQUERY) == TSDB_QUERY_TYPE_STABLE_SUBQUERY) { STableMetaInfo* pTableMetaInfo = tscGetMetaInfo(pQueryInfo, 0); if (pTableMetaInfo->pTableMeta == NULL){ code = tscGetTableMeta(pSql, pTableMetaInfo); assert(code == TSDB_CODE_SUCCESS); } assert((tscGetNumOfTags(pTableMetaInfo->pTableMeta) != 0) && pTableMetaInfo->vgroupIndex >= 0 && pSql->param != NULL); SRetrieveSupport *trs = (SRetrieveSupport *)pSql->param; SSqlObj * pParObj = trs->pParentSqlObj; assert(pParObj->signature == pParObj && trs->subqueryIndex == pTableMetaInfo->vgroupIndex && tscGetNumOfTags(pTableMetaInfo->pTableMeta) != 0); tscTrace("%p get metricMeta during super table query successfully", pSql); code = tscGetSTableVgroupInfo(pSql, 0); pRes->code = code; if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return; } else { // normal async query continues if (pCmd->parseFinished) { tscTrace("%p re-send data to vnode in table Meta callback since sql parsed completed", pSql); STableMetaInfo* pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); code = tscGetTableMeta(pSql, pTableMetaInfo); assert(code == TSDB_CODE_SUCCESS); if (pTableMetaInfo->pTableMeta) { // todo update the submit message according to the new table meta // 1. table uid, 2. ip address code = tscSendMsgToServer(pSql); if (code == TSDB_CODE_SUCCESS) return; } } else { code = tsParseSql(pSql, false); if ((pQueryInfo->type & TSDB_QUERY_TYPE_STMT_INSERT) == TSDB_QUERY_TYPE_STMT_INSERT) { STableMetaInfo* pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); code = tscGetTableMeta(pSql, pTableMetaInfo); assert(code == TSDB_CODE_SUCCESS && pTableMetaInfo->pTableMeta != NULL); (*pSql->fp)(pSql->param, pSql, code); return; } if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return; } } } else { // stream computing STableMetaInfo *pTableMetaInfo = tscGetTableMetaInfoFromCmd(pCmd, pCmd->clauseIndex, 0); code = tscGetTableMeta(pSql, pTableMetaInfo); pRes->code = code; if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return; if (code == TSDB_CODE_SUCCESS && UTIL_TABLE_IS_SUPER_TABLE(pTableMetaInfo)) { code = tscGetSTableVgroupInfo(pSql, pCmd->clauseIndex); pRes->code = code; if (code == TSDB_CODE_TSC_ACTION_IN_PROGRESS) return; } } if (code != TSDB_CODE_SUCCESS) { pSql->res.code = code; tscQueueAsyncRes(pSql); return; } if (pSql->pStream) { tscTrace("%p stream:%p meta is updated, start new query, command:%d", pSql, pSql->pStream, pSql->cmd.command); tsParseSql(pSql, false); sem_post(&pSql->rspSem); return; } else { tscTrace("%p get tableMeta successfully", pSql); } tscDoQuery(pSql); }