xxq250
/
homework-jianmu
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Merge branch '3.0' into feat/TS-2007

This commit is contained in:
Cary Xu 2022-11-10 15:49:15 +08:00
parent 8205dad2a1 11b20e149d
commit 69c1465142
157 changed files with 2679 additions and 3042 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Jenkinsfile2
View File

@ -430,8 +430,6 @@ pipeline {
rm -rf ${WKC}/debug
cd ${WKC}/tests/parallel_test
time ./container_build.sh -w ${WKDIR} -t 10 -e
rm -f /tmp/cases.task
./collect_cases.sh -e
'''
def extra_param = ""
def log_server_file = "/home/log_server.json"
@ -462,7 +460,7 @@ pipeline {
cd ${WKC}/tests/parallel_test
export DEFAULT_RETRY_TIME=2
date
''' + timeout_cmd + ''' time ./run.sh -e -m /home/m.json -t /tmp/cases.task -b ${BRANCH_NAME}_${BUILD_ID} -l ${WKDIR}/log -o 480 ''' + extra_param + '''
''' + timeout_cmd + ''' time ./run.sh -e -m /home/m.json -t cases.task -b ${BRANCH_NAME}_${BUILD_ID} -l ${WKDIR}/log -o 480 ''' + extra_param + '''
'''
}
}

15
cmake/cmake.version
View File

@ -65,13 +65,14 @@ ELSE ()
ENDIF ()
MESSAGE(STATUS "============= compile version parameter information start ============= ")
MESSAGE(STATUS "ver number:" ${TD_VER_NUMBER})
MESSAGE(STATUS "compatible ver number:" ${TD_VER_COMPATIBLE})
MESSAGE(STATUS "communit commit id:" ${TD_VER_GIT})
MESSAGE(STATUS "build date:" ${TD_VER_DATE})
MESSAGE(STATUS "ver type:" ${TD_VER_VERTYPE})
MESSAGE(STATUS "ver cpu:" ${TD_VER_CPUTYPE})
MESSAGE(STATUS "os type:" ${TD_VER_OSTYPE})
MESSAGE(STATUS "version: " ${TD_VER_NUMBER})
MESSAGE(STATUS "compatible: " ${TD_VER_COMPATIBLE})
MESSAGE(STATUS "commit id: " ${TD_VER_GIT})
MESSAGE(STATUS "build date: " ${TD_VER_DATE})
MESSAGE(STATUS "build type: " ${CMAKE_BUILD_TYPE})
MESSAGE(STATUS "type: " ${TD_VER_VERTYPE})
MESSAGE(STATUS "cpu: " ${TD_VER_CPUTYPE})
MESSAGE(STATUS "os: " ${TD_VER_OSTYPE})
MESSAGE(STATUS "============= compile version parameter information end ============= ")
STRING(REPLACE "." "_" TD_LIB_VER_NUMBER ${TD_VER_NUMBER})

2
include/common/tdatablock.h
View File

@ -244,7 +244,7 @@ int32_t blockDataAppendColInfo(SSDataBlock* pBlock, SColumnInfoData* pColIn
SColumnInfoData createColumnInfoData(int16_t type, int32_t bytes, int16_t colId);
SColumnInfoData* bdGetColumnInfoData(const SSDataBlock* pBlock, int32_t index);
void blockEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen, int32_t numOfCols, int8_t needCompress);
int32_t blockEncode(const SSDataBlock* pBlock, char* data, int32_t numOfCols);
const char* blockDecode(SSDataBlock* pBlock, const char* pData);
void blockDebugShowDataBlock(SSDataBlock* pBlock, const char* flag);

1
include/common/tmsg.h
View File

@ -190,6 +190,7 @@ typedef struct {
int64_t dbId;
int32_t vgVersion;
int32_t numOfTable; // unit is TSDB_TABLE_NUM_UNIT
int64_t stateTs;
} SBuildUseDBInput;
typedef struct SField {

3
include/libs/catalog/catalog.h
View File

@ -56,6 +56,7 @@ typedef struct SDbInfo {
int32_t vgVer;
int32_t tbNum;
int64_t dbId;
int64_t stateTs;
} SDbInfo;
typedef struct STablesReq {
@ -153,7 +154,7 @@ int32_t catalogInit(SCatalogCfg* cfg);
*/
int32_t catalogGetHandle(uint64_t clusterId, SCatalog** catalogHandle);
int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* version, int64_t* dbId, int32_t* tableNum);
int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* version, int64_t* dbId, int32_t* tableNum, int64_t* stateTs);
/**
* Get a DB's all vgroup info.

3
include/libs/qcom/query.h
View File

@ -127,6 +127,7 @@ typedef struct SDBVgInfo {
int16_t hashSuffix;
int8_t hashMethod;
int32_t numOfTable; // DB's table num, unit is TSDB_TABLE_NUM_UNIT
int64_t stateTs;
SHashObj* vgHash; // key:vgId, value:SVgroupInfo
} SDBVgInfo;
@ -203,6 +204,8 @@ int32_t taosAsyncExec(__async_exec_fn_t execFn, void* execParam, int32_t* code);
void destroySendMsgInfo(SMsgSendInfo* pMsgBody);
void destroyAhandle(void* ahandle);
int32_t asyncSendMsgToServerExt(void* pTransporter, SEpSet* epSet, int64_t* pTransporterId, SMsgSendInfo* pInfo,
bool persistHandle, void* ctx);

4
include/libs/transport/trpc.h
View File

@ -72,6 +72,7 @@ typedef struct SRpcMsg {
typedef void (*RpcCfp)(void *parent, SRpcMsg *, SEpSet *epset);
typedef bool (*RpcRfp)(int32_t code, tmsg_t msgType);
typedef bool (*RpcTfp)(int32_t code, tmsg_t msgType);
typedef void (*RpcDfp)(void *ahandle);
typedef struct SRpcInit {
char localFqdn[TSDB_FQDN_LEN];
@ -97,6 +98,9 @@ typedef struct SRpcInit {
// set up timeout for particular msg
RpcTfp tfp;
// destroy client ahandle;
RpcDfp dfp;
void *parent;
} SRpcInit;

2
include/util/tdef.h
View File

@ -290,7 +290,7 @@ typedef enum ELogicConditionType {
#define TSDB_DEFAULT_VN_PER_DB 2
#define TSDB_MIN_BUFFER_PER_VNODE 3 // unit MB
#define TSDB_MAX_BUFFER_PER_VNODE 16384 // unit MB
#define TSDB_DEFAULT_BUFFER_PER_VNODE 96
#define TSDB_DEFAULT_BUFFER_PER_VNODE 256
#define TSDB_MIN_PAGES_PER_VNODE 64
#define TSDB_MAX_PAGES_PER_VNODE (INT32_MAX - 1)
#define TSDB_DEFAULT_PAGES_PER_VNODE 256

3
source/client/src/clientEnv.c
View File

@ -140,12 +140,13 @@ void *openTransporter(const char *user, const char *auth, int32_t numOfThread) {
rpcInit.numOfThreads = numOfThread;
rpcInit.cfp = processMsgFromServer;
rpcInit.rfp = clientRpcRfp;
// rpcInit.tfp = clientRpcTfp;
rpcInit.sessions = 1024;
rpcInit.connType = TAOS_CONN_CLIENT;
rpcInit.user = (char *)user;
rpcInit.idleTime = tsShellActivityTimer * 1000;
rpcInit.compressSize = tsCompressMsgSize;
rpcInit.dfp = destroyAhandle;
void *pDnodeConn = rpcOpen(&rpcInit);
if (pDnodeConn == NULL) {
tscError("failed to init connection to server");

4
source/client/src/clientHb.c
View File

@ -61,7 +61,7 @@ static int32_t hbProcessDBInfoRsp(void *value, int32_t valueLen, struct SCatalog
int32_t numOfBatchs = taosArrayGetSize(batchUseRsp.pArray);
for (int32_t i = 0; i < numOfBatchs; ++i) {
SUseDbRsp *rsp = taosArrayGet(batchUseRsp.pArray, i);
tscDebug("hb db rsp, db:%s, vgVersion:%d, uid:%" PRIx64, rsp->db, rsp->vgVersion, rsp->uid);
tscDebug("hb db rsp, db:%s, vgVersion:%d, stateTs:%" PRId64 ", uid:%" PRIx64, rsp->db, rsp->vgVersion, rsp->stateTs, rsp->uid);
if (rsp->vgVersion < 0) {
code = catalogRemoveDB(pCatalog, rsp->db, rsp->uid);
@ -72,6 +72,7 @@ static int32_t hbProcessDBInfoRsp(void *value, int32_t valueLen, struct SCatalog
}
vgInfo->vgVersion = rsp->vgVersion;
vgInfo->stateTs = rsp->stateTs;
vgInfo->hashMethod = rsp->hashMethod;
vgInfo->hashPrefix = rsp->hashPrefix;
vgInfo->hashSuffix = rsp->hashSuffix;
@ -486,6 +487,7 @@ int32_t hbGetExpiredDBInfo(SClientHbKey *connKey, struct SCatalog *pCatalog, SCl
db->dbId = htobe64(db->dbId);
db->vgVersion = htonl(db->vgVersion);
db->numOfTable = htonl(db->numOfTable);
db->stateTs = htobe64(db->stateTs);
}
SKv kv = {

5
source/client/src/clientMsgHandler.c
View File

@ -181,7 +181,7 @@ int32_t processUseDbRsp(void* param, SDataBuf* pMsg, int32_t code) {
tDeserializeSUseDbRsp(pMsg->pData, pMsg->len, &usedbRsp);
struct SCatalog* pCatalog = NULL;
if (usedbRsp.vgVersion >= 0) {
if (usedbRsp.vgVersion >= 0) { // cached in local
uint64_t clusterId = pRequest->pTscObj->pAppInfo->clusterId;
int32_t code1 = catalogGetHandle(clusterId, &pCatalog);
if (code1 != TSDB_CODE_SUCCESS) {
@ -442,8 +442,7 @@ static int32_t buildShowVariablesRsp(SArray* pVars, SRetrieveTableRsp** pRsp) {
(*pRsp)->numOfRows = htonl(pBlock->info.rows);
(*pRsp)->numOfCols = htonl(SHOW_VARIABLES_RESULT_COLS);
int32_t len = 0;
blockEncode(pBlock, (*pRsp)->data, &len, SHOW_VARIABLES_RESULT_COLS, false);
int32_t len = blockEncode(pBlock, (*pRsp)->data, SHOW_VARIABLES_RESULT_COLS);
ASSERT(len == rspSize - sizeof(SRetrieveTableRsp));
blockDataDestroy(pBlock);

31
source/common/src/tdatablock.c
View File

@ -2197,7 +2197,9 @@ char* buildCtbNameByGroupId(const char* stbFullName, uint64_t groupId) {
return rname.ctbShortName;
}
void blockEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen, int32_t numOfCols, int8_t needCompress) {
int32_t blockEncode(const SSDataBlock* pBlock, char* data, int32_t numOfCols) {
int32_t dataLen = 0;
// todo extract method
int32_t* version = (int32_t*)data;
*version = 1;
@ -2238,7 +2240,7 @@ void blockEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen, int32_
int32_t* colSizes = (int32_t*)data;
data += numOfCols * sizeof(int32_t);
*dataLen = blockDataGetSerialMetaSize(numOfCols);
dataLen = blockDataGetSerialMetaSize(numOfCols);
int32_t numOfRows = pBlock->info.rows;
for (int32_t col = 0; col < numOfCols; ++col) {
@ -2255,26 +2257,23 @@ void blockEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen, int32_
}
data += metaSize;
(*dataLen) += metaSize;
dataLen += metaSize;
if (needCompress) {
colSizes[col] = blockCompressColData(pColRes, numOfRows, data, needCompress);
data += colSizes[col];
(*dataLen) += colSizes[col];
} else {
colSizes[col] = colDataGetLength(pColRes, numOfRows);
(*dataLen) += colSizes[col];
memmove(data, pColRes->pData, colSizes[col]);
data += colSizes[col];
}
colSizes[col] = colDataGetLength(pColRes, numOfRows);
dataLen += colSizes[col];
memmove(data, pColRes->pData, colSizes[col]);
data += colSizes[col];
colSizes[col] = htonl(colSizes[col]);
}
*actualLen = *dataLen;
*actualLen = dataLen;
*groupId = pBlock->info.groupId;
ASSERT(*dataLen > 0);
uDebug("build data block, actualLen:%d, rows:%d, cols:%d", *dataLen, *rows, *cols);
ASSERT(dataLen > 0);
uDebug("build data block, actualLen:%d, rows:%d, cols:%d", dataLen, *rows, *cols);
return dataLen;
}
const char* blockDecode(SSDataBlock* pBlock, const char* pData) {

3
source/dnode/mgmt/mgmt_dnode/src/dmHandle.c
View File

@ -307,8 +307,7 @@ int32_t dmProcessRetrieve(SDnodeMgmt *pMgmt, SRpcMsg *pMsg) {
pStart += sizeof(SSysTableSchema);
}
int32_t len = 0;
blockEncode(pBlock, pStart, &len, numOfCols, false);
int32_t len = blockEncode(pBlock, pStart, numOfCols);
pRsp->numOfRows = htonl(pBlock->info.rows);
pRsp->precision = TSDB_TIME_PRECISION_MILLI; // millisecond time precision

39
source/dnode/mnode/impl/src/mndDb.c
View File

@ -606,7 +606,7 @@ static int32_t mndProcessCreateDbReq(SRpcMsg *pReq) {
} else {
if (terrno == TSDB_CODE_MND_DB_IN_CREATING) {
if (mndSetRpcInfoForDbTrans(pMnode, pReq, MND_OPER_CREATE_DB, createReq.db) == 0) {
mInfo("db:%s, is creating and response after trans finished", createReq.db);
mInfo("db:%s, is creating and createdb response after trans finished", createReq.db);
code = TSDB_CODE_ACTION_IN_PROGRESS;
goto _OVER;
} else {
@ -1175,7 +1175,7 @@ int32_t mndExtractDbInfo(SMnode *pMnode, SDbObj *pDb, SUseDbRsp *pRsp, const SUs
int32_t numOfTable = mndGetDBTableNum(pDb, pMnode);
if (pReq == NULL || pReq->vgVersion < pDb->vgVersion || pReq->dbId != pDb->uid || numOfTable != pReq->numOfTable) {
if (pReq == NULL || pReq->vgVersion < pDb->vgVersion || pReq->dbId != pDb->uid || numOfTable != pReq->numOfTable || pReq->stateTs < pDb->stateTs) {
mndBuildDBVgroupInfo(pDb, pMnode, pRsp->pVgroupInfos);
}
@ -1225,6 +1225,14 @@ static int32_t mndProcessUseDbReq(SRpcMsg *pReq) {
usedbRsp.vgVersion = usedbReq.vgVersion;
usedbRsp.errCode = terrno;
if (terrno == TSDB_CODE_MND_DB_IN_CREATING) {
if (mndSetRpcInfoForDbTrans(pMnode, pReq, MND_OPER_CREATE_DB, usedbReq.db) == 0) {
mInfo("db:%s, is creating and usedb response after trans finished", usedbReq.db);
code = TSDB_CODE_ACTION_IN_PROGRESS;
goto _OVER;
}
}
mError("db:%s, failed to process use db req since %s", usedbReq.db, terrstr());
} else {
if (mndCheckDbPrivilege(pMnode, pReq->info.conn.user, MND_OPER_USE_DB, pDb) != 0) {
@ -1255,7 +1263,7 @@ static int32_t mndProcessUseDbReq(SRpcMsg *pReq) {
pReq->info.rspLen = contLen;
_OVER:
if (code != 0) {
if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
mError("db:%s, failed to process use db req since %s", usedbReq.db, terrstr());
}
@ -1275,12 +1283,31 @@ int32_t mndValidateDbInfo(SMnode *pMnode, SDbVgVersion *pDbs, int32_t numOfDbs,
for (int32_t i = 0; i < numOfDbs; ++i) {
SDbVgVersion *pDbVgVersion = &pDbs[i];
pDbVgVersion->dbId = htobe64(pDbVgVersion->dbId);
pDbVgVersion->dbId = be64toh(pDbVgVersion->dbId);
pDbVgVersion->vgVersion = htonl(pDbVgVersion->vgVersion);
pDbVgVersion->numOfTable = htonl(pDbVgVersion->numOfTable);
pDbVgVersion->stateTs = be64toh(pDbVgVersion->stateTs);
SUseDbRsp usedbRsp = {0};
if ((0 == strcasecmp(pDbVgVersion->dbFName, TSDB_INFORMATION_SCHEMA_DB) || (0 == strcasecmp(pDbVgVersion->dbFName, TSDB_PERFORMANCE_SCHEMA_DB)))) {
memcpy(usedbRsp.db, pDbVgVersion->dbFName, TSDB_DB_FNAME_LEN);
int32_t vgVersion = mndGetGlobalVgroupVersion(pMnode);
if (pDbVgVersion->vgVersion < vgVersion) {
usedbRsp.pVgroupInfos = taosArrayInit(10, sizeof(SVgroupInfo));
mndBuildDBVgroupInfo(NULL, pMnode, usedbRsp.pVgroupInfos);
usedbRsp.vgVersion = vgVersion++;
} else {
usedbRsp.vgVersion = pDbVgVersion->vgVersion;
}
usedbRsp.vgNum = taosArrayGetSize(usedbRsp.pVgroupInfos);
taosArrayPush(batchUseRsp.pArray, &usedbRsp);
continue;
}
SDbObj *pDb = mndAcquireDb(pMnode, pDbVgVersion->dbFName);
if (pDb == NULL) {
mTrace("db:%s, no exist", pDbVgVersion->dbFName);
@ -1293,8 +1320,8 @@ int32_t mndValidateDbInfo(SMnode *pMnode, SDbVgVersion *pDbs, int32_t numOfDbs,
int32_t numOfTable = mndGetDBTableNum(pDb, pMnode);
if (pDbVgVersion->vgVersion >= pDb->vgVersion && numOfTable == pDbVgVersion->numOfTable /* &&
pDbVgVersion->stateTs == pDb->stateTs */) {
if (pDbVgVersion->vgVersion >= pDb->vgVersion && numOfTable == pDbVgVersion->numOfTable &&
pDbVgVersion->stateTs == pDb->stateTs) {
mTrace("db:%s, valid dbinfo, vgVersion:%d stateTs:%" PRId64
" numOfTables:%d, not changed vgVersion:%d stateTs:%" PRId64 " numOfTables:%d",
pDbVgVersion->dbFName, pDbVgVersion->vgVersion, pDbVgVersion->stateTs, pDbVgVersion->numOfTable,

3
source/dnode/mnode/impl/src/mndShow.c
View File

@ -303,8 +303,7 @@ static int32_t mndProcessRetrieveSysTableReq(SRpcMsg *pReq) {
pStart += sizeof(SSysTableSchema);
}
int32_t len = 0;
blockEncode(pBlock, pStart, &len, pShow->pMeta->numOfColumns, false);
int32_t len = blockEncode(pBlock, pStart, pShow->pMeta->numOfColumns);
}
pRsp->numOfRows = htonl(rowsRead);

8
source/dnode/mnode/impl/src/mndTrans.c
View File

@ -938,11 +938,15 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
for (int32_t i = 0; i < size; ++i) {
SRpcHandleInfo *pInfo = taosArrayGet(pTrans->pRpcArray, i);
if (pInfo->handle != NULL) {
mInfo("trans:%d, send rsp, code:0x%x stage:%s app:%p", pTrans->id, code, mndTransStr(pTrans->stage),
pInfo->ahandle);
if (code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
code = TSDB_CODE_MND_TRANS_NETWORK_UNAVAILL;
}
if (i != 0 && code == 0) {
code = TSDB_CODE_RPC_REDIRECT;
}
mInfo("trans:%d, client:%d send rsp, code:0x%x stage:%s app:%p", pTrans->id, i, code, mndTransStr(pTrans->stage),
pInfo->ahandle);
SRpcMsg rspMsg = {.code = code, .info = *pInfo};
if (pTrans->originRpcType == TDMT_MND_CREATE_DB) {

44
source/dnode/vnode/src/inc/tsdb.h
View File

@ -110,7 +110,6 @@ static FORCE_INLINE int64_t tsdbLogicToFileSize(int64_t lSize, int32_t szPage) {
#define tsdbRowFromBlockData(BLOCKDATA, IROW) ((TSDBROW){.type = 1, .pBlockData = (BLOCKDATA), .iRow = (IROW)})
void tsdbRowGetColVal(TSDBROW *pRow, STSchema *pTSchema, int32_t iCol, SColVal *pColVal);
int32_t tPutTSDBRow(uint8_t *p, TSDBROW *pRow);
int32_t tGetTSDBRow(uint8_t *p, TSDBROW *pRow);
int32_t tsdbRowCmprFn(const void *p1, const void *p2);
// SRowIter
void tRowIterInit(SRowIter *pIter, TSDBROW *pRow, STSchema *pTSchema);
@ -210,11 +209,10 @@ void tsdbRefMemTable(SMemTable *pMemTable);
void tsdbUnrefMemTable(SMemTable *pMemTable);
SArray *tsdbMemTableGetTbDataArray(SMemTable *pMemTable);
// STbDataIter
int32_t tsdbTbDataIterCreate(STbData *pTbData, TSDBKEY *pFrom, int8_t backward, STbDataIter **ppIter);
void *tsdbTbDataIterDestroy(STbDataIter *pIter);
void tsdbTbDataIterOpen(STbData *pTbData, TSDBKEY *pFrom, int8_t backward, STbDataIter *pIter);
TSDBROW *tsdbTbDataIterGet(STbDataIter *pIter);
bool tsdbTbDataIterNext(STbDataIter *pIter);
int32_t tsdbTbDataIterCreate(STbData *pTbData, TSDBKEY *pFrom, int8_t backward, STbDataIter **ppIter);
void *tsdbTbDataIterDestroy(STbDataIter *pIter);
void tsdbTbDataIterOpen(STbData *pTbData, TSDBKEY *pFrom, int8_t backward, STbDataIter *pIter);
bool tsdbTbDataIterNext(STbDataIter *pIter);
// STbData
int32_t tsdbGetNRowsInTbData(STbData *pTbData);
// tsdbFile.c ==============================================================================================
@ -772,6 +770,40 @@ static FORCE_INLINE int32_t tsdbKeyCmprFn(const void *p1, const void *p2) {
return 0;
}
#define SL_NODE_FORWARD(n, l) ((n)->forwards[l])
#define SL_NODE_BACKWARD(n, l) ((n)->forwards[(n)->level + (l)])
#define SL_NODE_DATA(n) (&SL_NODE_BACKWARD(n, (n)->level))
static FORCE_INLINE int32_t tGetTSDBRow(uint8_t *p, TSDBROW *pRow) {
int32_t n = tGetI64(p, &pRow->version);
pRow->pTSRow = (STSRow *)(p + n);
n += pRow->pTSRow->len;
return n;
}
static FORCE_INLINE TSDBROW *tsdbTbDataIterGet(STbDataIter *pIter) {
if (pIter == NULL) return NULL;
if (pIter->pRow) {
return pIter->pRow;
}
if (pIter->backward) {
if (pIter->pNode == pIter->pTbData->sl.pHead) {
return NULL;
}
} else {
if (pIter->pNode == pIter->pTbData->sl.pTail) {
return NULL;
}
}
tGetTSDBRow((uint8_t *)SL_NODE_DATA(pIter->pNode), &pIter->row);
pIter->pRow = &pIter->row;
return pIter->pRow;
}
#ifdef __cplusplus
}
#endif

4
source/dnode/vnode/src/tq/tqExec.c
View File

@ -27,9 +27,7 @@ int32_t tqAddBlockDataToRsp(const SSDataBlock* pBlock, SMqDataRsp* pRsp, int32_t
pRetrieve->completed = 1;
pRetrieve->numOfRows = htonl(pBlock->info.rows);
// TODO enable compress
int32_t actualLen = 0;
blockEncode(pBlock, pRetrieve->data, &actualLen, numOfCols, false);
int32_t actualLen = blockEncode(pBlock, pRetrieve->data, numOfCols);
actualLen += sizeof(SRetrieveTableRsp);
ASSERT(actualLen <= dataStrLen);
taosArrayPush(pRsp->blockDataLen, &actualLen);

22
source/dnode/vnode/src/tsdb/tsdbCache.c
View File

@ -475,7 +475,7 @@ _err:
}
return code;
}
/*
static int32_t getTableDelIdx(SDelFReader *pDelFReader, tb_uid_t suid, tb_uid_t uid, SDelIdx *pDelIdx) {
int32_t code = 0;
SArray *pDelIdxArray = NULL;
@ -499,7 +499,7 @@ _err:
}
return code;
}
*/
typedef enum {
SFSLASTNEXTROW_FS,
SFSLASTNEXTROW_FILESET,
@ -997,8 +997,6 @@ static int32_t nextRowIterOpen(CacheNextRowIter *pIter, tb_uid_t uid, STsdb *pTs
pIter->pSkyline = taosArrayInit(32, sizeof(TSDBKEY));
SDelIdx delIdx;
SDelFile *pDelFile = pReadSnap->fs.pDelFile;
if (pDelFile) {
SDelFReader *pDelFReader;
@ -1006,18 +1004,20 @@ static int32_t nextRowIterOpen(CacheNextRowIter *pIter, tb_uid_t uid, STsdb *pTs
code = tsdbDelFReaderOpen(&pDelFReader, pDelFile, pTsdb);
if (code) goto _err;
code = getTableDelIdx(pDelFReader, suid, uid, &delIdx);
if (code) {
tsdbDelFReaderClose(&pDelFReader);
goto _err;
}
code = getTableDelSkyline(pMem, pIMem, pDelFReader, &delIdx, pIter->pSkyline);
SArray *pDelIdxArray = taosArrayInit(32, sizeof(SDelIdx));
code = tsdbReadDelIdx(pDelFReader, pDelIdxArray);
if (code) goto _err;
SDelIdx *delIdx = taosArraySearch(pDelIdxArray, &(SDelIdx){.suid = suid, .uid = uid}, tCmprDelIdx, TD_EQ);
code = getTableDelSkyline(pMem, pIMem, pDelFReader, delIdx, pIter->pSkyline);
if (code) {
tsdbDelFReaderClose(&pDelFReader);
goto _err;
}
taosArrayDestroy(pDelIdxArray);
tsdbDelFReaderClose(&pDelFReader);
} else {
code = getTableDelSkyline(pMem, pIMem, NULL, NULL, pIter->pSkyline);

25
source/dnode/vnode/src/tsdb/tsdbMemTable.c
View File

@ -294,31 +294,6 @@ bool tsdbTbDataIterNext(STbDataIter *pIter) {
return true;
}
TSDBROW *tsdbTbDataIterGet(STbDataIter *pIter) {
// we add here for commit usage
if (pIter == NULL) return NULL;
if (pIter->pRow) {
goto _exit;
}
if (pIter->backward) {
if (pIter->pNode == pIter->pTbData->sl.pHead) {
goto _exit;
}
} else {
if (pIter->pNode == pIter->pTbData->sl.pTail) {
goto _exit;
}
}
tGetTSDBRow((uint8_t *)SL_NODE_DATA(pIter->pNode), &pIter->row);
pIter->pRow = &pIter->row;
_exit:
return pIter->pRow;
}
static int32_t tsdbMemTableRehash(SMemTable *pMemTable) {
int32_t code = 0;

4
source/dnode/vnode/src/tsdb/tsdbRead.c
View File

@ -973,7 +973,7 @@ static void copyNumericCols(const SColData* pData, SFileBlockDumpInfo* pDumpInfo
int32_t mid = dumpedRows >> 1u;
int8_t* pts = (int8_t*)pColData->pData;
for (int32_t j = 0; j < mid; ++j) {
int64_t t = pts[j];
int8_t t = pts[j];
pts[j] = pts[dumpedRows - j - 1];
pts[dumpedRows - j - 1] = t;
}
@ -998,7 +998,7 @@ static void copyNumericCols(const SColData* pData, SFileBlockDumpInfo* pDumpInfo
int32_t mid = dumpedRows >> 1u;
int32_t* pts = (int32_t*)pColData->pData;
for (int32_t j = 0; j < mid; ++j) {
int64_t t = pts[j];
int32_t t = pts[j];
pts[j] = pts[dumpedRows - j - 1];
pts[dumpedRows - j - 1] = t;
}

12
source/dnode/vnode/src/tsdb/tsdbUtil.c
View File

@ -575,16 +575,6 @@ int32_t tPutTSDBRow(uint8_t *p, TSDBROW *pRow) {
return n;
}
int32_t tGetTSDBRow(uint8_t *p, TSDBROW *pRow) {
int32_t n = 0;
n += tGetI64(p, &pRow->version);
pRow->pTSRow = (STSRow *)(p + n);
n += pRow->pTSRow->len;
return n;
}
int32_t tsdbRowCmprFn(const void *p1, const void *p2) {
return tsdbKeyCmprFn(&TSDBROW_KEY((TSDBROW *)p1), &TSDBROW_KEY((TSDBROW *)p2));
}
@ -1053,7 +1043,7 @@ int32_t tBlockDataAppendRow(SBlockData *pBlockData, TSDBROW *pRow, STSchema *pTS
tRowIterInit(&rIter, pRow, pTSchema);
pColVal = tRowIterNext(&rIter);
for (int32_t iColData = 0; iColData < pBlockData->nColData; iColData++) {
SColData *pColData = tBlockDataGetColDataByIdx(pBlockData, iColData);
SColData *pColData = &((SColData *)pBlockData->aColData->pData)[iColData];
while (pColVal && pColVal->cid < pColData->cid) {
pColVal = tRowIterNext(&rIter);

6
source/dnode/vnode/src/vnd/vnodeModule.c
View File

@ -37,6 +37,12 @@ struct SVnodeGlobal vnodeGlobal;
static void* loop(void* arg);
static tsem_t canCommit = {0};
static void vnodeInitCommit() { tsem_init(&canCommit, 0, 4); };
void vnode_wait_commit() { tsem_wait(&canCommit); }
void vnode_done_commit() { tsem_wait(&canCommit); }
int vnodeInit(int nthreads) {
int8_t init;
int ret;

4
source/libs/catalog/src/catalog.c
View File

@ -715,10 +715,10 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* version, int64_t* dbId, int32_t* tableNum) {
int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* version, int64_t* dbId, int32_t* tableNum, int64_t* pStateTs) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == dbFName || NULL == version || NULL == dbId) {
if (NULL == pCtg || NULL == dbFName || NULL == version || NULL == dbId || NULL == tableNum || NULL == pStateTs) {
CTG_API_LEAVE(TSDB_CODE_CTG_INVALID_INPUT);
}

1
source/libs/catalog/src/ctgAsync.c
View File

@ -1998,6 +1998,7 @@ int32_t ctgLaunchGetDbInfoTask(SCtgTask* pTask) {
pInfo->vgVer = dbCache->vgCache.vgInfo->vgVersion;
pInfo->dbId = dbCache->dbId;
pInfo->tbNum = dbCache->vgCache.vgInfo->numOfTable;
pInfo->stateTs = dbCache->vgCache.vgInfo->stateTs;
ctgReleaseVgInfoToCache(pCtg, dbCache);
dbCache = NULL;

29
source/libs/catalog/src/ctgCache.c
View File

@ -1231,14 +1231,16 @@ int32_t ctgAddNewDBCache(SCatalog *pCtg, const char *dbFName, uint64_t dbId) {
CTG_CACHE_STAT_INC(numOfDb, 1);
SDbVgVersion vgVersion = {.dbId = newDBCache.dbId, .vgVersion = -1};
SDbVgVersion vgVersion = {.dbId = newDBCache.dbId, .vgVersion = -1, .stateTs = 0};
tstrncpy(vgVersion.dbFName, dbFName, sizeof(vgVersion.dbFName));
ctgDebug("db added to cache, dbFName:%s, dbId:0x%" PRIx64, dbFName, dbId);
CTG_ERR_RET(ctgMetaRentAdd(&pCtg->dbRent, &vgVersion, dbId, sizeof(SDbVgVersion)));
if (!IS_SYS_DBNAME(dbFName)) {
CTG_ERR_RET(ctgMetaRentAdd(&pCtg->dbRent, &vgVersion, dbId, sizeof(SDbVgVersion)));
ctgDebug("db added to rent, dbFName:%s, vgVersion:%d, dbId:0x%" PRIx64, dbFName, vgVersion.vgVersion, dbId);
ctgDebug("db added to rent, dbFName:%s, vgVersion:%d, dbId:0x%" PRIx64, dbFName, vgVersion.vgVersion, dbId);
}
return TSDB_CODE_SUCCESS;
@ -1563,7 +1565,7 @@ int32_t ctgOpUpdateVgroup(SCtgCacheOperation *operation) {
}
bool newAdded = false;
SDbVgVersion vgVersion = {.dbId = msg->dbId, .vgVersion = dbInfo->vgVersion, .numOfTable = dbInfo->numOfTable};
SDbVgVersion vgVersion = {.dbId = msg->dbId, .vgVersion = dbInfo->vgVersion, .numOfTable = dbInfo->numOfTable, .stateTs = dbInfo->stateTs};
SCtgDBCache *dbCache = NULL;
CTG_ERR_JRET(ctgGetAddDBCache(msg->pCtg, dbFName, msg->dbId, &dbCache));
@ -1579,15 +1581,15 @@ int32_t ctgOpUpdateVgroup(SCtgCacheOperation *operation) {
SDBVgInfo *vgInfo = vgCache->vgInfo;
if (dbInfo->vgVersion < vgInfo->vgVersion) {
ctgDebug("db vgVer is old, dbFName:%s, vgVer:%d, curVer:%d", dbFName, dbInfo->vgVersion, vgInfo->vgVersion);
ctgDebug("db updateVgroup is ignored, dbFName:%s, vgVer:%d, curVer:%d", dbFName, dbInfo->vgVersion, vgInfo->vgVersion);
ctgWUnlockVgInfo(dbCache);
goto _return;
}
if (dbInfo->vgVersion == vgInfo->vgVersion && dbInfo->numOfTable == vgInfo->numOfTable) {
ctgDebug("no new db vgVer or numOfTable, dbFName:%s, vgVer:%d, numOfTable:%d", dbFName, dbInfo->vgVersion,
dbInfo->numOfTable);
if (dbInfo->vgVersion == vgInfo->vgVersion && dbInfo->numOfTable == vgInfo->numOfTable && dbInfo->stateTs == vgInfo->stateTs) {
ctgDebug("no new db vgroup update info, dbFName:%s, vgVer:%d, numOfTable:%d, stateTs:%" PRId64, dbFName, dbInfo->vgVersion,
dbInfo->numOfTable, dbInfo->stateTs);
ctgWUnlockVgInfo(dbCache);
goto _return;
@ -1599,15 +1601,17 @@ int32_t ctgOpUpdateVgroup(SCtgCacheOperation *operation) {
vgCache->vgInfo = dbInfo;
msg->dbInfo = NULL;
ctgDebug("db vgInfo updated, dbFName:%s, vgVer:%d, dbId:0x%" PRIx64, dbFName, vgVersion.vgVersion, vgVersion.dbId);
ctgDebug("db vgInfo updated, dbFName:%s, vgVer:%d, stateTs:%" PRId64 ", dbId:0x%" PRIx64, dbFName, vgVersion.vgVersion, vgVersion.stateTs, vgVersion.dbId);
ctgWUnlockVgInfo(dbCache);
dbCache = NULL;
tstrncpy(vgVersion.dbFName, dbFName, sizeof(vgVersion.dbFName));
CTG_ERR_JRET(ctgMetaRentUpdate(&msg->pCtg->dbRent, &vgVersion, vgVersion.dbId, sizeof(SDbVgVersion),
ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare));
if (!IS_SYS_DBNAME(dbFName)) {
tstrncpy(vgVersion.dbFName, dbFName, sizeof(vgVersion.dbFName));
CTG_ERR_JRET(ctgMetaRentUpdate(&msg->pCtg->dbRent, &vgVersion, vgVersion.dbId, sizeof(SDbVgVersion),
ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare));
}
_return:
@ -2170,7 +2174,6 @@ void *ctgUpdateThreadFunc(void *param) {
CTG_RT_STAT_INC(numOfOpDequeue, 1);
ctgdShowCacheInfo();
ctgdShowClusterCache(pCtg);
}
qInfo("catalog update thread stopped");

26
source/libs/catalog/src/ctgDbg.c
View File

@ -78,7 +78,7 @@ void ctgdUserCallback(SMetaData *pResult, void *param, int32_t code) {
num = taosArrayGetSize(pResult->pDbInfo);
for (int32_t i = 0; i < num; ++i) {
SDbInfo *pDb = taosArrayGet(pResult->pDbInfo, i);
qDebug("db %d dbInfo: vgVer:%d, tbNum:%d, dbId:0x%" PRIx64, i, pDb->vgVer, pDb->tbNum, pDb->dbId);
qDebug("db %d dbInfo: vgVer:%d, tbNum:%d, stateTs:%" PRId64 " dbId:0x%" PRIx64, i, pDb->vgVer, pDb->tbNum, pDb->stateTs, pDb->dbId);
}
} else {
qDebug("empty db info");
@ -462,6 +462,7 @@ void ctgdShowDBCache(SCatalog *pCtg, SHashObj *dbHash) {
int32_t hashMethod = -1;
int16_t hashPrefix = 0;
int16_t hashSuffix = 0;
int64_t stateTs = 0;
int32_t vgNum = 0;
if (dbCache->vgCache.vgInfo) {
@ -469,16 +470,35 @@ void ctgdShowDBCache(SCatalog *pCtg, SHashObj *dbHash) {
hashMethod = dbCache->vgCache.vgInfo->hashMethod;
hashPrefix = dbCache->vgCache.vgInfo->hashPrefix;
hashSuffix = dbCache->vgCache.vgInfo->hashSuffix;
stateTs = dbCache->vgCache.vgInfo->stateTs;
if (dbCache->vgCache.vgInfo->vgHash) {
vgNum = taosHashGetSize(dbCache->vgCache.vgInfo->vgHash);
}
}
ctgDebug("[%d] db [%.*s][0x%" PRIx64
"] %s: metaNum:%d, stbNum:%d, vgVersion:%d, hashMethod:%d, prefix:%d, suffix:%d, vgNum:%d",
i, (int32_t)len, dbFName, dbCache->dbId, dbCache->deleted ? "deleted" : "", metaNum, stbNum, vgVersion,
"] %s: metaNum:%d, stbNum:%d, vgVersion:%d, stateTs:%" PRId64 ", hashMethod:%d, prefix:%d, suffix:%d, vgNum:%d",
i, (int32_t)len, dbFName, dbCache->dbId, dbCache->deleted ? "deleted" : "", metaNum, stbNum, vgVersion, stateTs,
hashMethod, hashPrefix, hashSuffix, vgNum);
if (dbCache->vgCache.vgInfo) {
int32_t i = 0;
void *pVgIter = taosHashIterate(dbCache->vgCache.vgInfo->vgHash, NULL);
while (pVgIter) {
SVgroupInfo * pVg = (SVgroupInfo *)pVgIter;
ctgDebug("The %04dth VG [id:%d, hashBegin:%u, hashEnd:%u, numOfTable:%d, epNum:%d, inUse:%d]",
i++, pVg->vgId, pVg->hashBegin, pVg->hashEnd, pVg->numOfTable, pVg->epSet.numOfEps, pVg->epSet.inUse);
for (int32_t n = 0; n < pVg->epSet.numOfEps; ++n) {
SEp *pEp = &pVg->epSet.eps[n];
ctgDebug("\tEp %d [fqdn:%s, port:%d]", n, pEp->fqdn, pEp->port);
}
pVgIter = taosHashIterate(dbCache->vgCache.vgInfo->vgHash, pVgIter);
}
}
pIter = taosHashIterate(dbHash, pIter);
}
}

3
source/libs/catalog/test/catalogTests.cpp
View File

@ -2486,7 +2486,8 @@ TEST(dbVgroup, getSetDbVgroupCase) {
int32_t dbVer = 0;
int64_t dbId = 0;
int32_t tbNum = 0;
code = catalogGetDBVgVersion(pCtg, ctgTestDbname, &dbVer, &dbId, &tbNum);
int64_t stateTs = 0;
code = catalogGetDBVgVersion(pCtg, ctgTestDbname, &dbVer, &dbId, &tbNum, &stateTs);
ASSERT_EQ(code, 0);
ASSERT_EQ(dbVer, ctgTestVgVersion);
ASSERT_EQ(dbId, ctgTestDbId);

3
source/libs/command/src/command.c
View File

@ -39,8 +39,7 @@ static int32_t buildRetrieveTableRsp(SSDataBlock* pBlock, int32_t numOfCols, SRe
(*pRsp)->numOfRows = htonl(pBlock->info.rows);
(*pRsp)->numOfCols = htonl(numOfCols);
int32_t len = 0;
blockEncode(pBlock, (*pRsp)->data, &len, numOfCols, false);
int32_t len = blockEncode(pBlock, (*pRsp)->data, numOfCols);
ASSERT(len == rspSize - sizeof(SRetrieveTableRsp));
return TSDB_CODE_SUCCESS;

3
source/libs/command/src/explain.c
View File

@ -1610,8 +1610,7 @@ int32_t qExplainGetRspFromCtx(void *ctx, SRetrieveTableRsp **pRsp) {
rsp->completed = 1;
rsp->numOfRows = htonl(rowNum);
int32_t len = 0;
blockEncode(pBlock, rsp->data, &len, taosArrayGetSize(pBlock->pDataBlock), 0);
int32_t len = blockEncode(pBlock, rsp->data, taosArrayGetSize(pBlock->pDataBlock));
ASSERT(len == rspSize - sizeof(SRetrieveTableRsp));
rsp->compLen = htonl(len);

12
source/libs/executor/inc/executorimpl.h
View File

@ -842,10 +842,8 @@ typedef struct SJoinOperatorInfo {
#define OPTR_IS_OPENED(_optr) (((_optr)->status & OP_OPENED) == OP_OPENED)
#define OPTR_SET_OPENED(_optr) ((_optr)->status |= OP_OPENED)
void doDestroyExchangeOperatorInfo(void* param);
SOperatorFpSet createOperatorFpSet(__optr_open_fn_t openFn, __optr_fn_t nextFn, __optr_fn_t streamFn,
__optr_fn_t cleanup, __optr_close_fn_t closeFn, __optr_explain_fn_t explain);
SOperatorFpSet createOperatorFpSet(__optr_open_fn_t openFn, __optr_fn_t nextFn, __optr_fn_t cleanup,
__optr_close_fn_t closeFn, __optr_explain_fn_t explain);
int32_t operatorDummyOpenFn(SOperatorInfo* pOperator);
int32_t appendDownstream(SOperatorInfo* p, SOperatorInfo** pDownstream, int32_t num);
@ -881,7 +879,11 @@ STimeWindow getFirstQualifiedTimeWindow(int64_t ts, STimeWindow* pWindow, SInter
int32_t getTableScanInfo(SOperatorInfo* pOperator, int32_t* order, int32_t* scanFlag);
int32_t getBufferPgSize(int32_t rowSize, uint32_t* defaultPgsz, uint32_t* defaultBufsz);
void doSetOperatorCompleted(SOperatorInfo* pOperator);
void doDestroyExchangeOperatorInfo(void* param);
void setOperatorCompleted(SOperatorInfo* pOperator);
void setOperatorInfo(SOperatorInfo* pOperator, const char* name, int32_t type, bool blocking, int32_t status, void* pInfo,
SExecTaskInfo* pTaskInfo);
void doFilter(SSDataBlock* pBlock, SFilterInfo* pFilterInfo, SColMatchInfo* pColMatchInfo);
int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int32_t numOfExpr,
SSDataBlock* pBlock, int32_t rows, const char* idStr, STableMetaCacheInfo * pCache);

15
source/libs/executor/src/cachescanoperator.c
View File

@ -93,16 +93,11 @@ SOperatorInfo* createCacherowsScanOperator(SLastRowScanPhysiNode* pScanNode, SRe
p->pCtx = createSqlFunctionCtx(p->pExprInfo, p->numOfExprs, &p->rowEntryInfoOffset);
}
pOperator->name = "LastrowScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_LAST_ROW_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "CachedRowScanOperator", QUERY_NODE_PHYSICAL_PLAN_LAST_ROW_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = taosArrayGetSize(pInfo->pRes->pDataBlock);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doScanCache, NULL, NULL, destroyLastrowScanOperator, NULL);
createOperatorFpSet(operatorDummyOpenFn, doScanCache, NULL, destroyLastrowScanOperator, NULL);
pOperator->cost.openCost = 0;
return pOperator;
@ -126,7 +121,7 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
uint64_t suid = tableListGetSuid(pTableList);
int32_t size = tableListGetSize(pTableList);
if (size == 0) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -182,7 +177,7 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
pInfo->indexOfBufferedRes += 1;
return pRes;
} else {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
} else {
@ -234,7 +229,7 @@ SSDataBlock* doScanCache(SOperatorInfo* pOperator) {
}
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
}

2
source/libs/executor/src/dataDispatcher.c
View File

@ -76,7 +76,7 @@ static void toDataCacheEntry(SDataDispatchHandle* pHandle, const SInputData* pIn
pEntry->dataLen = 0;
pBuf->useSize = sizeof(SDataCacheEntry);
blockEncode(pInput->pData, pEntry->data, &pEntry->dataLen, numOfCols, pEntry->compressed);
pEntry->dataLen = blockEncode(pInput->pData, pEntry->data, numOfCols);
ASSERT(pEntry->numOfRows == *(int32_t*)(pEntry->data + 8));
ASSERT(pEntry->numOfCols == *(int32_t*)(pEntry->data + 8 + 4));

638
source/libs/executor/src/exchangeoperator.c Normal file
View File

@ -0,0 +1,638 @@
/*
* Copyright (c) 2019 TAOS Data, Inc. <jhtao@taosdata.com>
*
* 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 <http://www.gnu.org/licenses/>.
*/
#include "filter.h"
#include "function.h"
#include "functionMgt.h"
#include "os.h"
#include "querynodes.h"
#include "tfill.h"
#include "tname.h"
#include "tref.h"
#include "tdatablock.h"
#include "tglobal.h"
#include "tmsg.h"
#include "tsort.h"
#include "ttime.h"
#include "executorimpl.h"
#include "index.h"
#include "query.h"
#include "tcompare.h"
#include "thash.h"
#include "ttypes.h"
#include "vnode.h"
typedef struct SFetchRspHandleWrapper {
uint32_t exchangeId;
int32_t sourceIndex;
} SFetchRspHandleWrapper;
static void destroyExchangeOperatorInfo(void* param);
static void freeBlock(void* pParam);
static void freeSourceDataInfo(void* param);
static void* setAllSourcesCompleted(SOperatorInfo* pOperator, int64_t startTs);
static int32_t loadRemoteDataCallback(void* param, SDataBuf* pMsg, int32_t code);
static int32_t doSendFetchDataRequest(SExchangeInfo* pExchangeInfo, SExecTaskInfo* pTaskInfo, int32_t sourceIndex);
static int32_t getCompletedSources(const SArray* pArray);
static int32_t prepareConcurrentlyLoad(SOperatorInfo* pOperator);
static int32_t seqLoadRemoteData(SOperatorInfo* pOperator);
static int32_t prepareLoadRemoteData(SOperatorInfo* pOperator);
static void concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SExchangeInfo* pExchangeInfo,
SExecTaskInfo* pTaskInfo) {
int32_t code = 0;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSourceDataInfo);
int32_t completed = getCompletedSources(pExchangeInfo->pSourceDataInfo);
if (completed == totalSources) {
setAllSourcesCompleted(pOperator, pExchangeInfo->openedTs);
return;
}
while (1) {
tsem_wait(&pExchangeInfo->ready);
for (int32_t i = 0; i < totalSources; ++i) {
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, i);
if (pDataInfo->status == EX_SOURCE_DATA_EXHAUSTED) {
continue;
}
if (pDataInfo->status != EX_SOURCE_DATA_READY) {
continue;
}
if (pDataInfo->code != TSDB_CODE_SUCCESS) {
code = pDataInfo->code;
goto _error;
}
SRetrieveTableRsp* pRsp = pDataInfo->pRsp;
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, i);
// todo
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pRsp->numOfRows == 0) {
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
qDebug("%s vgId:%d, taskId:0x%" PRIx64 " execId:%d index:%d completed, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 ", try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pDataInfo->totalRows,
pExchangeInfo->loadInfo.totalRows, i + 1, totalSources);
taosMemoryFreeClear(pDataInfo->pRsp);
break;
}
SRetrieveTableRsp* pRetrieveRsp = pDataInfo->pRsp;
int32_t index = 0;
char* pStart = pRetrieveRsp->data;
while (index++ < pRetrieveRsp->numOfBlocks) {
SSDataBlock* pb = createOneDataBlock(pExchangeInfo->pDummyBlock, false);
code = extractDataBlockFromFetchRsp(pb, pStart, NULL, &pStart);
if (code != 0) {
taosMemoryFreeClear(pDataInfo->pRsp);
goto _error;
}
taosArrayPush(pExchangeInfo->pResultBlockList, &pb);
}
updateLoadRemoteInfo(pLoadInfo, pRetrieveRsp->numOfRows, pRetrieveRsp->compLen, pExchangeInfo->openedTs, pOperator);
if (pRsp->completed == 1) {
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64
" execId:%d index:%d completed, blocks:%d, numOfRows:%d, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64
", total:%.2f Kb, try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pRsp->numOfBlocks,
pRsp->numOfRows, pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize / 1024.0,
i + 1, totalSources);
} else {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64
" execId:%d blocks:%d, numOfRows:%d, totalRows:%" PRIu64 ", total:%.2f Kb",
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRsp->numOfBlocks,
pRsp->numOfRows, pLoadInfo->totalRows, pLoadInfo->totalSize / 1024.0);
}
taosMemoryFreeClear(pDataInfo->pRsp);
if (pDataInfo->status != EX_SOURCE_DATA_EXHAUSTED) {
pDataInfo->status = EX_SOURCE_DATA_NOT_READY;
code = doSendFetchDataRequest(pExchangeInfo, pTaskInfo, i);
if (code != TSDB_CODE_SUCCESS) {
taosMemoryFreeClear(pDataInfo->pRsp);
goto _error;
}
}
return;
} // end loop
int32_t complete1 = getCompletedSources(pExchangeInfo->pSourceDataInfo);
if (complete1 == totalSources) {
qDebug("all sources are completed, %s", GET_TASKID(pTaskInfo));
return;
}
}
_error:
pTaskInfo->code = code;
}
static SSDataBlock* doLoadRemoteDataImpl(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
pTaskInfo->code = pOperator->fpSet._openFn(pOperator);
if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
return NULL;
}
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pOperator->status == OP_EXEC_DONE) {
qDebug("%s all %" PRIzu " source(s) are exhausted, total rows:%" PRIu64 " bytes:%" PRIu64 ", elapsed:%.2f ms",
GET_TASKID(pTaskInfo), totalSources, pLoadInfo->totalRows, pLoadInfo->totalSize,
pLoadInfo->totalElapsed / 1000.0);
return NULL;
}
size_t size = taosArrayGetSize(pExchangeInfo->pResultBlockList);
if (size == 0 || pExchangeInfo->rspBlockIndex >= size) {
pExchangeInfo->rspBlockIndex = 0;
taosArrayClearEx(pExchangeInfo->pResultBlockList, freeBlock);
if (pExchangeInfo->seqLoadData) {
seqLoadRemoteData(pOperator);
} else {
concurrentlyLoadRemoteDataImpl(pOperator, pExchangeInfo, pTaskInfo);
}
if (taosArrayGetSize(pExchangeInfo->pResultBlockList) == 0) {
return NULL;
}
}
// we have buffered retrieved datablock, return it directly
return taosArrayGetP(pExchangeInfo->pResultBlockList, pExchangeInfo->rspBlockIndex++);
}
static SSDataBlock* doLoadRemoteData(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
if (pOperator->status == OP_EXEC_DONE) {
return NULL;
}
while (1) {
SSDataBlock* pBlock = doLoadRemoteDataImpl(pOperator);
if (pBlock == NULL) {
return NULL;
}
SLimitInfo* pLimitInfo = &pExchangeInfo->limitInfo;
if (hasLimitOffsetInfo(pLimitInfo)) {
int32_t status = handleLimitOffset(pOperator, pLimitInfo, pBlock, false);
if (status == PROJECT_RETRIEVE_CONTINUE) {
continue;
} else if (status == PROJECT_RETRIEVE_DONE) {
size_t rows = pBlock->info.rows;
pExchangeInfo->limitInfo.numOfOutputRows += rows;
if (rows == 0) {
setOperatorCompleted(pOperator);
return NULL;
} else {
return pBlock;
}
}
} else {
return pBlock;
}
}
}
static int32_t initDataSource(int32_t numOfSources, SExchangeInfo* pInfo, const char* id) {
pInfo->pSourceDataInfo = taosArrayInit(numOfSources, sizeof(SSourceDataInfo));
if (pInfo->pSourceDataInfo == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
for (int32_t i = 0; i < numOfSources; ++i) {
SSourceDataInfo dataInfo = {0};
dataInfo.status = EX_SOURCE_DATA_NOT_READY;
dataInfo.taskId = id;
dataInfo.index = i;
SSourceDataInfo* pDs = taosArrayPush(pInfo->pSourceDataInfo, &dataInfo);
if (pDs == NULL) {
taosArrayDestroy(pInfo->pSourceDataInfo);
return TSDB_CODE_OUT_OF_MEMORY;
}
}
return TSDB_CODE_SUCCESS;
}
static int32_t initExchangeOperator(SExchangePhysiNode* pExNode, SExchangeInfo* pInfo, const char* id) {
size_t numOfSources = LIST_LENGTH(pExNode->pSrcEndPoints);
if (numOfSources == 0) {
qError("%s invalid number: %d of sources in exchange operator", id, (int32_t)numOfSources);
return TSDB_CODE_INVALID_PARA;
}
pInfo->pSources = taosArrayInit(numOfSources, sizeof(SDownstreamSourceNode));
if (pInfo->pSources == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
for (int32_t i = 0; i < numOfSources; ++i) {
SDownstreamSourceNode* pNode = (SDownstreamSourceNode*)nodesListGetNode((SNodeList*)pExNode->pSrcEndPoints, i);
taosArrayPush(pInfo->pSources, pNode);
}
initLimitInfo(pExNode->node.pLimit, pExNode->node.pSlimit, &pInfo->limitInfo);
pInfo->self = taosAddRef(exchangeObjRefPool, pInfo);
return initDataSource(numOfSources, pInfo, id);
}
SOperatorInfo* createExchangeOperatorInfo(void* pTransporter, SExchangePhysiNode* pExNode, SExecTaskInfo* pTaskInfo) {
SExchangeInfo* pInfo = taosMemoryCalloc(1, sizeof(SExchangeInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
if (pInfo == NULL || pOperator == NULL) {
goto _error;
}
int32_t code = initExchangeOperator(pExNode, pInfo, GET_TASKID(pTaskInfo));
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
tsem_init(&pInfo->ready, 0, 0);
pInfo->pDummyBlock = createResDataBlock(pExNode->node.pOutputDataBlockDesc);
pInfo->pResultBlockList = taosArrayInit(1, POINTER_BYTES);
pInfo->seqLoadData = false;
pInfo->pTransporter = pTransporter;
setOperatorInfo(pOperator, "ExchangeOperator", QUERY_NODE_PHYSICAL_PLAN_EXCHANGE, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = taosArrayGetSize(pInfo->pDummyBlock->pDataBlock);
pOperator->fpSet = createOperatorFpSet(prepareLoadRemoteData, doLoadRemoteData, NULL, destroyExchangeOperatorInfo, NULL);
return pOperator;
_error:
if (pInfo != NULL) {
doDestroyExchangeOperatorInfo(pInfo);
}
taosMemoryFreeClear(pOperator);
pTaskInfo->code = code;
return NULL;
}
void destroyExchangeOperatorInfo(void* param) {
SExchangeInfo* pExInfo = (SExchangeInfo*)param;
taosRemoveRef(exchangeObjRefPool, pExInfo->self);
}
void freeBlock(void* pParam) {
SSDataBlock* pBlock = *(SSDataBlock**)pParam;
blockDataDestroy(pBlock);
}
void freeSourceDataInfo(void* p) {
SSourceDataInfo* pInfo = (SSourceDataInfo*)p;
taosMemoryFreeClear(pInfo->pRsp);
}
void doDestroyExchangeOperatorInfo(void* param) {
SExchangeInfo* pExInfo = (SExchangeInfo*)param;
taosArrayDestroy(pExInfo->pSources);
taosArrayDestroyEx(pExInfo->pSourceDataInfo, freeSourceDataInfo);
if (pExInfo->pResultBlockList != NULL) {
taosArrayDestroyEx(pExInfo->pResultBlockList, freeBlock);
pExInfo->pResultBlockList = NULL;
}
blockDataDestroy(pExInfo->pDummyBlock);
tsem_destroy(&pExInfo->ready);
taosMemoryFreeClear(param);
}
int32_t loadRemoteDataCallback(void* param, SDataBuf* pMsg, int32_t code) {
SFetchRspHandleWrapper* pWrapper = (SFetchRspHandleWrapper*)param;
SExchangeInfo* pExchangeInfo = taosAcquireRef(exchangeObjRefPool, pWrapper->exchangeId);
if (pExchangeInfo == NULL) {
qWarn("failed to acquire exchange operator, since it may have been released");
taosMemoryFree(pMsg->pData);
return TSDB_CODE_SUCCESS;
}
int32_t index = pWrapper->sourceIndex;
SSourceDataInfo* pSourceDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, index);
if (code == TSDB_CODE_SUCCESS) {
pSourceDataInfo->pRsp = pMsg->pData;
SRetrieveTableRsp* pRsp = pSourceDataInfo->pRsp;
pRsp->numOfRows = htonl(pRsp->numOfRows);
pRsp->compLen = htonl(pRsp->compLen);
pRsp->numOfCols = htonl(pRsp->numOfCols);
pRsp->useconds = htobe64(pRsp->useconds);
pRsp->numOfBlocks = htonl(pRsp->numOfBlocks);
ASSERT(pRsp != NULL);
qDebug("%s fetch rsp received, index:%d, blocks:%d, rows:%d", pSourceDataInfo->taskId, index, pRsp->numOfBlocks,
pRsp->numOfRows);
} else {
taosMemoryFree(pMsg->pData);
pSourceDataInfo->code = code;
qDebug("%s fetch rsp received, index:%d, error:%s", pSourceDataInfo->taskId, index, tstrerror(code));
}
pSourceDataInfo->status = EX_SOURCE_DATA_READY;
tsem_post(&pExchangeInfo->ready);
taosReleaseRef(exchangeObjRefPool, pWrapper->exchangeId);
return TSDB_CODE_SUCCESS;
}
int32_t doSendFetchDataRequest(SExchangeInfo* pExchangeInfo, SExecTaskInfo* pTaskInfo, int32_t sourceIndex) {
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, sourceIndex);
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, sourceIndex);
ASSERT(pDataInfo->status == EX_SOURCE_DATA_NOT_READY);
SFetchRspHandleWrapper* pWrapper = taosMemoryCalloc(1, sizeof(SFetchRspHandleWrapper));
pWrapper->exchangeId = pExchangeInfo->self;
pWrapper->sourceIndex = sourceIndex;
if (pSource->localExec) {
SDataBuf pBuf = {0};
int32_t code =
(*pTaskInfo->localFetch.fp)(pTaskInfo->localFetch.handle, pSource->schedId, pTaskInfo->id.queryId,
pSource->taskId, 0, pSource->execId, &pBuf.pData, pTaskInfo->localFetch.explainRes);
loadRemoteDataCallback(pWrapper, &pBuf, code);
taosMemoryFree(pWrapper);
} else {
SResFetchReq* pMsg = taosMemoryCalloc(1, sizeof(SResFetchReq));
if (NULL == pMsg) {
pTaskInfo->code = TSDB_CODE_QRY_OUT_OF_MEMORY;
taosMemoryFree(pWrapper);
return pTaskInfo->code;
}
qDebug("%s build fetch msg and send to vgId:%d, ep:%s, taskId:0x%" PRIx64 ", execId:%d, %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->addr.epSet.eps[0].fqdn, pSource->taskId,
pSource->execId, sourceIndex, totalSources);
pMsg->header.vgId = htonl(pSource->addr.nodeId);
pMsg->sId = htobe64(pSource->schedId);
pMsg->taskId = htobe64(pSource->taskId);
pMsg->queryId = htobe64(pTaskInfo->id.queryId);
pMsg->execId = htonl(pSource->execId);
// send the fetch remote task result reques
SMsgSendInfo* pMsgSendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo));
if (NULL == pMsgSendInfo) {
taosMemoryFreeClear(pMsg);
taosMemoryFree(pWrapper);
qError("%s prepare message %d failed", GET_TASKID(pTaskInfo), (int32_t)sizeof(SMsgSendInfo));
pTaskInfo->code = TSDB_CODE_QRY_OUT_OF_MEMORY;
return pTaskInfo->code;
}
pMsgSendInfo->param = pWrapper;
pMsgSendInfo->paramFreeFp = taosMemoryFree;
pMsgSendInfo->msgInfo.pData = pMsg;
pMsgSendInfo->msgInfo.len = sizeof(SResFetchReq);
pMsgSendInfo->msgType = pSource->fetchMsgType;
pMsgSendInfo->fp = loadRemoteDataCallback;
int64_t transporterId = 0;
int32_t code =
asyncSendMsgToServer(pExchangeInfo->pTransporter, &pSource->addr.epSet, &transporterId, pMsgSendInfo);
}
return TSDB_CODE_SUCCESS;
}
void updateLoadRemoteInfo(SLoadRemoteDataInfo* pInfo, int32_t numOfRows, int32_t dataLen, int64_t startTs,
SOperatorInfo* pOperator) {
pInfo->totalRows += numOfRows;
pInfo->totalSize += dataLen;
pInfo->totalElapsed += (taosGetTimestampUs() - startTs);
pOperator->resultInfo.totalRows += numOfRows;
}
int32_t extractDataBlockFromFetchRsp(SSDataBlock* pRes, char* pData, SArray* pColList, char** pNextStart) {
if (pColList == NULL) { // data from other sources
blockDataCleanup(pRes);
*pNextStart = (char*)blockDecode(pRes, pData);
} else { // extract data according to pColList
char* pStart = pData;
int32_t numOfCols = htonl(*(int32_t*)pStart);
pStart += sizeof(int32_t);
// todo refactor:extract method
SSysTableSchema* pSchema = (SSysTableSchema*)pStart;
for (int32_t i = 0; i < numOfCols; ++i) {
SSysTableSchema* p = (SSysTableSchema*)pStart;
p->colId = htons(p->colId);
p->bytes = htonl(p->bytes);
pStart += sizeof(SSysTableSchema);
}
SSDataBlock* pBlock = createDataBlock();
for (int32_t i = 0; i < numOfCols; ++i) {
SColumnInfoData idata = createColumnInfoData(pSchema[i].type, pSchema[i].bytes, pSchema[i].colId);
blockDataAppendColInfo(pBlock, &idata);
}
blockDecode(pBlock, pStart);
blockDataEnsureCapacity(pRes, pBlock->info.rows);
// data from mnode
pRes->info.rows = pBlock->info.rows;
relocateColumnData(pRes, pColList, pBlock->pDataBlock, false);
blockDataDestroy(pBlock);
}
// todo move this to time window aggregator, since the primary timestamp may not be known by exchange operator.
blockDataUpdateTsWindow(pRes, 0);
return TSDB_CODE_SUCCESS;
}
void* setAllSourcesCompleted(SOperatorInfo* pOperator, int64_t startTs) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
int64_t el = taosGetTimestampUs() - startTs;
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
pLoadInfo->totalElapsed += el;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
qDebug("%s all %" PRIzu " sources are exhausted, total rows: %" PRIu64 " bytes:%" PRIu64 ", elapsed:%.2f ms",
GET_TASKID(pTaskInfo), totalSources, pLoadInfo->totalRows, pLoadInfo->totalSize,
pLoadInfo->totalElapsed / 1000.0);
setOperatorCompleted(pOperator);
return NULL;
}
int32_t getCompletedSources(const SArray* pArray) {
size_t total = taosArrayGetSize(pArray);
int32_t completed = 0;
for (int32_t k = 0; k < total; ++k) {
SSourceDataInfo* p = taosArrayGet(pArray, k);
if (p->status == EX_SOURCE_DATA_EXHAUSTED) {
completed += 1;
}
}
return completed;
}
int32_t prepareConcurrentlyLoad(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
int64_t startTs = taosGetTimestampUs();
// Asynchronously send all fetch requests to all sources.
for (int32_t i = 0; i < totalSources; ++i) {
int32_t code = doSendFetchDataRequest(pExchangeInfo, pTaskInfo, i);
if (code != TSDB_CODE_SUCCESS) {
pTaskInfo->code = code;
return code;
}
}
int64_t endTs = taosGetTimestampUs();
qDebug("%s send all fetch requests to %" PRIzu " sources completed, elapsed:%.2fms", GET_TASKID(pTaskInfo),
totalSources, (endTs - startTs) / 1000.0);
pOperator->status = OP_RES_TO_RETURN;
pOperator->cost.openCost = taosGetTimestampUs() - startTs;
tsem_wait(&pExchangeInfo->ready);
tsem_post(&pExchangeInfo->ready);
return TSDB_CODE_SUCCESS;
}
int32_t seqLoadRemoteData(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
int64_t startTs = taosGetTimestampUs();
while (1) {
if (pExchangeInfo->current >= totalSources) {
setAllSourcesCompleted(pOperator, startTs);
return TSDB_CODE_SUCCESS;
}
doSendFetchDataRequest(pExchangeInfo, pTaskInfo, pExchangeInfo->current);
tsem_wait(&pExchangeInfo->ready);
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, pExchangeInfo->current);
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, pExchangeInfo->current);
if (pDataInfo->code != TSDB_CODE_SUCCESS) {
qError("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d error happens, code:%s", GET_TASKID(pTaskInfo),
pSource->addr.nodeId, pSource->taskId, pSource->execId, tstrerror(pDataInfo->code));
pOperator->pTaskInfo->code = pDataInfo->code;
return pOperator->pTaskInfo->code;
}
SRetrieveTableRsp* pRsp = pDataInfo->pRsp;
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pRsp->numOfRows == 0) {
qDebug("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d %d of total completed, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 " try next",
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pExchangeInfo->current + 1,
pDataInfo->totalRows, pLoadInfo->totalRows);
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
pExchangeInfo->current += 1;
taosMemoryFreeClear(pDataInfo->pRsp);
continue;
}
SRetrieveTableRsp* pRetrieveRsp = pDataInfo->pRsp;
char* pStart = pRetrieveRsp->data;
int32_t code = extractDataBlockFromFetchRsp(NULL, pStart, NULL, &pStart);
if (pRsp->completed == 1) {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 ", totalBytes:%" PRIu64 " try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRetrieveRsp->numOfRows,
pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize, pExchangeInfo->current + 1,
totalSources);
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
pExchangeInfo->current += 1;
} else {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, totalRows:%" PRIu64
", totalBytes:%" PRIu64,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRetrieveRsp->numOfRows,
pLoadInfo->totalRows, pLoadInfo->totalSize);
}
updateLoadRemoteInfo(pLoadInfo, pRetrieveRsp->numOfRows, pRetrieveRsp->compLen, startTs, pOperator);
pDataInfo->totalRows += pRetrieveRsp->numOfRows;
taosMemoryFreeClear(pDataInfo->pRsp);
return TSDB_CODE_SUCCESS;
}
}
int32_t prepareLoadRemoteData(SOperatorInfo* pOperator) {
if (OPTR_IS_OPENED(pOperator)) {
return TSDB_CODE_SUCCESS;
}
int64_t st = taosGetTimestampUs();
SExchangeInfo* pExchangeInfo = pOperator->info;
if (!pExchangeInfo->seqLoadData) {
int32_t code = prepareConcurrentlyLoad(pOperator);
if (code != TSDB_CODE_SUCCESS) {
return code;
}
pExchangeInfo->openedTs = taosGetTimestampUs();
}
OPTR_SET_OPENED(pOperator);
pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
return TSDB_CODE_SUCCESS;
}

21
source/libs/executor/src/executor.c
View File

@ -1106,3 +1106,24 @@ int32_t qStreamPrepareScan(qTaskInfo_t tinfo, STqOffsetVal* pOffset, int8_t subT
}
return 0;
}
void qProcessRspMsg(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
SMsgSendInfo* pSendInfo = (SMsgSendInfo*)pMsg->info.ahandle;
assert(pMsg->info.ahandle != NULL);
SDataBuf buf = {.len = pMsg->contLen, .pData = NULL};
if (pMsg->contLen > 0) {
buf.pData = taosMemoryCalloc(1, pMsg->contLen);
if (buf.pData == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
pMsg->code = TSDB_CODE_OUT_OF_MEMORY;
} else {
memcpy(buf.pData, pMsg->pCont, pMsg->contLen);
}
}
pSendInfo->fp(pSendInfo->param, &buf, pMsg->code);
rpcFreeCont(pMsg->pCont);
destroySendMsgInfo(pSendInfo);
}

663
source/libs/executor/src/executorimpl.c
View File

@ -32,7 +32,6 @@
#include "index.h"
#include "query.h"
#include "tcompare.h"
#include "tcompression.h"
#include "thash.h"
#include "ttypes.h"
#include "vnode.h"
@ -91,11 +90,11 @@ static void destroySortOperatorInfo(void* param);
static void destroyAggOperatorInfo(void* param);
static void destroyIntervalOperatorInfo(void* param);
static void destroyExchangeOperatorInfo(void* param);
static void destroyOperatorInfo(SOperatorInfo* pOperator);
void doSetOperatorCompleted(SOperatorInfo* pOperator) {
void setOperatorCompleted(SOperatorInfo* pOperator) {
pOperator->status = OP_EXEC_DONE;
ASSERT(pOperator->pTaskInfo != NULL);
@ -103,14 +102,24 @@ void doSetOperatorCompleted(SOperatorInfo* pOperator) {
setTaskStatus(pOperator->pTaskInfo, TASK_COMPLETED);
}
void setOperatorInfo(SOperatorInfo* pOperator, const char* name, int32_t type, bool blocking, int32_t status,
void* pInfo, SExecTaskInfo* pTaskInfo) {
pOperator->name = (char*)name;
pOperator->operatorType = type;
pOperator->blocking = blocking;
pOperator->status = status;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
}
int32_t operatorDummyOpenFn(SOperatorInfo* pOperator) {
OPTR_SET_OPENED(pOperator);
pOperator->cost.openCost = 0;
return TSDB_CODE_SUCCESS;
}
SOperatorFpSet createOperatorFpSet(__optr_open_fn_t openFn, __optr_fn_t nextFn, __optr_fn_t streamFn,
__optr_fn_t cleanup, __optr_close_fn_t closeFn, __optr_explain_fn_t explain) {
SOperatorFpSet createOperatorFpSet(__optr_open_fn_t openFn, __optr_fn_t nextFn, __optr_fn_t cleanup,
__optr_close_fn_t closeFn, __optr_explain_fn_t explain) {
SOperatorFpSet fpSet = {
._openFn = openFn,
.getNextFn = nextFn,
@ -1652,596 +1661,6 @@ int32_t appendDownstream(SOperatorInfo* p, SOperatorInfo** pDownstream, int32_t
return TSDB_CODE_SUCCESS;
}
typedef struct SFetchRspHandleWrapper {
uint32_t exchangeId;
int32_t sourceIndex;
} SFetchRspHandleWrapper;
int32_t loadRemoteDataCallback(void* param, SDataBuf* pMsg, int32_t code) {
SFetchRspHandleWrapper* pWrapper = (SFetchRspHandleWrapper*)param;
SExchangeInfo* pExchangeInfo = taosAcquireRef(exchangeObjRefPool, pWrapper->exchangeId);
if (pExchangeInfo == NULL) {
qWarn("failed to acquire exchange operator, since it may have been released");
taosMemoryFree(pMsg->pData);
return TSDB_CODE_SUCCESS;
}
int32_t index = pWrapper->sourceIndex;
SSourceDataInfo* pSourceDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, index);
if (code == TSDB_CODE_SUCCESS) {
pSourceDataInfo->pRsp = pMsg->pData;
SRetrieveTableRsp* pRsp = pSourceDataInfo->pRsp;
pRsp->numOfRows = htonl(pRsp->numOfRows);
pRsp->compLen = htonl(pRsp->compLen);
pRsp->numOfCols = htonl(pRsp->numOfCols);
pRsp->useconds = htobe64(pRsp->useconds);
pRsp->numOfBlocks = htonl(pRsp->numOfBlocks);
ASSERT(pRsp != NULL);
qDebug("%s fetch rsp received, index:%d, blocks:%d, rows:%d", pSourceDataInfo->taskId, index, pRsp->numOfBlocks,
pRsp->numOfRows);
} else {
taosMemoryFree(pMsg->pData);
pSourceDataInfo->code = code;
qDebug("%s fetch rsp received, index:%d, error:%s", pSourceDataInfo->taskId, index, tstrerror(code));
}
pSourceDataInfo->status = EX_SOURCE_DATA_READY;
tsem_post(&pExchangeInfo->ready);
taosReleaseRef(exchangeObjRefPool, pWrapper->exchangeId);
return TSDB_CODE_SUCCESS;
}
void qProcessRspMsg(void* parent, SRpcMsg* pMsg, SEpSet* pEpSet) {
SMsgSendInfo* pSendInfo = (SMsgSendInfo*)pMsg->info.ahandle;
assert(pMsg->info.ahandle != NULL);
SDataBuf buf = {.len = pMsg->contLen, .pData = NULL};
if (pMsg->contLen > 0) {
buf.pData = taosMemoryCalloc(1, pMsg->contLen);
if (buf.pData == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
pMsg->code = TSDB_CODE_OUT_OF_MEMORY;
} else {
memcpy(buf.pData, pMsg->pCont, pMsg->contLen);
}
}
pSendInfo->fp(pSendInfo->param, &buf, pMsg->code);
rpcFreeCont(pMsg->pCont);
destroySendMsgInfo(pSendInfo);
}
static int32_t doSendFetchDataRequest(SExchangeInfo* pExchangeInfo, SExecTaskInfo* pTaskInfo, int32_t sourceIndex) {
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, sourceIndex);
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, sourceIndex);
ASSERT(pDataInfo->status == EX_SOURCE_DATA_NOT_READY);
SFetchRspHandleWrapper* pWrapper = taosMemoryCalloc(1, sizeof(SFetchRspHandleWrapper));
pWrapper->exchangeId = pExchangeInfo->self;
pWrapper->sourceIndex = sourceIndex;
if (pSource->localExec) {
SDataBuf pBuf = {0};
int32_t code =
(*pTaskInfo->localFetch.fp)(pTaskInfo->localFetch.handle, pSource->schedId, pTaskInfo->id.queryId,
pSource->taskId, 0, pSource->execId, &pBuf.pData, pTaskInfo->localFetch.explainRes);
loadRemoteDataCallback(pWrapper, &pBuf, code);
taosMemoryFree(pWrapper);
} else {
SResFetchReq* pMsg = taosMemoryCalloc(1, sizeof(SResFetchReq));
if (NULL == pMsg) {
pTaskInfo->code = TSDB_CODE_QRY_OUT_OF_MEMORY;
taosMemoryFree(pWrapper);
return pTaskInfo->code;
}
qDebug("%s build fetch msg and send to vgId:%d, ep:%s, taskId:0x%" PRIx64 ", execId:%d, %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->addr.epSet.eps[0].fqdn, pSource->taskId,
pSource->execId, sourceIndex, totalSources);
pMsg->header.vgId = htonl(pSource->addr.nodeId);
pMsg->sId = htobe64(pSource->schedId);
pMsg->taskId = htobe64(pSource->taskId);
pMsg->queryId = htobe64(pTaskInfo->id.queryId);
pMsg->execId = htonl(pSource->execId);
// send the fetch remote task result reques
SMsgSendInfo* pMsgSendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo));
if (NULL == pMsgSendInfo) {
taosMemoryFreeClear(pMsg);
taosMemoryFree(pWrapper);
qError("%s prepare message %d failed", GET_TASKID(pTaskInfo), (int32_t)sizeof(SMsgSendInfo));
pTaskInfo->code = TSDB_CODE_QRY_OUT_OF_MEMORY;
return pTaskInfo->code;
}
pMsgSendInfo->param = pWrapper;
pMsgSendInfo->paramFreeFp = taosMemoryFree;
pMsgSendInfo->msgInfo.pData = pMsg;
pMsgSendInfo->msgInfo.len = sizeof(SResFetchReq);
pMsgSendInfo->msgType = pSource->fetchMsgType;
pMsgSendInfo->fp = loadRemoteDataCallback;
int64_t transporterId = 0;
int32_t code =
asyncSendMsgToServer(pExchangeInfo->pTransporter, &pSource->addr.epSet, &transporterId, pMsgSendInfo);
}
return TSDB_CODE_SUCCESS;
}
void updateLoadRemoteInfo(SLoadRemoteDataInfo* pInfo, int32_t numOfRows, int32_t dataLen, int64_t startTs,
SOperatorInfo* pOperator) {
pInfo->totalRows += numOfRows;
pInfo->totalSize += dataLen;
pInfo->totalElapsed += (taosGetTimestampUs() - startTs);
pOperator->resultInfo.totalRows += numOfRows;
}
int32_t extractDataBlockFromFetchRsp(SSDataBlock* pRes, char* pData, SArray* pColList, char** pNextStart) {
if (pColList == NULL) { // data from other sources
blockDataCleanup(pRes);
*pNextStart = (char*)blockDecode(pRes, pData);
} else { // extract data according to pColList
char* pStart = pData;
int32_t numOfCols = htonl(*(int32_t*)pStart);
pStart += sizeof(int32_t);
// todo refactor:extract method
SSysTableSchema* pSchema = (SSysTableSchema*)pStart;
for (int32_t i = 0; i < numOfCols; ++i) {
SSysTableSchema* p = (SSysTableSchema*)pStart;
p->colId = htons(p->colId);
p->bytes = htonl(p->bytes);
pStart += sizeof(SSysTableSchema);
}
SSDataBlock* pBlock = createDataBlock();
for (int32_t i = 0; i < numOfCols; ++i) {
SColumnInfoData idata = createColumnInfoData(pSchema[i].type, pSchema[i].bytes, pSchema[i].colId);
blockDataAppendColInfo(pBlock, &idata);
}
blockDecode(pBlock, pStart);
blockDataEnsureCapacity(pRes, pBlock->info.rows);
// data from mnode
pRes->info.rows = pBlock->info.rows;
relocateColumnData(pRes, pColList, pBlock->pDataBlock, false);
blockDataDestroy(pBlock);
}
// todo move this to time window aggregator, since the primary timestamp may not be known by exchange operator.
blockDataUpdateTsWindow(pRes, 0);
return TSDB_CODE_SUCCESS;
}
static void* setAllSourcesCompleted(SOperatorInfo* pOperator, int64_t startTs) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
int64_t el = taosGetTimestampUs() - startTs;
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
pLoadInfo->totalElapsed += el;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
qDebug("%s all %" PRIzu " sources are exhausted, total rows: %" PRIu64 " bytes:%" PRIu64 ", elapsed:%.2f ms",
GET_TASKID(pTaskInfo), totalSources, pLoadInfo->totalRows, pLoadInfo->totalSize,
pLoadInfo->totalElapsed / 1000.0);
doSetOperatorCompleted(pOperator);
return NULL;
}
static int32_t getCompletedSources(const SArray* pArray) {
size_t total = taosArrayGetSize(pArray);
int32_t completed = 0;
for (int32_t k = 0; k < total; ++k) {
SSourceDataInfo* p = taosArrayGet(pArray, k);
if (p->status == EX_SOURCE_DATA_EXHAUSTED) {
completed += 1;
}
}
return completed;
}
static void concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SExchangeInfo* pExchangeInfo,
SExecTaskInfo* pTaskInfo) {
int32_t code = 0;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSourceDataInfo);
int32_t completed = getCompletedSources(pExchangeInfo->pSourceDataInfo);
if (completed == totalSources) {
setAllSourcesCompleted(pOperator, pExchangeInfo->openedTs);
return;
}
while (1) {
tsem_wait(&pExchangeInfo->ready);
for (int32_t i = 0; i < totalSources; ++i) {
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, i);
if (pDataInfo->status == EX_SOURCE_DATA_EXHAUSTED) {
continue;
}
if (pDataInfo->status != EX_SOURCE_DATA_READY) {
continue;
}
if (pDataInfo->code != TSDB_CODE_SUCCESS) {
code = pDataInfo->code;
goto _error;
}
SRetrieveTableRsp* pRsp = pDataInfo->pRsp;
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, i);
// todo
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pRsp->numOfRows == 0) {
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
qDebug("%s vgId:%d, taskId:0x%" PRIx64 " execId:%d index:%d completed, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 ", try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pDataInfo->totalRows,
pExchangeInfo->loadInfo.totalRows, i + 1, totalSources);
taosMemoryFreeClear(pDataInfo->pRsp);
break;
}
SRetrieveTableRsp* pRetrieveRsp = pDataInfo->pRsp;
int32_t index = 0;
char* pStart = pRetrieveRsp->data;
while (index++ < pRetrieveRsp->numOfBlocks) {
SSDataBlock* pb = createOneDataBlock(pExchangeInfo->pDummyBlock, false);
code = extractDataBlockFromFetchRsp(pb, pStart, NULL, &pStart);
if (code != 0) {
taosMemoryFreeClear(pDataInfo->pRsp);
goto _error;
}
taosArrayPush(pExchangeInfo->pResultBlockList, &pb);
}
updateLoadRemoteInfo(pLoadInfo, pRetrieveRsp->numOfRows, pRetrieveRsp->compLen, pExchangeInfo->openedTs, pOperator);
if (pRsp->completed == 1) {
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64
" execId:%d index:%d completed, blocks:%d, numOfRows:%d, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64
", total:%.2f Kb, try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pRsp->numOfBlocks,
pRsp->numOfRows, pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize / 1024.0,
i + 1, totalSources);
} else {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64
" execId:%d blocks:%d, numOfRows:%d, totalRows:%" PRIu64 ", total:%.2f Kb",
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRsp->numOfBlocks,
pRsp->numOfRows, pLoadInfo->totalRows, pLoadInfo->totalSize / 1024.0);
}
taosMemoryFreeClear(pDataInfo->pRsp);
if (pDataInfo->status != EX_SOURCE_DATA_EXHAUSTED) {
pDataInfo->status = EX_SOURCE_DATA_NOT_READY;
code = doSendFetchDataRequest(pExchangeInfo, pTaskInfo, i);
if (code != TSDB_CODE_SUCCESS) {
taosMemoryFreeClear(pDataInfo->pRsp);
goto _error;
}
}
return;
} // end loop
int32_t complete1 = getCompletedSources(pExchangeInfo->pSourceDataInfo);
if (complete1 == totalSources) {
qDebug("all sources are completed, %s", GET_TASKID(pTaskInfo));
return;
}
}
_error:
pTaskInfo->code = code;
}
static int32_t prepareConcurrentlyLoad(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
int64_t startTs = taosGetTimestampUs();
// Asynchronously send all fetch requests to all sources.
for (int32_t i = 0; i < totalSources; ++i) {
int32_t code = doSendFetchDataRequest(pExchangeInfo, pTaskInfo, i);
if (code != TSDB_CODE_SUCCESS) {
pTaskInfo->code = code;
return code;
}
}
int64_t endTs = taosGetTimestampUs();
qDebug("%s send all fetch requests to %" PRIzu " sources completed, elapsed:%.2fms", GET_TASKID(pTaskInfo),
totalSources, (endTs - startTs) / 1000.0);
pOperator->status = OP_RES_TO_RETURN;
pOperator->cost.openCost = taosGetTimestampUs() - startTs;
tsem_wait(&pExchangeInfo->ready);
tsem_post(&pExchangeInfo->ready);
return TSDB_CODE_SUCCESS;
}
static int32_t seqLoadRemoteData(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
int64_t startTs = taosGetTimestampUs();
while (1) {
if (pExchangeInfo->current >= totalSources) {
setAllSourcesCompleted(pOperator, startTs);
return TSDB_CODE_SUCCESS;
}
doSendFetchDataRequest(pExchangeInfo, pTaskInfo, pExchangeInfo->current);
tsem_wait(&pExchangeInfo->ready);
SSourceDataInfo* pDataInfo = taosArrayGet(pExchangeInfo->pSourceDataInfo, pExchangeInfo->current);
SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, pExchangeInfo->current);
if (pDataInfo->code != TSDB_CODE_SUCCESS) {
qError("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d error happens, code:%s", GET_TASKID(pTaskInfo),
pSource->addr.nodeId, pSource->taskId, pSource->execId, tstrerror(pDataInfo->code));
pOperator->pTaskInfo->code = pDataInfo->code;
return pOperator->pTaskInfo->code;
}
SRetrieveTableRsp* pRsp = pDataInfo->pRsp;
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pRsp->numOfRows == 0) {
qDebug("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d %d of total completed, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 " try next",
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pExchangeInfo->current + 1,
pDataInfo->totalRows, pLoadInfo->totalRows);
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
pExchangeInfo->current += 1;
taosMemoryFreeClear(pDataInfo->pRsp);
continue;
}
SRetrieveTableRsp* pRetrieveRsp = pDataInfo->pRsp;
char* pStart = pRetrieveRsp->data;
int32_t code = extractDataBlockFromFetchRsp(NULL, pStart, NULL, &pStart);
if (pRsp->completed == 1) {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, rowsOfSource:%" PRIu64
", totalRows:%" PRIu64 ", totalBytes:%" PRIu64 " try next %d/%" PRIzu,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRetrieveRsp->numOfRows,
pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize, pExchangeInfo->current + 1,
totalSources);
pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED;
pExchangeInfo->current += 1;
} else {
qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, totalRows:%" PRIu64
", totalBytes:%" PRIu64,
GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRetrieveRsp->numOfRows,
pLoadInfo->totalRows, pLoadInfo->totalSize);
}
updateLoadRemoteInfo(pLoadInfo, pRetrieveRsp->numOfRows, pRetrieveRsp->compLen, startTs, pOperator);
pDataInfo->totalRows += pRetrieveRsp->numOfRows;
taosMemoryFreeClear(pDataInfo->pRsp);
return TSDB_CODE_SUCCESS;
}
}
static int32_t prepareLoadRemoteData(SOperatorInfo* pOperator) {
if (OPTR_IS_OPENED(pOperator)) {
return TSDB_CODE_SUCCESS;
}
int64_t st = taosGetTimestampUs();
SExchangeInfo* pExchangeInfo = pOperator->info;
if (!pExchangeInfo->seqLoadData) {
int32_t code = prepareConcurrentlyLoad(pOperator);
if (code != TSDB_CODE_SUCCESS) {
return code;
}
pExchangeInfo->openedTs = taosGetTimestampUs();
}
OPTR_SET_OPENED(pOperator);
pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
return TSDB_CODE_SUCCESS;
}
static void freeBlock(void* pParam) {
SSDataBlock* pBlock = *(SSDataBlock**)pParam;
blockDataDestroy(pBlock);
}
static SSDataBlock* doLoadRemoteDataImpl(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
pTaskInfo->code = pOperator->fpSet._openFn(pOperator);
if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
return NULL;
}
size_t totalSources = taosArrayGetSize(pExchangeInfo->pSources);
SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo;
if (pOperator->status == OP_EXEC_DONE) {
qDebug("%s all %" PRIzu " source(s) are exhausted, total rows:%" PRIu64 " bytes:%" PRIu64 ", elapsed:%.2f ms",
GET_TASKID(pTaskInfo), totalSources, pLoadInfo->totalRows, pLoadInfo->totalSize,
pLoadInfo->totalElapsed / 1000.0);
return NULL;
}
size_t size = taosArrayGetSize(pExchangeInfo->pResultBlockList);
if (size == 0 || pExchangeInfo->rspBlockIndex >= size) {
pExchangeInfo->rspBlockIndex = 0;
taosArrayClearEx(pExchangeInfo->pResultBlockList, freeBlock);
if (pExchangeInfo->seqLoadData) {
seqLoadRemoteData(pOperator);
} else {
concurrentlyLoadRemoteDataImpl(pOperator, pExchangeInfo, pTaskInfo);
}
if (taosArrayGetSize(pExchangeInfo->pResultBlockList) == 0) {
return NULL;
}
}
// we have buffered retrieved datablock, return it directly
return taosArrayGetP(pExchangeInfo->pResultBlockList, pExchangeInfo->rspBlockIndex++);
}
static SSDataBlock* doLoadRemoteData(SOperatorInfo* pOperator) {
SExchangeInfo* pExchangeInfo = pOperator->info;
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
if (pOperator->status == OP_EXEC_DONE) {
return NULL;
}
while (1) {
SSDataBlock* pBlock = doLoadRemoteDataImpl(pOperator);
if (pBlock == NULL) {
return NULL;
}
SLimitInfo* pLimitInfo = &pExchangeInfo->limitInfo;
if (hasLimitOffsetInfo(pLimitInfo)) {
int32_t status = handleLimitOffset(pOperator, pLimitInfo, pBlock, false);
if (status == PROJECT_RETRIEVE_CONTINUE) {
continue;
} else if (status == PROJECT_RETRIEVE_DONE) {
size_t rows = pBlock->info.rows;
pExchangeInfo->limitInfo.numOfOutputRows += rows;
if (rows == 0) {
doSetOperatorCompleted(pOperator);
return NULL;
} else {
return pBlock;
}
}
} else {
return pBlock;
}
}
}
static int32_t initDataSource(int32_t numOfSources, SExchangeInfo* pInfo, const char* id) {
pInfo->pSourceDataInfo = taosArrayInit(numOfSources, sizeof(SSourceDataInfo));
if (pInfo->pSourceDataInfo == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
for (int32_t i = 0; i < numOfSources; ++i) {
SSourceDataInfo dataInfo = {0};
dataInfo.status = EX_SOURCE_DATA_NOT_READY;
dataInfo.taskId = id;
dataInfo.index = i;
SSourceDataInfo* pDs = taosArrayPush(pInfo->pSourceDataInfo, &dataInfo);
if (pDs == NULL) {
taosArrayDestroy(pInfo->pSourceDataInfo);
return TSDB_CODE_OUT_OF_MEMORY;
}
}
return TSDB_CODE_SUCCESS;
}
static int32_t initExchangeOperator(SExchangePhysiNode* pExNode, SExchangeInfo* pInfo, const char* id) {
size_t numOfSources = LIST_LENGTH(pExNode->pSrcEndPoints);
if (numOfSources == 0) {
qError("%s invalid number: %d of sources in exchange operator", id, (int32_t)numOfSources);
return TSDB_CODE_INVALID_PARA;
}
pInfo->pSources = taosArrayInit(numOfSources, sizeof(SDownstreamSourceNode));
if (pInfo->pSources == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
for (int32_t i = 0; i < numOfSources; ++i) {
SDownstreamSourceNode* pNode = (SDownstreamSourceNode*)nodesListGetNode((SNodeList*)pExNode->pSrcEndPoints, i);
taosArrayPush(pInfo->pSources, pNode);
}
initLimitInfo(pExNode->node.pLimit, pExNode->node.pSlimit, &pInfo->limitInfo);
pInfo->self = taosAddRef(exchangeObjRefPool, pInfo);
return initDataSource(numOfSources, pInfo, id);
}
SOperatorInfo* createExchangeOperatorInfo(void* pTransporter, SExchangePhysiNode* pExNode, SExecTaskInfo* pTaskInfo) {
SExchangeInfo* pInfo = taosMemoryCalloc(1, sizeof(SExchangeInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
if (pInfo == NULL || pOperator == NULL) {
goto _error;
}
int32_t code = initExchangeOperator(pExNode, pInfo, GET_TASKID(pTaskInfo));
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
tsem_init(&pInfo->ready, 0, 0);
pInfo->pDummyBlock = createResDataBlock(pExNode->node.pOutputDataBlockDesc);
pInfo->pResultBlockList = taosArrayInit(1, POINTER_BYTES);
pInfo->seqLoadData = false;
pInfo->pTransporter = pTransporter;
pOperator->name = "ExchangeOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_EXCHANGE;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->exprSupp.numOfExprs = taosArrayGetSize(pInfo->pDummyBlock->pDataBlock);
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet =
createOperatorFpSet(prepareLoadRemoteData, doLoadRemoteData, NULL, NULL, destroyExchangeOperatorInfo, NULL);
return pOperator;
_error:
if (pInfo != NULL) {
doDestroyExchangeOperatorInfo(pInfo);
}
taosMemoryFreeClear(pOperator);
pTaskInfo->code = code;
return NULL;
}
static int32_t doInitAggInfoSup(SAggSupporter* pAggSup, SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t keyBufSize,
const char* pKey);
@ -2386,7 +1805,7 @@ static SSDataBlock* getAggregateResult(SOperatorInfo* pOperator) {
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
pTaskInfo->code = pOperator->fpSet._openFn(pOperator);
if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -2396,7 +1815,7 @@ static SSDataBlock* getAggregateResult(SOperatorInfo* pOperator) {
doFilter(pInfo->pRes, pOperator->exprSupp.pFilterInfo, NULL);
if (!hasRemainResults(&pAggInfo->groupResInfo)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -2645,7 +2064,7 @@ static SSDataBlock* doFillImpl(SOperatorInfo* pOperator) {
SSDataBlock* pBlock = pDownstream->fpSet.getNextFn(pDownstream);
if (pBlock == NULL) {
if (pInfo->totalInputRows == 0) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -2722,7 +2141,7 @@ static SSDataBlock* doFill(SOperatorInfo* pOperator) {
while (true) {
fillResult = doFillImpl(pOperator);
if (fillResult == NULL) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -2952,15 +2371,10 @@ SOperatorInfo* createAggregateOperatorInfo(SOperatorInfo* downstream, SAggPhysiN
pInfo->binfo.mergeResultBlock = pAggNode->mergeDataBlock;
pInfo->groupId = UINT64_MAX;
pOperator->name = "TableAggregate";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_HASH_AGG;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "TableAggregate", QUERY_NODE_PHYSICAL_PLAN_HASH_AGG, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(doOpenAggregateOptr, getAggregateResult, NULL, NULL, destroyAggOperatorInfo, NULL);
createOperatorFpSet(doOpenAggregateOptr, getAggregateResult, NULL, destroyAggOperatorInfo, NULL);
if (downstream->operatorType == QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN) {
STableScanInfo* pTableScanInfo = downstream->info;
@ -3024,33 +2438,6 @@ void destroyFillOperatorInfo(void* param) {
taosMemoryFreeClear(param);
}
void destroyExchangeOperatorInfo(void* param) {
SExchangeInfo* pExInfo = (SExchangeInfo*)param;
taosRemoveRef(exchangeObjRefPool, pExInfo->self);
}
void freeSourceDataInfo(void* p) {
SSourceDataInfo* pInfo = (SSourceDataInfo*)p;
taosMemoryFreeClear(pInfo->pRsp);
}
void doDestroyExchangeOperatorInfo(void* param) {
SExchangeInfo* pExInfo = (SExchangeInfo*)param;
taosArrayDestroy(pExInfo->pSources);
taosArrayDestroyEx(pExInfo->pSourceDataInfo, freeSourceDataInfo);
if (pExInfo->pResultBlockList != NULL) {
taosArrayDestroyEx(pExInfo->pResultBlockList, freeBlock);
pExInfo->pResultBlockList = NULL;
}
blockDataDestroy(pExInfo->pDummyBlock);
tsem_destroy(&pExInfo->ready);
taosMemoryFreeClear(param);
}
static int32_t initFillInfo(SFillOperatorInfo* pInfo, SExprInfo* pExpr, int32_t numOfCols, SExprInfo* pNotFillExpr,
int32_t numOfNotFillCols, SNodeListNode* pValNode, STimeWindow win, int32_t capacity,
const char* id, SInterval* pInterval, int32_t fillType, int32_t order) {
@ -3182,15 +2569,9 @@ SOperatorInfo* createFillOperatorInfo(SOperatorInfo* downstream, SFillPhysiNode*
goto _error;
}
pOperator->name = "FillOperator";
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_FILL;
setOperatorInfo(pOperator, "FillOperator", QUERY_NODE_PHYSICAL_PLAN_FILL, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = pInfo->numOfExpr;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doFill, NULL, NULL, destroyFillOperatorInfo, NULL);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doFill, NULL, destroyFillOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
return pOperator;

38
source/libs/executor/src/groupoperator.c
View File

@ -316,7 +316,7 @@ static SSDataBlock* buildGroupResultDataBlock(SOperatorInfo* pOperator) {
doFilter(pRes, pOperator->exprSupp.pFilterInfo, NULL);
if (!hasRemainResults(&pInfo->groupResInfo)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -438,15 +438,10 @@ SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SAggPhysiNode*
}
initResultRowInfo(&pInfo->binfo.resultRowInfo);
pOperator->name = "GroupbyAggOperator";
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "GroupbyAggOperator", 0, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, hashGroupbyAggregate, NULL, NULL, destroyGroupOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, hashGroupbyAggregate, NULL, destroyGroupOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
@ -654,7 +649,7 @@ static SSDataBlock* buildPartitionResult(SOperatorInfo* pOperator) {
// try next group data
++pInfo->groupIndex;
if (pInfo->groupIndex >= taosArrayGetSize(pInfo->sortedGroupArray)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
clearPartitionOperator(pInfo);
return NULL;
}
@ -821,17 +816,12 @@ SOperatorInfo* createPartitionOperatorInfo(SOperatorInfo* downstream, SPartition
goto _error;
}
pOperator->name = "PartitionOperator";
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PARTITION;
setOperatorInfo(pOperator, "PartitionOperator", QUERY_NODE_PHYSICAL_PLAN_PARTITION, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = numOfCols;
pOperator->exprSupp.pExprInfo = pExprInfo;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, hashPartition, NULL, NULL, destroyPartitionOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, hashPartition, NULL, destroyPartitionOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
return pOperator;
@ -909,7 +899,10 @@ static SSDataBlock* buildStreamPartitionResult(SOperatorInfo* pOperator) {
void* pData = colDataGetVarData(pCol, 0);
// TODO check tbname validity
if (pData != (void*)-1) {
memcpy(pDest->info.parTbName, varDataVal(pData), varDataLen(pData));
memset(pDest->info.parTbName, 0, TSDB_TABLE_NAME_LEN);
int32_t len = TMIN(varDataLen(pData), TSDB_TABLE_NAME_LEN);
memcpy(pDest->info.parTbName, varDataVal(pData), len);
/*pDest->info.parTbName[len + 1] = 0;*/
} else {
pDest->info.parTbName[0] = 0;
}
@ -965,7 +958,7 @@ static SSDataBlock* doStreamHashPartition(SOperatorInfo* pOperator) {
pInfo->pInputDataBlock = NULL;
SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
if (pBlock == NULL) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
printDataBlock(pBlock, "stream partitionby recv");
@ -1106,15 +1099,10 @@ SOperatorInfo* createStreamPartitionOperatorInfo(SOperatorInfo* downstream, SStr
int32_t numOfCols = 0;
SExprInfo* pExprInfo = createExprInfo(pPartNode->part.pTargets, NULL, &numOfCols);
pOperator->name = "StreamPartitionOperator";
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_PARTITION;
setOperatorInfo(pOperator, "StreamPartitionOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_PARTITION, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = numOfCols;
pOperator->exprSupp.pExprInfo = pExprInfo;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamHashPartition, NULL, NULL,
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamHashPartition, NULL,
destroyStreamPartitionOperatorInfo, NULL);
initParDownStream(downstream, &pInfo->partitionSup, &pInfo->scalarSup);

15
source/libs/executor/src/joinoperator.c
View File

@ -73,14 +73,10 @@ SOperatorInfo* createMergeJoinOperatorInfo(SOperatorInfo** pDownstream, int32_t
initResultSizeInfo(&pOperator->resultInfo, 4096);
pInfo->pRes = pResBlock;
pOperator->name = "MergeJoinOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_JOIN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
setOperatorInfo(pOperator, "MergeJoinOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE_JOIN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.pExprInfo = pExprInfo;
pOperator->exprSupp.numOfExprs = numOfCols;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
extractTimeCondition(pInfo, pDownstream, numOfDownstream, pJoinNode);
@ -121,8 +117,7 @@ SOperatorInfo* createMergeJoinOperatorInfo(SOperatorInfo** pDownstream, int32_t
pInfo->inputOrder = TSDB_ORDER_DESC;
}
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doMergeJoin, NULL, NULL, destroyMergeJoinOperator, NULL);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doMergeJoin, NULL, destroyMergeJoinOperator, NULL);
code = appendDownstream(pOperator, pDownstream, numOfDownstream);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
@ -372,13 +367,13 @@ static void doMergeJoinImpl(struct SOperatorInfo* pOperator, SSDataBlock* pRes)
if (leftTs == rightTs) {
mergeJoinJoinDownstreamTsRanges(pOperator, leftTs, pRes, &nrows);
} else if (asc && leftTs < rightTs || !asc && leftTs > rightTs) {
} else if ((asc && leftTs < rightTs) || (!asc && leftTs > rightTs)) {
pJoinInfo->leftPos += 1;
if (pJoinInfo->leftPos >= pJoinInfo->pLeft->info.rows) {
continue;
}
} else if (asc && leftTs > rightTs || !asc && leftTs < rightTs) {
} else if ((asc && leftTs > rightTs) || (!asc && leftTs < rightTs)) {
pJoinInfo->rightPos += 1;
if (pJoinInfo->rightPos >= pJoinInfo->pRight->info.rows) {
continue;

28
source/libs/executor/src/projectoperator.c
View File

@ -98,13 +98,9 @@ SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SProjectPhys
}
pInfo->pPseudoColInfo = setRowTsColumnOutputInfo(pOperator->exprSupp.pCtx, numOfCols);
pOperator->name = "ProjectOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_PROJECT;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doProjectOperation, NULL, NULL,
setOperatorInfo(pOperator, "ProjectOperator", QUERY_NODE_PHYSICAL_PLAN_PROJECT, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doProjectOperation, NULL,
destroyProjectOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
@ -153,7 +149,7 @@ static int32_t setInfoForNewGroup(SSDataBlock* pBlock, SLimitInfo* pLimitInfo, S
if (pLimitInfo->currentGroupId != 0 && pLimitInfo->currentGroupId != pBlock->info.groupId) {
pLimitInfo->numOfOutputGroups += 1;
if ((pLimitInfo->slimit.limit > 0) && (pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return PROJECT_RETRIEVE_DONE;
}
@ -187,7 +183,7 @@ static int32_t doIngroupLimitOffset(SLimitInfo* pLimitInfo, uint64_t groupId, SS
// TODO: optimize it later when partition by + limit
if ((pLimitInfo->slimit.limit == -1 && pLimitInfo->currentGroupId == 0) ||
(pLimitInfo->slimit.limit > 0 && pLimitInfo->slimit.limit <= pLimitInfo->numOfOutputGroups)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
}
@ -252,7 +248,7 @@ SSDataBlock* doProjectOperation(SOperatorInfo* pOperator) {
}
qDebug("set op close, exec %d, status %d rows %d", pTaskInfo->execModel, pOperator->status,
pFinalRes->info.rows);
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
if (pTaskInfo->execModel == OPTR_EXEC_MODEL_QUEUE) {
@ -400,14 +396,8 @@ SOperatorInfo* createIndefinitOutputOperatorInfo(SOperatorInfo* downstream, SPhy
pInfo->binfo.pRes = pResBlock;
pInfo->pPseudoColInfo = setRowTsColumnOutputInfo(pSup->pCtx, numOfExpr);
pOperator->name = "IndefinitOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doApplyIndefinitFunction, NULL, NULL,
destroyIndefinitOperatorInfo, NULL);
setOperatorInfo(pOperator, "IndefinitOperator", QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doApplyIndefinitFunction, NULL, destroyIndefinitOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -499,7 +489,7 @@ SSDataBlock* doApplyIndefinitFunction(SOperatorInfo* pOperator) {
// The downstream exec may change the value of the newgroup, so use a local variable instead.
SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
if (pBlock == NULL) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -628,7 +618,7 @@ SSDataBlock* doGenerateSourceData(SOperatorInfo* pOperator) {
pOperator->resultInfo.totalRows += pRes->info.rows;
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
if (pOperator->cost.openCost == 0) {
pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
}

151
source/libs/executor/src/scanoperator.c
View File

@ -284,7 +284,7 @@ static int32_t doDynamicPruneDataBlock(SOperatorInfo* pOperator, SDataBlockInfo*
}
static bool doFilterByBlockSMA(SFilterInfo* pFilterInfo, SColumnDataAgg** pColsAgg, int32_t numOfCols,
int32_t numOfRows) {
int32_t numOfRows) {
if (pColsAgg == NULL || pFilterInfo == NULL) {
return true;
}
@ -345,7 +345,7 @@ static void doSetTagColumnData(STableScanInfo* pTableScanInfo, SSDataBlock* pBlo
// todo handle the slimit info
void applyLimitOffset(SLimitInfo* pLimitInfo, SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo, SOperatorInfo* pOperator) {
SLimit* pLimit = &pLimitInfo->limit;
SLimit* pLimit = &pLimitInfo->limit;
const char* id = GET_TASKID(pTaskInfo);
if (pLimit->offset > 0 && pLimitInfo->remainOffset > 0) {
@ -499,7 +499,7 @@ static void prepareForDescendingScan(STableScanInfo* pTableScanInfo, SqlFunction
typedef struct STableCachedVal {
const char* pName;
STag* pTags;
STag* pTags;
} STableCachedVal;
static void freeTableCachedVal(void* param) {
@ -513,13 +513,11 @@ static void freeTableCachedVal(void* param) {
taosMemoryFree(pVal);
}
//const void *key, size_t keyLen, void *value
static void freeCachedMetaItem(const void *key, size_t keyLen, void *value) {
freeTableCachedVal(value);
}
// const void *key, size_t keyLen, void *value
static void freeCachedMetaItem(const void* key, size_t keyLen, void* value) { freeTableCachedVal(value); }
int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int32_t numOfExpr,
SSDataBlock* pBlock, int32_t rows, const char* idStr, STableMetaCacheInfo* pCache) {
int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int32_t numOfExpr, SSDataBlock* pBlock,
int32_t rows, const char* idStr, STableMetaCacheInfo* pCache) {
// currently only the tbname pseudo column
if (numOfExpr <= 0) {
return TSDB_CODE_SUCCESS;
@ -531,11 +529,11 @@ int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int
int32_t backupRows = pBlock->info.rows;
pBlock->info.rows = rows;
bool freeReader = false;
bool freeReader = false;
STableCachedVal val = {0};
SMetaReader mr = {0};
LRUHandle* h = NULL;
LRUHandle* h = NULL;
// 1. check if it is existed in meta cache
if (pCache == NULL) {
@ -582,7 +580,8 @@ int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int
val = *pVal;
freeReader = true;
int32_t ret = taosLRUCacheInsert(pCache->pTableMetaEntryCache, &pBlock->info.uid, sizeof(uint64_t), pVal, sizeof(STableCachedVal), freeCachedMetaItem, NULL, TAOS_LRU_PRIORITY_LOW);
int32_t ret = taosLRUCacheInsert(pCache->pTableMetaEntryCache, &pBlock->info.uid, sizeof(uint64_t), pVal,
sizeof(STableCachedVal), freeCachedMetaItem, NULL, TAOS_LRU_PRIORITY_LOW);
if (ret != TAOS_LRU_STATUS_OK) {
qError("failed to put meta into lru cache, code:%d, %s", ret, idStr);
freeTableCachedVal(pVal);
@ -594,13 +593,13 @@ int32_t addTagPseudoColumnData(SReadHandle* pHandle, const SExprInfo* pExpr, int
taosLRUCacheRelease(pCache->pTableMetaEntryCache, h, false);
}
qDebug("retrieve table meta from cache:%"PRIu64", hit:%"PRIu64 " miss:%"PRIu64", %s", pCache->metaFetch, pCache->cacheHit,
(pCache->metaFetch - pCache->cacheHit), idStr);
qDebug("retrieve table meta from cache:%" PRIu64 ", hit:%" PRIu64 " miss:%" PRIu64 ", %s", pCache->metaFetch,
pCache->cacheHit, (pCache->metaFetch - pCache->cacheHit), idStr);
}
for (int32_t j = 0; j < numOfExpr; ++j) {
const SExprInfo* pExpr1 = &pExpr[j];
int32_t dstSlotId = pExpr1->base.resSchema.slotId;
int32_t dstSlotId = pExpr1->base.resSchema.slotId;
SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, dstSlotId);
colInfoDataCleanup(pColInfoData, pBlock->info.rows);
@ -652,7 +651,7 @@ void setTbNameColData(const SSDataBlock* pBlock, SColumnInfoData* pColInfoData,
fmGetScalarFuncExecFuncs(functionId, &fpSet);
size_t len = TSDB_TABLE_FNAME_LEN + VARSTR_HEADER_SIZE;
char buf[TSDB_TABLE_FNAME_LEN + VARSTR_HEADER_SIZE] = {0};
char buf[TSDB_TABLE_FNAME_LEN + VARSTR_HEADER_SIZE] = {0};
STR_TO_VARSTR(buf, name)
SColumnInfoData infoData = createColumnInfoData(TSDB_DATA_TYPE_VARCHAR, len, 1);
@ -820,7 +819,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
} else { // scan table group by group sequentially
if (pInfo->currentGroupId == -1) {
if ((++pInfo->currentGroupId) >= tableListGetOutputGroups(pTaskInfo->pTableInfoList)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -843,7 +842,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
}
if ((++pInfo->currentGroupId) >= tableListGetOutputGroups(pTaskInfo->pTableInfoList)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -865,7 +864,7 @@ static SSDataBlock* doTableScan(SOperatorInfo* pOperator) {
return result;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
}
@ -904,12 +903,12 @@ SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode,
goto _error;
}
SScanPhysiNode* pScanNode = &pTableScanNode->scan;
SScanPhysiNode* pScanNode = &pTableScanNode->scan;
SDataBlockDescNode* pDescNode = pScanNode->node.pOutputDataBlockDesc;
int32_t numOfCols = 0;
int32_t code = extractColMatchInfo(pScanNode->pScanCols, pDescNode, &numOfCols, COL_MATCH_FROM_COL_ID,
&pInfo->matchInfo);
int32_t code =
extractColMatchInfo(pScanNode->pScanCols, pDescNode, &numOfCols, COL_MATCH_FROM_COL_ID, &pInfo->matchInfo);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
@ -947,22 +946,17 @@ SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode,
pInfo->currentGroupId = -1;
pInfo->assignBlockUid = pTableScanNode->assignBlockUid;
pOperator->name = "TableScanOperator"; // for debug purpose
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "TableScanOperator", QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = numOfCols;
pOperator->pTaskInfo = pTaskInfo;
pInfo->metaCache.pTableMetaEntryCache = taosLRUCacheInit(1024*128, -1, .5);
pInfo->metaCache.pTableMetaEntryCache = taosLRUCacheInit(1024 * 128, -1, .5);
if (pInfo->metaCache.pTableMetaEntryCache == NULL) {
code = terrno;
goto _error;
}
taosLRUCacheSetStrictCapacity(pInfo->metaCache.pTableMetaEntryCache, false);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableScan, NULL, NULL, destroyTableScanOperatorInfo,
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableScan, NULL, destroyTableScanOperatorInfo,
getTableScannerExecInfo);
// for non-blocking operator, the open cost is always 0
@ -986,14 +980,8 @@ SOperatorInfo* createTableSeqScanOperatorInfo(void* pReadHandle, SExecTaskInfo*
pInfo->dataReader = pReadHandle;
// pInfo->prevGroupId = -1;
pOperator->name = "TableSeqScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TABLE_SEQ_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableScanImpl, NULL, NULL, NULL, NULL);
setOperatorInfo(pOperator, "TableSeqScanOperator", QUERY_NODE_PHYSICAL_PLAN_TABLE_SEQ_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableScanImpl, NULL, NULL, NULL);
return pOperator;
}
@ -1049,8 +1037,8 @@ static SSDataBlock* doBlockInfoScan(SOperatorInfo* pOperator) {
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
STableBlockDistInfo blockDistInfo = {.minRows = INT_MAX, .maxRows = INT_MIN};
int32_t code = doGetTableRowSize(pBlockScanInfo->readHandle.meta, pBlockScanInfo->uid, (int32_t*)&blockDistInfo.rowSize,
GET_TASKID(pTaskInfo));
int32_t code = doGetTableRowSize(pBlockScanInfo->readHandle.meta, pBlockScanInfo->uid,
(int32_t*)&blockDistInfo.rowSize, GET_TASKID(pTaskInfo));
if (code != TSDB_CODE_SUCCESS) {
T_LONG_JMP(pTaskInfo->env, code);
}
@ -1148,15 +1136,8 @@ SOperatorInfo* createDataBlockInfoScanOperator(SReadHandle* readHandle, SBlockDi
goto _error;
}
pOperator->name = "DataBlockDistScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_BLOCK_DIST_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doBlockInfoScan, NULL, NULL, destroyBlockDistScanOperatorInfo, NULL);
setOperatorInfo(pOperator, "DataBlockDistScanOperator", QUERY_NODE_PHYSICAL_PLAN_BLOCK_DIST_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doBlockInfoScan, NULL, destroyBlockDistScanOperatorInfo, NULL);
return pOperator;
_error:
@ -1599,8 +1580,10 @@ static void calBlockTbName(SExprSupp* pTbNameCalSup, SSDataBlock* pBlock) {
void* pData = colDataGetData(pCol, 0);
// TODO check tbname validation
if (pData != (void*)-1 && pData != NULL) {
memcpy(pBlock->info.parTbName, varDataVal(pData), TMIN(varDataLen(pData), TSDB_TABLE_NAME_LEN));
pBlock->info.parTbName[TSDB_TABLE_NAME_LEN - 1] = 0;
memset(pBlock->info.parTbName, 0, TSDB_TABLE_NAME_LEN);
int32_t len = TMIN(varDataLen(pData), TSDB_TABLE_NAME_LEN);
memcpy(pBlock->info.parTbName, varDataVal(pData), len);
/*pBlock->info.parTbName[len + 1] = 0;*/
} else {
pBlock->info.parTbName[0] = 0;
}
@ -2368,11 +2351,9 @@ SOperatorInfo* createRawScanOperatorInfo(SReadHandle* pHandle, SExecTaskInfo* pT
pInfo->vnode = pHandle->vnode;
pInfo->sContext = pHandle->sContext;
pOperator->name = "RawScanOperator";
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "RawScanOperator", QUERY_NODE_PHYSICAL_PLAN_TABLE_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(NULL, doRawScan, NULL, NULL, destroyRawScanOperatorInfo, NULL);
pOperator->fpSet = createOperatorFpSet(NULL, doRawScan, NULL, destroyRawScanOperatorInfo, NULL);
return pOperator;
_end:
@ -2556,16 +2537,11 @@ SOperatorInfo* createStreamScanOperatorInfo(SReadHandle* pHandle, STableScanPhys
pInfo->assignBlockUid = pTableScanNode->assignBlockUid;
pInfo->partitionSup.needCalc = false;
pOperator->name = "StreamScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "StreamScanOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = taosArrayGetSize(pInfo->pRes->pDataBlock);
pOperator->pTaskInfo = pTaskInfo;
__optr_fn_t nextFn = pTaskInfo->execModel == OPTR_EXEC_MODEL_STREAM ? doStreamScan : doQueueScan;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, nextFn, NULL, NULL, destroyStreamScanOperatorInfo, NULL);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, nextFn, NULL, destroyStreamScanOperatorInfo, NULL);
return pOperator;
@ -2900,7 +2876,7 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
}
blockDataDestroy(dataBlock);
pInfo->loadInfo.totalRows += pInfo->pRes->info.rows;
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return (pInfo->pRes->info.rows == 0) ? NULL : pInfo->pRes;
}
@ -2953,7 +2929,7 @@ static SSDataBlock* sysTableScanUserTags(SOperatorInfo* pOperator) {
if (ret != 0) {
metaCloseTbCursor(pInfo->pCur);
pInfo->pCur = NULL;
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
pInfo->loadInfo.totalRows += pInfo->pRes->info.rows;
@ -3743,7 +3719,7 @@ static SSDataBlock* sysTableBuildUserTablesByUids(SOperatorInfo* pOperator) {
}
if (i >= taosArrayGetSize(pIdx->uids)) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
} else {
pIdx->lastIdx = i + 1;
}
@ -3925,7 +3901,7 @@ static SSDataBlock* sysTableBuildUserTables(SOperatorInfo* pOperator) {
if (ret != 0) {
metaCloseTbCursor(pInfo->pCur);
pInfo->pCur = NULL;
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
pInfo->loadInfo.totalRows += pInfo->pRes->info.rows;
@ -3947,7 +3923,7 @@ static SSDataBlock* sysTableScanUserTables(SOperatorInfo* pOperator) {
doFilterResult(pInfo->pRes, pOperator->exprSupp.pFilterInfo);
pInfo->loadInfo.totalRows += pInfo->pRes->info.rows;
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return (pInfo->pRes->info.rows == 0) ? NULL : pInfo->pRes;
} else {
if (pInfo->showRewrite == false) {
@ -4199,15 +4175,9 @@ SOperatorInfo* createSysTableScanOperatorInfo(void* readHandle, SSystemTableScan
pInfo->readHandle = *(SReadHandle*)readHandle;
}
pOperator->name = "SysTableScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_SYSTABLE_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "SysTableScanOperator", QUERY_NODE_PHYSICAL_PLAN_SYSTABLE_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = taosArrayGetSize(pInfo->pRes->pDataBlock);
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doSysTableScan, NULL, NULL, destroySysScanOperator, NULL);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doSysTableScan, NULL, destroySysScanOperator, NULL);
return pOperator;
_error:
@ -4283,7 +4253,7 @@ static SSDataBlock* doTagScan(SOperatorInfo* pOperator) {
count += 1;
if (++pInfo->curPos >= size) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
}
@ -4319,7 +4289,7 @@ SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysi
int32_t numOfExprs = 0;
SExprInfo* pExprInfo = createExprInfo(pPhyNode->pScanPseudoCols, NULL, &numOfExprs);
int32_t code = initExprSupp(&pOperator->exprSupp, pExprInfo, numOfExprs);
int32_t code = initExprSupp(&pOperator->exprSupp, pExprInfo, numOfExprs);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
@ -4335,18 +4305,11 @@ SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysi
pInfo->readHandle = *pReadHandle;
pInfo->curPos = 0;
pOperator->name = "TagScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TAG_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "TagScanOperator", QUERY_NODE_PHYSICAL_PLAN_TAG_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
initResultSizeInfo(&pOperator->resultInfo, 4096);
blockDataEnsureCapacity(pInfo->pRes, pOperator->resultInfo.capacity);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTagScan, NULL, NULL, destroyTagScanOperatorInfo, NULL);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTagScan, NULL, destroyTagScanOperatorInfo, NULL);
return pOperator;
@ -4455,7 +4418,7 @@ static int32_t loadDataBlockFromOneTable(SOperatorInfo* pOperator, STableMergeSc
T_LONG_JMP(pTaskInfo->env, code);
}
if (pOperator->exprSupp.pFilterInfo!= NULL) {
if (pOperator->exprSupp.pFilterInfo != NULL) {
int64_t st = taosGetTimestampMs();
doFilter(pBlock, pOperator->exprSupp.pFilterInfo, &pTableScanInfo->matchInfo);
@ -4713,7 +4676,7 @@ SSDataBlock* doTableMergeScan(SOperatorInfo* pOperator) {
pInfo->hasGroupId = true;
if (tableListSize == 0) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
pInfo->tableStartIndex = 0;
@ -4732,7 +4695,7 @@ SSDataBlock* doTableMergeScan(SOperatorInfo* pOperator) {
} else {
stopGroupTableMergeScan(pOperator);
if (pInfo->tableEndIndex >= tableListSize - 1) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
pInfo->tableStartIndex = pInfo->tableEndIndex + 1;
@ -4831,7 +4794,6 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN
pInfo->sample.seed = taosGetTimestampSec();
pInfo->dataBlockLoadFlag = pTableScanNode->dataRequired;
code = filterInitFromNode((SNode*)pTableScanNode->scan.node.pConditions, &pOperator->exprSupp.pFilterInfo, 0);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
@ -4853,15 +4815,10 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN
int32_t rowSize = pInfo->pResBlock->info.rowSize;
pInfo->bufPageSize = getProperSortPageSize(rowSize);
pOperator->name = "TableMergeScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TABLE_MERGE_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "TableMergeScanOperator", QUERY_NODE_PHYSICAL_PLAN_TABLE_MERGE_SCAN, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.numOfExprs = numOfCols;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableMergeScan, NULL, NULL,
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableMergeScan, NULL,
destroyTableMergeScanOperatorInfo, getTableMergeScanExplainExecInfo);
pOperator->cost.openCost = 0;
return pOperator;

37
source/libs/executor/src/sortoperator.c
View File

@ -53,11 +53,7 @@ SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSortPhysiNode*
pInfo->pSortInfo = createSortInfo(pSortNode->pSortKeys);
initLimitInfo(pSortNode->node.pLimit, pSortNode->node.pSlimit, &pInfo->limitInfo);
pOperator->name = "SortOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_SORT;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "SortOperator", QUERY_NODE_PHYSICAL_PLAN_SORT, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->exprSupp.pExprInfo = pExprInfo;
pOperator->exprSupp.numOfExprs = numOfCols;
@ -67,7 +63,7 @@ SOperatorInfo* createSortOperatorInfo(SOperatorInfo* downstream, SSortPhysiNode*
// TODO dynamic set the available sort buffer
pOperator->fpSet =
createOperatorFpSet(doOpenSortOperator, doSort, NULL, NULL, destroySortOperatorInfo, getExplainExecInfo);
createOperatorFpSet(doOpenSortOperator, doSort, NULL, destroySortOperatorInfo, getExplainExecInfo);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -214,7 +210,7 @@ SSDataBlock* doSort(SOperatorInfo* pOperator) {
pBlock = getSortedBlockData(pInfo->pSortHandle, pInfo->binfo.pRes, pOperator->resultInfo.capacity,
pInfo->matchInfo.pList, pInfo);
if (pBlock == NULL) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -428,7 +424,7 @@ SSDataBlock* doGroupSort(SOperatorInfo* pOperator) {
pInfo->prefetchedSortInput = pOperator->pDownstream[0]->fpSet.getNextFn(pOperator->pDownstream[0]);
if (pInfo->prefetchedSortInput == NULL) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
pInfo->currGroupId = pInfo->prefetchedSortInput->info.groupId;
@ -453,7 +449,7 @@ SSDataBlock* doGroupSort(SOperatorInfo* pOperator) {
beginSortGroup(pOperator);
} else if (pInfo->childOpStatus == CHILD_OP_FINISHED) {
finishSortGroup(pOperator);
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
}
@ -509,15 +505,8 @@ SOperatorInfo* createGroupSortOperatorInfo(SOperatorInfo* downstream, SGroupSort
}
pInfo->pSortInfo = createSortInfo(pSortPhyNode->pSortKeys);
pOperator->name = "GroupSortOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doGroupSort, NULL, NULL, destroyGroupSortOperatorInfo,
setOperatorInfo(pOperator, "GroupSortOperator", QUERY_NODE_PHYSICAL_PLAN_GROUP_SORT, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doGroupSort, NULL, destroyGroupSortOperatorInfo,
getGroupSortExplainExecInfo);
code = appendDownstream(pOperator, &downstream, 1);
@ -705,7 +694,7 @@ SSDataBlock* doMultiwayMerge(SOperatorInfo* pOperator) {
if (pBlock != NULL) {
pOperator->resultInfo.totalRows += pBlock->info.rows;
} else {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
return pBlock;
@ -774,14 +763,8 @@ SOperatorInfo* createMultiwayMergeOperatorInfo(SOperatorInfo** downStreams, size
pInfo->bufPageSize = getProperSortPageSize(rowSize);
pInfo->sortBufSize = pInfo->bufPageSize * (numStreams + 1); // one additional is reserved for merged result.
pOperator->name = "MultiwayMerge";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->fpSet = createOperatorFpSet(doOpenMultiwayMergeOperator, doMultiwayMerge, NULL, NULL,
setOperatorInfo(pOperator, "MultiwayMergeOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(doOpenMultiwayMergeOperator, doMultiwayMerge, NULL,
destroyMultiwayMergeOperatorInfo, getMultiwayMergeExplainExecInfo);
code = appendDownstream(pOperator, downStreams, numStreams);

14
source/libs/executor/src/tfill.c
View File

@ -1443,7 +1443,7 @@ static SSDataBlock* doStreamFill(SOperatorInfo* pOperator) {
printDataBlock(pInfo->pRes, "stream fill");
return pInfo->pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
resetStreamFillInfo(pInfo);
return NULL;
}
@ -1512,7 +1512,7 @@ static SSDataBlock* doStreamFill(SOperatorInfo* pOperator) {
}
if (pInfo->pRes->info.rows == 0) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
resetStreamFillInfo(pInfo);
return NULL;
}
@ -1690,15 +1690,9 @@ SOperatorInfo* createStreamFillOperatorInfo(SOperatorInfo* downstream, SStreamFi
}
pInfo->srcRowIndex = 0;
pOperator->name = "StreamFillOperator";
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_FILL;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "StreamFillOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_FILL, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doStreamFill, NULL, NULL, destroyStreamFillOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, doStreamFill, NULL, destroyStreamFillOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {

149
source/libs/executor/src/timewindowoperator.c
View File

@ -1221,7 +1221,7 @@ static SSDataBlock* doStateWindowAgg(SOperatorInfo* pOperator) {
pTaskInfo->code = pOperator->fpSet._openFn(pOperator);
if (pTaskInfo->code != TSDB_CODE_SUCCESS) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -1232,7 +1232,7 @@ static SSDataBlock* doStateWindowAgg(SOperatorInfo* pOperator) {
bool hasRemain = hasRemainResults(&pInfo->groupResInfo);
if (!hasRemain) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -1269,7 +1269,7 @@ static SSDataBlock* doBuildIntervalResult(SOperatorInfo* pOperator) {
bool hasRemain = hasRemainResults(&pInfo->groupResInfo);
if (!hasRemain) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -1739,7 +1739,6 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SIntervalPh
ASSERT(as.calTrigger != STREAM_TRIGGER_MAX_DELAY);
pOperator->pTaskInfo = pTaskInfo;
pInfo->win = pTaskInfo->window;
pInfo->inputOrder = (pPhyNode->window.inputTsOrder == ORDER_ASC) ? TSDB_ORDER_ASC : TSDB_ORDER_DESC;
pInfo->resultTsOrder = (pPhyNode->window.outputTsOrder == ORDER_ASC) ? TSDB_ORDER_ASC : TSDB_ORDER_DESC;
@ -1777,15 +1776,10 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SIntervalPh
}
initResultRowInfo(&pInfo->binfo.resultRowInfo);
pOperator->name = "TimeIntervalAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "TimeIntervalAggOperator", QUERY_NODE_PHYSICAL_PLAN_HASH_INTERVAL, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(doOpenIntervalAgg, doBuildIntervalResult, NULL, NULL, destroyIntervalOperatorInfo, NULL);
createOperatorFpSet(doOpenIntervalAgg, doBuildIntervalResult, NULL, destroyIntervalOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -1890,7 +1884,7 @@ static SSDataBlock* doSessionWindowAgg(SOperatorInfo* pOperator) {
bool hasRemain = hasRemainResults(&pInfo->groupResInfo);
if (!hasRemain) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -1933,7 +1927,7 @@ static SSDataBlock* doSessionWindowAgg(SOperatorInfo* pOperator) {
bool hasRemain = hasRemainResults(&pInfo->groupResInfo);
if (!hasRemain) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -2281,7 +2275,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -2330,7 +2324,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
}
@ -2342,7 +2336,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
pSliceInfo->current =
taosTimeAdd(pSliceInfo->current, pInterval->interval, pInterval->intervalUnit, pInterval->precision);
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
}
@ -2365,7 +2359,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
}
@ -2386,7 +2380,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
} else {
@ -2448,7 +2442,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
}
@ -2463,7 +2457,7 @@ static SSDataBlock* doTimeslice(SOperatorInfo* pOperator) {
}
if (pSliceInfo->current > pSliceInfo->win.ekey) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
}
@ -2557,15 +2551,9 @@ SOperatorInfo* createTimeSliceOperatorInfo(SOperatorInfo* downstream, SPhysiNode
pScanInfo->cond.twindows = pInfo->win;
pScanInfo->cond.type = TIMEWINDOW_RANGE_EXTERNAL;
pOperator->name = "TimeSliceOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_INTERP_FUNC;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->pTaskInfo = pTaskInfo;
setOperatorInfo(pOperator, "TimeSliceOperator", QUERY_NODE_PHYSICAL_PLAN_INTERP_FUNC, false, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doTimeslice, NULL, NULL, destroyTimeSliceOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, doTimeslice, NULL, destroyTimeSliceOperatorInfo, NULL);
blockDataEnsureCapacity(pInfo->pRes, pOperator->resultInfo.capacity);
@ -2633,15 +2621,10 @@ SOperatorInfo* createStatewindowOperatorInfo(SOperatorInfo* downstream, SStateWi
initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pTaskInfo->window);
pInfo->tsSlotId = tsSlotId;
pOperator->name = "StateWindowOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_STATE;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->pTaskInfo = pTaskInfo;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "StateWindowOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE_STATE, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(openStateWindowAggOptr, doStateWindowAgg, NULL, NULL, destroyStateWindowOperatorInfo, NULL);
createOperatorFpSet(openStateWindowAggOptr, doStateWindowAgg, NULL, destroyStateWindowOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -2711,14 +2694,9 @@ SOperatorInfo* createSessionAggOperatorInfo(SOperatorInfo* downstream, SSessionW
goto _error;
}
pOperator->name = "SessionWindowAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_SESSION;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "SessionWindowAggOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE_SESSION, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doSessionWindowAgg, NULL, NULL, destroySWindowOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, doSessionWindowAgg, NULL, destroySWindowOperatorInfo, NULL);
pOperator->pTaskInfo = pTaskInfo;
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -3134,7 +3112,7 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
return pInfo->binfo.pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
if (!IS_FINAL_OP(pInfo)) {
clearFunctionContext(&pOperator->exprSupp);
// semi interval operator clear disk buffer
@ -3403,7 +3381,7 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
pOperator->info = pInfo;
pOperator->fpSet =
createOperatorFpSet(NULL, doStreamFinalIntervalAgg, NULL, NULL, destroyStreamFinalIntervalOperatorInfo, NULL);
createOperatorFpSet(NULL, doStreamFinalIntervalAgg, NULL, destroyStreamFinalIntervalOperatorInfo, NULL);
if (pPhyNode->type == QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL) {
initIntervalDownStream(downstream, pPhyNode->type, &pInfo->aggSup, &pInfo->interval, &pInfo->twAggSup);
}
@ -4027,7 +4005,7 @@ static SSDataBlock* doStreamSessionAgg(SOperatorInfo* pOperator) {
return pBInfo->pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -4133,7 +4111,7 @@ static SSDataBlock* doStreamSessionAgg(SOperatorInfo* pOperator) {
return pBInfo->pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -4200,13 +4178,11 @@ SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SPh
pInfo->pGroupIdTbNameMap =
taosHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), false, HASH_NO_LOCK);
pOperator->name = "StreamSessionWindowAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_SESSION;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamSessionAgg, NULL, NULL,
destroyStreamSessionAggOperatorInfo, NULL);
setOperatorInfo(pOperator, "StreamSessionWindowAggOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_SESSION, true,
OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doStreamSessionAgg, NULL, destroyStreamSessionAggOperatorInfo, NULL);
if (downstream) {
initDownStream(downstream, &pInfo->streamAggSup, pInfo->twAggSup.waterMark, pOperator->operatorType,
pInfo->primaryTsIndex);
@ -4257,7 +4233,7 @@ static SSDataBlock* doStreamSessionSemiAgg(SOperatorInfo* pOperator) {
clearFunctionContext(&pOperator->exprSupp);
// semi interval operator clear disk buffer
clearStreamSessionOperator(pInfo);
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
}
@ -4336,7 +4312,7 @@ static SSDataBlock* doStreamSessionSemiAgg(SOperatorInfo* pOperator) {
clearFunctionContext(&pOperator->exprSupp);
// semi interval operator clear disk buffer
clearStreamSessionOperator(pInfo);
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -4347,20 +4323,21 @@ SOperatorInfo* createStreamFinalSessionAggOperatorInfo(SOperatorInfo* downstream
if (pOperator == NULL) {
goto _error;
}
SStreamSessionAggOperatorInfo* pInfo = pOperator->info;
if (pPhyNode->type == QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_SESSION) {
pInfo->isFinal = true;
pOperator->name = "StreamSessionFinalAggOperator";
} else {
pInfo->isFinal = false;
pInfo->isFinal = (pPhyNode->type == QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_SESSION);
char* name = (pInfo->isFinal)? "StreamSessionFinalAggOperator":"StreamSessionSemiAggOperator";
if (pPhyNode->type != QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_SESSION) {
pInfo->pUpdateRes = createSpecialDataBlock(STREAM_CLEAR);
blockDataEnsureCapacity(pInfo->pUpdateRes, 128);
pOperator->name = "StreamSessionSemiAggOperator";
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamSessionSemiAgg, NULL, NULL,
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamSessionSemiAgg, NULL,
destroyStreamSessionAggOperatorInfo, NULL);
}
setOperatorInfo(pOperator, name, pPhyNode->type , false, OP_NOT_OPENED, pInfo, pTaskInfo);
pInfo->pGroupIdTbNameMap =
taosHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), false, HASH_NO_LOCK);
@ -4590,7 +4567,7 @@ static SSDataBlock* doStreamStateAgg(SOperatorInfo* pOperator) {
return pBInfo->pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -4656,7 +4633,7 @@ static SSDataBlock* doStreamStateAgg(SOperatorInfo* pOperator) {
printDataBlock(pBInfo->pRes, "single state");
return pBInfo->pRes;
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
return NULL;
}
@ -4721,14 +4698,9 @@ SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhys
pInfo->pGroupIdTbNameMap =
taosHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), false, HASH_NO_LOCK);
pOperator->name = "StreamStateAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_STATE;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->pTaskInfo = pTaskInfo;
pOperator->info = pInfo;
setOperatorInfo(pOperator, "StreamStateAggOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_STATE, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doStreamStateAgg, NULL, NULL, destroyStreamStateOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, doStreamStateAgg, NULL, destroyStreamStateOperatorInfo, NULL);
initDownStream(downstream, &pInfo->streamAggSup, pInfo->twAggSup.waterMark, pOperator->operatorType,
pInfo->primaryTsIndex);
code = appendDownstream(pOperator, &downstream, 1);
@ -4876,7 +4848,7 @@ static void doMergeAlignedIntervalAgg(SOperatorInfo* pOperator) {
cleanupAfterGroupResultGen(pMiaInfo, pRes);
}
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
break;
}
@ -5001,16 +4973,10 @@ SOperatorInfo* createMergeAlignedIntervalOperatorInfo(SOperatorInfo* downstream,
initResultRowInfo(&iaInfo->binfo.resultRowInfo);
blockDataEnsureCapacity(iaInfo->binfo.pRes, pOperator->resultInfo.capacity);
pOperator->name = "TimeMergeAlignedIntervalAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->pTaskInfo = pTaskInfo;
pOperator->info = miaInfo;
setOperatorInfo(pOperator, "TimeMergeAlignedIntervalAggOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE_ALIGNED_INTERVAL, false, OP_NOT_OPENED, miaInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, mergeAlignedIntervalAgg, NULL, NULL, destroyMAIOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, mergeAlignedIntervalAgg, NULL, destroyMAIOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -5254,7 +5220,7 @@ static SSDataBlock* doMergeIntervalAgg(SOperatorInfo* pOperator) {
}
if (pRes->info.rows == 0) {
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
}
size_t rows = pRes->info.rows;
@ -5313,16 +5279,9 @@ SOperatorInfo* createMergeIntervalOperatorInfo(SOperatorInfo* downstream, SMerge
}
initResultRowInfo(&pIntervalInfo->binfo.resultRowInfo);
pOperator->name = "TimeMergeIntervalAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->pTaskInfo = pTaskInfo;
pOperator->info = pMergeIntervalInfo;
setOperatorInfo(pOperator, "TimeMergeIntervalAggOperator", QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL, false, OP_NOT_OPENED, pMergeIntervalInfo, pTaskInfo);
pOperator->fpSet =
createOperatorFpSet(operatorDummyOpenFn, doMergeIntervalAgg, NULL, NULL, destroyMergeIntervalOperatorInfo, NULL);
createOperatorFpSet(operatorDummyOpenFn, doMergeIntervalAgg, NULL, destroyMergeIntervalOperatorInfo, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -5366,7 +5325,7 @@ static SSDataBlock* doStreamIntervalAgg(SOperatorInfo* pOperator) {
}
deleteIntervalDiscBuf(pInfo->pState, NULL, pInfo->twAggSup.maxTs - pInfo->twAggSup.deleteMark, &pInfo->interval,
&pInfo->delKey);
doSetOperatorCompleted(pOperator);
setOperatorCompleted(pOperator);
streamStateCommit(pTaskInfo->streamInfo.pState);
return NULL;
}
@ -5550,12 +5509,8 @@ SOperatorInfo* createStreamIntervalOperatorInfo(SOperatorInfo* downstream, SPhys
pInfo->pGroupIdTbNameMap =
taosHashInit(1024, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), false, HASH_NO_LOCK);
pOperator->name = "StreamIntervalOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL;
pOperator->blocking = true;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamIntervalAgg, NULL, NULL,
setOperatorInfo(pOperator, "StreamIntervalOperator", QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL, true, OP_NOT_OPENED, pInfo, pTaskInfo);
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doStreamIntervalAgg, NULL,
destroyStreamFinalIntervalOperatorInfo, NULL);
initIntervalDownStream(downstream, pPhyNode->type, &pInfo->aggSup, &pInfo->interval, &pInfo->twAggSup);

87
source/libs/function/src/builtinsimpl.c
View File

@ -3096,27 +3096,86 @@ int32_t lastFunction(SqlFunctionCtx* pCtx) {
}
#else
int64_t* pts = (int64_t*)pInput->pPTS->pData;
for (int32_t i = pInput->startRowIndex; i < pInput->numOfRows + pInput->startRowIndex; ++i) {
if (pInputCol->hasNull && colDataIsNull(pInputCol, pInput->totalRows, i, pColAgg)) {
continue;
}
numOfElems++;
#if 0
for (int32_t i = pInput->startRowIndex; i < pInput->numOfRows + pInput->startRowIndex; ++i) {
if (pInputCol->hasNull && colDataIsNull(pInputCol, pInput->totalRows, i, pColAgg)) {
continue;
}
char* data = colDataGetData(pInputCol, i);
TSKEY cts = pts[i];
if (pResInfo->numOfRes == 0 || pInfo->ts < cts) {
doSaveCurrentVal(pCtx, i, cts, type, data);
pResInfo->numOfRes = 1;
numOfElems++;
if (pResInfo->numOfRes == 0 || pInfo->ts < pts[i]) {
char* data = colDataGetData(pInputCol, i);
doSaveCurrentVal(pCtx, i, pts[i], type, data);
pResInfo->numOfRes = 1;
}
}
}
#else
if (!pInputCol->hasNull) {
numOfElems = 1;
int32_t round = pInput->numOfRows >> 2;
int32_t reminder = pInput->numOfRows & 0x03;
int32_t tick = 0;
for (int32_t i = pInput->startRowIndex; tick < round; i += 4, tick += 1) {
int64_t cts = pts[i];
int32_t chosen = i;
if (cts < pts[i + 1]) {
cts = pts[i + 1];
chosen = i + 1;
}
if (cts < pts[i + 2]) {
cts = pts[i + 2];
chosen = i + 2;
}
if (cts < pts[i + 3]) {
cts = pts[i + 3];
chosen = i + 3;
}
if (pResInfo->numOfRes == 0 || pInfo->ts < cts) {
char* data = colDataGetData(pInputCol, chosen);
doSaveCurrentVal(pCtx, i, cts, type, data);
pResInfo->numOfRes = 1;
}
}
for (int32_t i = pInput->startRowIndex + round * 4; i < pInput->startRowIndex + pInput->numOfRows; ++i) {
if (pResInfo->numOfRes == 0 || pInfo->ts < pts[i]) {
char* data = colDataGetData(pInputCol, i);
doSaveCurrentVal(pCtx, i, pts[i], type, data);
pResInfo->numOfRes = 1;
}
}
} else {
for (int32_t i = pInput->startRowIndex; i < pInput->startRowIndex + pInput->numOfRows; ++i) {
if (pInputCol->hasNull && colDataIsNull(pInputCol, pInput->totalRows, i, pColAgg)) {
continue;
}
numOfElems++;
if (pResInfo->numOfRes == 0 || pInfo->ts < pts[i]) {
char* data = colDataGetData(pInputCol, i);
doSaveCurrentVal(pCtx, i, pts[i], type, data);
pResInfo->numOfRes = 1;
}
}
}
#endif
#endif
// save selectivity value for column consisted of all null values
if (numOfElems == 0) {
firstlastSaveTupleData(pCtx->pSrcBlock, pInput->startRowIndex, pCtx, pInfo);
}
SET_VAL(pResInfo, numOfElems, 1);
// SET_VAL(pResInfo, numOfElems, 1);
return TSDB_CODE_SUCCESS;
}
@ -3266,8 +3325,8 @@ int32_t lastRowFunction(SqlFunctionCtx* pCtx) {
#if 0
int32_t blockDataOrder = (startKey <= endKey) ? TSDB_ORDER_ASC : TSDB_ORDER_DESC;
// the optimized version only function if all tuples in one block are monotonious increasing or descreasing.
// this is NOT always works if project operator exists in downstream.
// the optimized version only valid if all tuples in one block are monotonious increasing or descreasing.
// this assumption is NOT always works if project operator exists in downstream.
if (blockDataOrder == TSDB_ORDER_ASC) {
for (int32_t i = pInput->numOfRows + pInput->startRowIndex - 1; i >= pInput->startRowIndex; --i) {
char* data = colDataGetData(pInputCol, i);

2
source/libs/parser/inc/parUtil.h
View File

@ -108,7 +108,7 @@ int32_t getTableMetaFromCache(SParseMetaCache* pMetaCache, const SName* pName, S
int32_t getDbVgInfoFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, SArray** pVgInfo);
int32_t getTableVgroupFromCache(SParseMetaCache* pMetaCache, const SName* pName, SVgroupInfo* pVgroup);
int32_t getDbVgVersionFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, int32_t* pVersion, int64_t* pDbId,
int32_t* pTableNum);
int32_t* pTableNum, int64_t* pStateTs);
int32_t getDbCfgFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, SDbCfgInfo* pInfo);
int32_t getUserAuthFromCache(SParseMetaCache* pMetaCache, const char* pUser, const char* pDbFName, AUTH_TYPE type,
bool* pPass);

47
source/libs/parser/src/parTranslater.c
View File

@ -466,14 +466,14 @@ static int32_t getTableHashVgroup(STranslateContext* pCxt, const char* pDbName,
}
static int32_t getDBVgVersion(STranslateContext* pCxt, const char* pDbFName, int32_t* pVersion, int64_t* pDbId,
int32_t* pTableNum) {
int32_t* pTableNum, int64_t* pStateTs) {
SParseContext* pParCxt = pCxt->pParseCxt;
int32_t code = collectUseDatabaseImpl(pDbFName, pCxt->pDbs);
if (TSDB_CODE_SUCCESS == code) {
if (pParCxt->async) {
code = getDbVgVersionFromCache(pCxt->pMetaCache, pDbFName, pVersion, pDbId, pTableNum);
code = getDbVgVersionFromCache(pCxt->pMetaCache, pDbFName, pVersion, pDbId, pTableNum, pStateTs);
} else {
code = catalogGetDBVgVersion(pParCxt->pCatalog, pDbFName, pVersion, pDbId, pTableNum);
code = catalogGetDBVgVersion(pParCxt->pCatalog, pDbFName, pVersion, pDbId, pTableNum, pStateTs);
}
}
if (TSDB_CODE_SUCCESS != code) {
@ -1252,6 +1252,19 @@ static bool dataTypeEqual(const SDataType* l, const SDataType* r) {
return (l->type == r->type && l->bytes == r->bytes && l->precision == r->precision && l->scale == r->scale);
}
// 0 means equal, 1 means the left shall prevail, -1 means the right shall prevail
static int32_t dataTypeComp(const SDataType* l, const SDataType* r) {
if (l->type != r->type) {
return 1;
}
if (l->bytes != r->bytes) {
return l->bytes > r->bytes ? 1 : -1;
}
return (l->precision == r->precision && l->scale == r->scale) ? 0 : 1;
}
static EDealRes translateOperator(STranslateContext* pCxt, SOperatorNode* pOp) {
if (isMultiResFunc(pOp->pLeft)) {
return generateDealNodeErrMsg(pCxt, TSDB_CODE_PAR_WRONG_VALUE_TYPE, ((SExprNode*)(pOp->pLeft))->aliasName);
@ -1876,10 +1889,15 @@ static EDealRes translateCaseWhen(STranslateContext* pCxt, SCaseWhenNode* pCaseW
}
pWhenThen->pWhen = pIsTrue;
}
if (first) {
first = false;
if (first || dataTypeComp(&pCaseWhen->node.resType, &((SExprNode*)pNode)->resType) < 0) {
pCaseWhen->node.resType = ((SExprNode*)pNode)->resType;
} else if (!dataTypeEqual(&pCaseWhen->node.resType, &((SExprNode*)pNode)->resType)) {
}
first = false;
}
FOREACH(pNode, pCaseWhen->pWhenThenList) {
SWhenThenNode* pWhenThen = (SWhenThenNode*)pNode;
if (!dataTypeEqual(&pCaseWhen->node.resType, &((SExprNode*)pNode)->resType)) {
SNode* pCastFunc = NULL;
pCxt->errCode = createCastFunc(pCxt, pWhenThen->pThen, pCaseWhen->node.resType, &pCastFunc);
if (TSDB_CODE_SUCCESS != pCxt->errCode) {
@ -1889,6 +1907,7 @@ static EDealRes translateCaseWhen(STranslateContext* pCxt, SCaseWhenNode* pCaseW
pWhenThen->node.resType = pCaseWhen->node.resType;
}
}
if (NULL != pCaseWhen->pElse && !dataTypeEqual(&pCaseWhen->node.resType, &((SExprNode*)pCaseWhen->pElse)->resType)) {
SNode* pCastFunc = NULL;
pCxt->errCode = createCastFunc(pCxt, pCaseWhen->pElse, pCaseWhen->node.resType, &pCastFunc);
@ -3431,19 +3450,6 @@ static SNode* createSetOperProject(const char* pTableAlias, SNode* pNode) {
return (SNode*)pCol;
}
// 0 means equal, 1 means the left shall prevail, -1 means the right shall prevail
static int32_t dataTypeComp(const SDataType* l, const SDataType* r) {
if (l->type != r->type) {
return 1;
}
if (l->bytes != r->bytes) {
return l->bytes > r->bytes ? 1 : -1;
}
return (l->precision == r->precision && l->scale == r->scale) ? 0 : 1;
}
static int32_t translateSetOperProject(STranslateContext* pCxt, SSetOperator* pSetOperator) {
SNodeList* pLeftProjections = getProjectList(pSetOperator->pLeft);
SNodeList* pRightProjections = getProjectList(pSetOperator->pRight);
@ -4969,7 +4975,8 @@ static int32_t translateUseDatabase(STranslateContext* pCxt, SUseDatabaseStmt* p
SName name = {0};
tNameSetDbName(&name, pCxt->pParseCxt->acctId, pStmt->dbName, strlen(pStmt->dbName));
tNameExtractFullName(&name, usedbReq.db);
int32_t code = getDBVgVersion(pCxt, usedbReq.db, &usedbReq.vgVersion, &usedbReq.dbId, &usedbReq.numOfTable);
int32_t code =
getDBVgVersion(pCxt, usedbReq.db, &usedbReq.vgVersion, &usedbReq.dbId, &usedbReq.numOfTable, &usedbReq.stateTs);
if (TSDB_CODE_SUCCESS == code) {
code = buildCmdMsg(pCxt, TDMT_MND_USE_DB, (FSerializeFunc)tSerializeSUseDbReq, &usedbReq);
}

3
source/libs/parser/src/parUtil.c
View File

@ -876,13 +876,14 @@ int32_t reserveDbVgVersionInCache(int32_t acctId, const char* pDb, SParseMetaCac
}
int32_t getDbVgVersionFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, int32_t* pVersion, int64_t* pDbId,
int32_t* pTableNum) {
int32_t* pTableNum, int64_t* pStateTs) {
SDbInfo* pDbInfo = NULL;
int32_t code = getMetaDataFromHash(pDbFName, strlen(pDbFName), pMetaCache->pDbInfo, (void**)&pDbInfo);
if (TSDB_CODE_SUCCESS == code) {
*pVersion = pDbInfo->vgVer;
*pDbId = pDbInfo->dbId;
*pTableNum = pDbInfo->tbNum;
*pStateTs = pDbInfo->stateTs;
}
return code;
}

2
source/libs/parser/test/mockCatalog.cpp
View File

@ -249,7 +249,7 @@ int32_t __catalogGetTableDistVgInfo(SCatalog* pCtg, SRequestConnInfo* pConn, con
}
int32_t __catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* version, int64_t* dbId,
int32_t* tableNum) {
int32_t* tableNum, int64_t* stateTs) {
return 0;
}

6
source/libs/qcom/src/queryUtil.c
View File

@ -146,6 +146,12 @@ void destroySendMsgInfo(SMsgSendInfo* pMsgBody) {
}
taosMemoryFreeClear(pMsgBody);
}
void destroyAhandle(void *ahandle) {
SMsgSendInfo *pSendInfo = ahandle;
if (pSendInfo == NULL) return;
destroySendMsgInfo(pSendInfo);
}
int32_t asyncSendMsgToServerExt(void* pTransporter, SEpSet* epSet, int64_t* pTransporterId, SMsgSendInfo* pInfo,
bool persistHandle, void* rpcCtx) {

4
source/libs/qcom/src/querymsg.c
View File

@ -41,8 +41,9 @@ int32_t queryBuildUseDbOutput(SUseDbOutput *pOut, SUseDbRsp *usedbRsp) {
pOut->dbVgroup->hashMethod = usedbRsp->hashMethod;
pOut->dbVgroup->hashPrefix = usedbRsp->hashPrefix;
pOut->dbVgroup->hashSuffix = usedbRsp->hashSuffix;
pOut->dbVgroup->stateTs = usedbRsp->stateTs;
qDebug("Got %d vgroup for db %s", usedbRsp->vgNum, usedbRsp->db);
qDebug("Got %d vgroup for db %s, vgVersion:%d, stateTs:%" PRId64, usedbRsp->vgNum, usedbRsp->db, usedbRsp->vgVersion, usedbRsp->stateTs);
if (usedbRsp->vgNum <= 0) {
return TSDB_CODE_SUCCESS;
@ -103,6 +104,7 @@ int32_t queryBuildUseDbMsg(void *input, char **msg, int32_t msgSize, int32_t *ms
usedbReq.vgVersion = pInput->vgVersion;
usedbReq.dbId = pInput->dbId;
usedbReq.numOfTable = pInput->numOfTable;
usedbReq.stateTs = pInput->stateTs;
int32_t bufLen = tSerializeSUseDbReq(NULL, 0, &usedbReq);
void *pBuf = (*mallcFp)(bufLen);

3
source/libs/scheduler/src/schRemote.c
View File

@ -425,6 +425,7 @@ int32_t schHandleCallback(void *param, SDataBuf *pMsg, int32_t rspCode) {
_return:
taosMemoryFreeClear(pMsg->pData);
taosMemoryFreeClear(pMsg->pEpSet);
qDebug("end to handle rsp msg, type:%s, handle:%p, code:%s", TMSG_INFO(pMsg->msgType), pMsg->handle,
tstrerror(rspCode));
@ -438,6 +439,7 @@ int32_t schHandleDropCallback(void *param, SDataBuf *pMsg, int32_t code) {
code);
if (pMsg) {
taosMemoryFree(pMsg->pData);
taosMemoryFree(pMsg->pEpSet);
}
return TSDB_CODE_SUCCESS;
}
@ -492,6 +494,7 @@ _return:
tFreeSSchedulerHbRsp(&rsp);
taosMemoryFree(pMsg->pData);
taosMemoryFree(pMsg->pEpSet);
SCH_RET(code);
}

8
source/libs/scheduler/src/schTask.c
View File

@ -439,6 +439,8 @@ int32_t schHandleRedirect(SSchJob *pJob, SSchTask *pTask, SDataBuf *pData, int32
code = schDoTaskRedirect(pJob, pTask, pData, rspCode);
taosMemoryFree(pData->pData);
taosMemoryFree(pData->pEpSet);
pData->pData = NULL;
pData->pEpSet = NULL;
SCH_RET(code);
@ -446,6 +448,8 @@ _return:
taosMemoryFree(pData->pData);
taosMemoryFree(pData->pEpSet);
pData->pData = NULL;
pData->pEpSet = NULL;
SCH_RET(schProcessOnTaskFailure(pJob, pTask, code));
}
@ -942,7 +946,7 @@ int32_t schLaunchLocalTask(SSchJob *pJob, SSchTask *pTask) {
}
SCH_ERR_JRET(qWorkerProcessLocalQuery(schMgmt.queryMgmt, schMgmt.sId, pJob->queryId, pTask->taskId, pJob->refId,
pTask->execId, &qwMsg, explainRes));
pTask->execId, &qwMsg, explainRes));
if (SCH_IS_EXPLAIN_JOB(pJob)) {
SCH_ERR_RET(schHandleExplainRes(explainRes));
@ -1115,7 +1119,7 @@ int32_t schExecLocalFetch(SSchJob *pJob, SSchTask *pTask) {
}
SCH_ERR_JRET(qWorkerProcessLocalFetch(schMgmt.queryMgmt, schMgmt.sId, pJob->queryId, pTask->taskId, pJob->refId,
pTask->execId, &pRsp, explainRes));
pTask->execId, &pRsp, explainRes));
if (SCH_IS_EXPLAIN_JOB(pJob)) {
SCH_ERR_RET(schHandleExplainRes(explainRes));

6
source/libs/stream/src/streamDispatch.c
View File

@ -118,8 +118,7 @@ int32_t streamBroadcastToChildren(SStreamTask* pTask, const SSDataBlock* pBlock)
pRetrieve->ekey = htobe64(pBlock->info.window.ekey);
pRetrieve->version = htobe64(pBlock->info.version);
int32_t actualLen = 0;
blockEncode(pBlock, pRetrieve->data, &actualLen, numOfCols, false);
int32_t actualLen = blockEncode(pBlock, pRetrieve->data, numOfCols);
SStreamRetrieveReq req = {
.streamId = pTask->streamId,
@ -200,8 +199,7 @@ static int32_t streamAddBlockToDispatchMsg(const SSDataBlock* pBlock, SStreamDis
int32_t numOfCols = (int32_t)taosArrayGetSize(pBlock->pDataBlock);
pRetrieve->numOfCols = htonl(numOfCols);
int32_t actualLen = 0;
blockEncode(pBlock, pRetrieve->data, &actualLen, numOfCols, false);
int32_t actualLen = blockEncode(pBlock, pRetrieve->data, numOfCols);
actualLen += sizeof(SRetrieveTableRsp);
ASSERT(actualLen <= dataStrLen);
taosArrayPush(pReq->dataLen, &actualLen);

4
source/libs/sync/inc/syncAppendEntries.h
View File

@ -20,12 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
// TLA+ Spec
// HandleAppendEntriesRequest(i, j, m) ==

4
source/libs/sync/inc/syncAppendEntriesReply.h
View File

@ -20,12 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
// TLA+ Spec
// HandleAppendEntriesResponse(i, j, m) ==

4
source/libs/sync/inc/syncCommit.h
View File

@ -20,11 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "taosdef.h"
// \* Leader i advances its commitIndex.
// \* This is done as a separate step from handling AppendEntries responses,

4
source/libs/sync/inc/syncElection.h
View File

@ -20,11 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "taosdef.h"
// TLA+ Spec
// RequestVote(i, j) ==

4
source/libs/sync/inc/syncIndexMgr.h
View File

@ -20,11 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "taosdef.h"
// SIndexMgr -----------------------------
typedef struct SSyncIndexMgr {

5
source/libs/sync/inc/syncMessage.h
View File

@ -20,12 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "cJSON.h"
#include "syncInt.h"
#include "taosdef.h"
// ---------------------------------------------
cJSON* syncRpcMsg2Json(SRpcMsg* pRpcMsg);

5
source/libs/sync/inc/syncRaftCfg.h
View File

@ -20,12 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "cJSON.h"
#include "syncInt.h"
#include "taosdef.h"
#define CONFIG_FILE_LEN 2048

5
source/libs/sync/inc/syncRaftEntry.h
View File

@ -20,13 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
#include "tref.h"
#include "tskiplist.h"
typedef struct SSyncRaftEntry {

4
source/libs/sync/inc/syncRaftLog.h
View File

@ -20,12 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncRaftEntry.h"
#include "taosdef.h"
#include "wal.h"
typedef struct SSyncLogStoreData {

5
source/libs/sync/inc/syncRaftStore.h
View File

@ -20,12 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "cJSON.h"
#include "syncInt.h"
#include "taosdef.h"
#define RAFT_STORE_BLOCK_SIZE 512
#define RAFT_STORE_PATH_LEN (TSDB_FILENAME_LEN * 2)

4
source/libs/sync/inc/syncReplication.h
View File

@ -20,11 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "taosdef.h"
// TLA+ Spec
// AppendEntries(i, j) ==

4
source/libs/sync/inc/syncRequestVote.h
View File

@ -20,12 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
// TLA+ Spec
// HandleRequestVoteRequest(i, j, m) ==

4
source/libs/sync/inc/syncRequestVoteReply.h
View File

@ -20,12 +20,8 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
// TLA+ Spec
// HandleRequestVoteResponse(i, j, m) ==

4
source/libs/sync/inc/syncRespMgr.h
View File

@ -20,11 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "taosdef.h"
typedef struct SRespStub {
SRpcMsg rpcMsg;

19
source/libs/sync/inc/syncSnapshot.h
View File

@ -20,13 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "cJSON.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
#define SYNC_SNAPSHOT_SEQ_INVALID -2
#define SYNC_SNAPSHOT_SEQ_FORCE_CLOSE -3
@ -36,7 +30,6 @@ extern "C" {
#define SYNC_SNAPSHOT_RETRY_MS 5000
//---------------------------------------------------
typedef struct SSyncSnapshotSender {
bool start;
int32_t seq;
@ -65,12 +58,6 @@ int32_t snapshotSenderStop(SSyncSnapshotSender *pSender, bool finis
int32_t snapshotSend(SSyncSnapshotSender *pSender);
int32_t snapshotReSend(SSyncSnapshotSender *pSender);
cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender);
char *snapshotSender2Str(SSyncSnapshotSender *pSender);
int32_t syncNodeStartSnapshot(SSyncNode *pSyncNode, SRaftId *pDestId);
//---------------------------------------------------
typedef struct SSyncSnapshotReceiver {
bool start;
int32_t ack;
@ -83,7 +70,6 @@ typedef struct SSyncSnapshotReceiver {
// init when create
SSyncNode *pSyncNode;
} SSyncSnapshotReceiver;
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId fromId);
@ -93,13 +79,10 @@ int32_t snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver);
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverForceStop(SSyncSnapshotReceiver *pReceiver);
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);
char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver);
//---------------------------------------------------
// on message
int32_t syncNodeOnSnapshot(SSyncNode *ths, SyncSnapshotSend *pMsg);
int32_t syncNodeOnSnapshotReply(SSyncNode *ths, SyncSnapshotRsp *pMsg);
int32_t syncNodeStartSnapshot(SSyncNode *pSyncNode, SRaftId *pDestId);
#ifdef __cplusplus
}

5
source/libs/sync/inc/syncTimeout.h
View File

@ -20,12 +20,7 @@
extern "C" {
#endif
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
// TLA+ Spec
// Timeout(i) == /\ state[i] \in {Follower, Candidate}

1
source/libs/sync/inc/syncTools.h
View File

@ -728,7 +728,6 @@ int32_t syncNodeOnHeartbeat(SSyncNode* ths, SyncHeartbeat* pMsg);
int32_t syncNodeOnHeartbeatReply(SSyncNode* ths, SyncHeartbeatReply* pMsg);
int32_t syncNodeOnClientRequest(SSyncNode* ths, SRpcMsg* pMsg, SyncIndex* pRetIndex);
int32_t syncNodeOnTimer(SSyncNode* ths, SyncTimeout* pMsg);
int32_t syncNodeOnLocalCmd(SSyncNode* ths, SyncLocalCmd* pMsg);
// -----------------------------------------

8
source/libs/sync/inc/syncUtil.h
View File

@ -24,17 +24,15 @@ extern "C" {
uint64_t syncUtilAddr2U64(const char* host, uint16_t port);
void syncUtilU642Addr(uint64_t u64, char* host, int64_t len, uint16_t* port);
void syncUtilnodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet);
void syncUtilraftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet);
bool syncUtilnodeInfo2raftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId* raftId);
void syncUtilNodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet);
void syncUtilRaftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet);
bool syncUtilNodeInfo2RaftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId* raftId);
bool syncUtilSameId(const SRaftId* pId1, const SRaftId* pId2);
bool syncUtilEmptyId(const SRaftId* pId);
int32_t syncUtilElectRandomMS(int32_t min, int32_t max);
int32_t syncUtilQuorum(int32_t replicaNum);
cJSON* syncUtilNodeInfo2Json(const SNodeInfo* p);
cJSON* syncUtilRaftId2Json(const SRaftId* p);
char* syncUtilRaftId2Str(const SRaftId* p);
const char* syncStr(ESyncState state);
char* syncUtilPrintBin(char* ptr, uint32_t len);
char* syncUtilPrintBin2(char* ptr, uint32_t len);

4
source/libs/sync/inc/syncVoteMgr.h
View File

@ -39,8 +39,6 @@ void voteGrantedUpdate(SVotesGranted *pVotesGranted, SSyncNode *pSyncN
bool voteGrantedMajority(SVotesGranted *pVotesGranted);
void voteGrantedVote(SVotesGranted *pVotesGranted, SyncRequestVoteReply *pMsg);
void voteGrantedReset(SVotesGranted *pVotesGranted, SyncTerm term);
cJSON *voteGranted2Json(SVotesGranted *pVotesGranted);
char *voteGranted2Str(SVotesGranted *pVotesGranted);
typedef struct SVotesRespond {
SRaftId (*replicas)[TSDB_MAX_REPLICA];
@ -56,8 +54,6 @@ void votesRespondUpdate(SVotesRespond *pVotesRespond, SSyncNode *pSync
bool votesResponded(SVotesRespond *pVotesRespond, const SRaftId *pRaftId);
void votesRespondAdd(SVotesRespond *pVotesRespond, const SyncRequestVoteReply *pMsg);
void votesRespondReset(SVotesRespond *pVotesRespond, SyncTerm term);
cJSON *votesRespond2Json(SVotesRespond *pVotesRespond);
char *votesRespond2Str(SVotesRespond *pVotesRespond);
#ifdef __cplusplus
}

16
source/libs/sync/src/syncMain.c
View File

@ -757,7 +757,7 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
// init internal
pSyncNode->myNodeInfo = pSyncNode->pRaftCfg->cfg.nodeInfo[pSyncNode->pRaftCfg->cfg.myIndex];
if (!syncUtilnodeInfo2raftId(&pSyncNode->myNodeInfo, pSyncNode->vgId, &pSyncNode->myRaftId)) {
if (!syncUtilNodeInfo2RaftId(&pSyncNode->myNodeInfo, pSyncNode->vgId, &pSyncNode->myRaftId)) {
sError("vgId:%d, failed to determine my raft member id", pSyncNode->vgId);
goto _error;
}
@ -772,7 +772,7 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
}
}
for (int32_t i = 0; i < pSyncNode->peersNum; ++i) {
if (!syncUtilnodeInfo2raftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &pSyncNode->peersId[i])) {
if (!syncUtilNodeInfo2RaftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &pSyncNode->peersId[i])) {
sError("vgId:%d, failed to determine raft member id, peer:%d", pSyncNode->vgId, i);
goto _error;
}
@ -781,7 +781,7 @@ SSyncNode* syncNodeOpen(SSyncInfo* pSyncInfo) {
// init replicaNum, replicasId
pSyncNode->replicaNum = pSyncNode->pRaftCfg->cfg.replicaNum;
for (int32_t i = 0; i < pSyncNode->pRaftCfg->cfg.replicaNum; ++i) {
if (!syncUtilnodeInfo2raftId(&pSyncNode->pRaftCfg->cfg.nodeInfo[i], pSyncNode->vgId, &pSyncNode->replicasId[i])) {
if (!syncUtilNodeInfo2RaftId(&pSyncNode->pRaftCfg->cfg.nodeInfo[i], pSyncNode->vgId, &pSyncNode->replicasId[i])) {
sError("vgId:%d, failed to determine raft member id, replica:%d", pSyncNode->vgId, i);
goto _error;
}
@ -1213,7 +1213,7 @@ int32_t syncNodeRestartHeartbeatTimer(SSyncNode* pSyncNode) {
// utils --------------
int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilraftId2EpSet(destRaftId, &epSet);
syncUtilRaftId2EpSet(destRaftId, &epSet);
if (pSyncNode->syncSendMSg != NULL) {
// htonl
syncUtilMsgHtoN(pMsg->pCont);
@ -1230,7 +1230,7 @@ int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRp
int32_t syncNodeSendMsgByInfo(const SNodeInfo* nodeInfo, SSyncNode* pSyncNode, SRpcMsg* pMsg) {
SEpSet epSet;
syncUtilnodeInfo2EpSet(nodeInfo, &epSet);
syncUtilNodeInfo2EpSet(nodeInfo, &epSet);
if (pSyncNode->syncSendMSg != NULL) {
// htonl
syncUtilMsgHtoN(pMsg->pCont);
@ -1344,7 +1344,7 @@ void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncInde
// init internal
pSyncNode->myNodeInfo = pSyncNode->pRaftCfg->cfg.nodeInfo[pSyncNode->pRaftCfg->cfg.myIndex];
syncUtilnodeInfo2raftId(&pSyncNode->myNodeInfo, pSyncNode->vgId, &pSyncNode->myRaftId);
syncUtilNodeInfo2RaftId(&pSyncNode->myNodeInfo, pSyncNode->vgId, &pSyncNode->myRaftId);
// init peersNum, peers, peersId
pSyncNode->peersNum = pSyncNode->pRaftCfg->cfg.replicaNum - 1;
@ -1356,13 +1356,13 @@ void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncInde
}
}
for (int32_t i = 0; i < pSyncNode->peersNum; ++i) {
syncUtilnodeInfo2raftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &pSyncNode->peersId[i]);
syncUtilNodeInfo2RaftId(&pSyncNode->peersNodeInfo[i], pSyncNode->vgId, &pSyncNode->peersId[i]);
}
// init replicaNum, replicasId
pSyncNode->replicaNum = pSyncNode->pRaftCfg->cfg.replicaNum;
for (int32_t i = 0; i < pSyncNode->pRaftCfg->cfg.replicaNum; ++i) {
syncUtilnodeInfo2raftId(&pSyncNode->pRaftCfg->cfg.nodeInfo[i], pSyncNode->vgId, &pSyncNode->replicasId[i]);
syncUtilNodeInfo2RaftId(&pSyncNode->pRaftCfg->cfg.nodeInfo[i], pSyncNode->vgId, &pSyncNode->replicasId[i]);
}
// update quorum first

1
source/libs/sync/src/syncRaftEntry.c
View File

@ -16,6 +16,7 @@
#define _DEFAULT_SOURCE
#include "syncRaftEntry.h"
#include "syncUtil.h"
#include "tref.h"
SSyncRaftEntry* syncEntryBuild(int32_t dataLen) {
int32_t bytes = sizeof(SSyncRaftEntry) + dataLen;

6
source/libs/sync/src/syncRaftLog.c
View File

@ -294,6 +294,12 @@ static int32_t raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIn
return 0;
}
// need not truncate
SyncIndex walCommitVer = walGetCommittedVer(pWal);
if (fromIndex <= walCommitVer) {
return 0;
}
int32_t code = walRollback(pWal, fromIndex);
if (code != 0) {
int32_t err = terrno;

136
source/libs/sync/src/syncSnapshot.c
View File

@ -13,33 +13,25 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "syncSnapshot.h"
#include "syncIndexMgr.h"
#include "syncRaftCfg.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "wal.h"
//----------------------------------
static void snapshotSenderUpdateProgress(SSyncSnapshotSender *pSender, SyncSnapshotRsp *pMsg);
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncSnapshotSend *pBeginMsg);
static void snapshotReceiverGotData(SSyncSnapshotReceiver *pReceiver, SyncSnapshotSend *pMsg);
static int32_t snapshotReceiverFinish(SSyncSnapshotReceiver *pReceiver, SyncSnapshotSend *pMsg);
//----------------------------------
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
bool condition = (pSyncNode->pFsm->FpSnapshotStartRead != NULL) && (pSyncNode->pFsm->FpSnapshotStopRead != NULL) &&
(pSyncNode->pFsm->FpSnapshotDoRead != NULL);
SSyncSnapshotSender *pSender = NULL;
if (condition) {
pSender = taosMemoryMalloc(sizeof(SSyncSnapshotSender));
pSender = taosMemoryCalloc(1, sizeof(SSyncSnapshotSender));
if (pSender == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
memset(pSender, 0, sizeof(*pSender));
pSender->start = false;
pSender->seq = SYNC_SNAPSHOT_SEQ_INVALID;
@ -249,64 +241,6 @@ static void snapshotSenderUpdateProgress(SSyncSnapshotSender *pSender, SyncSnaps
++(pSender->seq);
}
cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender) {
char u64buf[128];
cJSON *pRoot = cJSON_CreateObject();
if (pSender != NULL) {
cJSON_AddNumberToObject(pRoot, "start", pSender->start);
cJSON_AddNumberToObject(pRoot, "seq", pSender->seq);
cJSON_AddNumberToObject(pRoot, "ack", pSender->ack);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pReader);
cJSON_AddStringToObject(pRoot, "pReader", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pCurrentBlock);
cJSON_AddStringToObject(pRoot, "pCurrentBlock", u64buf);
cJSON_AddNumberToObject(pRoot, "blockLen", pSender->blockLen);
if (pSender->pCurrentBlock != NULL) {
char *s;
s = syncUtilPrintBin((char *)(pSender->pCurrentBlock), pSender->blockLen);
cJSON_AddStringToObject(pRoot, "pCurrentBlock", s);
taosMemoryFree(s);
s = syncUtilPrintBin2((char *)(pSender->pCurrentBlock), pSender->blockLen);
cJSON_AddStringToObject(pRoot, "pCurrentBlock2", s);
taosMemoryFree(s);
}
cJSON *pSnapshot = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pSender->snapshot.lastApplyIndex);
cJSON_AddStringToObject(pSnapshot, "lastApplyIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pSender->snapshot.lastApplyTerm);
cJSON_AddStringToObject(pSnapshot, "lastApplyTerm", u64buf);
cJSON_AddItemToObject(pRoot, "snapshot", pSnapshot);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pSender->sendingMS);
cJSON_AddStringToObject(pRoot, "sendingMS", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
cJSON_AddNumberToObject(pRoot, "replicaIndex", pSender->replicaIndex);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pSender->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
// snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pSender->privateTerm);
// cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
cJSON_AddNumberToObject(pRoot, "finish", pSender->finish);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SSyncSnapshotSender", pRoot);
return pJson;
}
char *snapshotSender2Str(SSyncSnapshotSender *pSender) {
cJSON *pJson = snapshotSender2Json(pSender);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
int32_t syncNodeStartSnapshot(SSyncNode *pSyncNode, SRaftId *pDestId) {
sNTrace(pSyncNode, "starting snapshot ...");
@ -335,16 +269,17 @@ int32_t syncNodeStartSnapshot(SSyncNode *pSyncNode, SRaftId *pDestId) {
return 0;
}
// -------------------------------------
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId fromId) {
bool condition = (pSyncNode->pFsm->FpSnapshotStartWrite != NULL) && (pSyncNode->pFsm->FpSnapshotStopWrite != NULL) &&
(pSyncNode->pFsm->FpSnapshotDoWrite != NULL);
SSyncSnapshotReceiver *pReceiver = NULL;
if (condition) {
pReceiver = taosMemoryMalloc(sizeof(SSyncSnapshotReceiver));
ASSERT(pReceiver != NULL);
memset(pReceiver, 0, sizeof(*pReceiver));
pReceiver = taosMemoryCalloc(1, sizeof(SSyncSnapshotReceiver));
if (pReceiver == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
pReceiver->start = false;
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
@ -530,63 +465,6 @@ static void snapshotReceiverGotData(SSyncSnapshotReceiver *pReceiver, SyncSnapsh
}
}
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) {
char u64buf[128];
cJSON *pRoot = cJSON_CreateObject();
if (pReceiver != NULL) {
cJSON_AddNumberToObject(pRoot, "start", pReceiver->start);
cJSON_AddNumberToObject(pRoot, "ack", pReceiver->ack);
snprintf(u64buf, sizeof(u64buf), "%p", pReceiver->pWriter);
cJSON_AddStringToObject(pRoot, "pWriter", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pReceiver->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
cJSON *pFromId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pReceiver->fromId.addr);
cJSON_AddStringToObject(pFromId, "addr", u64buf);
{
uint64_t u64 = pReceiver->fromId.addr;
cJSON *pTmp = pFromId;
char host[128] = {0};
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pFromId, "vgId", pReceiver->fromId.vgId);
cJSON_AddItemToObject(pRoot, "fromId", pFromId);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pReceiver->snapshot.lastApplyIndex);
cJSON_AddStringToObject(pRoot, "snapshot.lastApplyIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pReceiver->snapshot.lastApplyTerm);
cJSON_AddStringToObject(pRoot, "snapshot.lastApplyTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pReceiver->snapshot.lastConfigIndex);
cJSON_AddStringToObject(pRoot, "snapshot.lastConfigIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pReceiver->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%" PRId64, pReceiver->startTime);
cJSON_AddStringToObject(pRoot, "startTime", u64buf);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SSyncSnapshotReceiver", pRoot);
return pJson;
}
char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver) {
cJSON *pJson = snapshotReceiver2Json(pReceiver);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
SyncIndex syncNodeGetSnapBeginIndex(SSyncNode *ths) {
SyncIndex snapStart = SYNC_INDEX_INVALID;

27
source/libs/sync/src/syncTimeout.c
View File

@ -13,28 +13,23 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#define _DEFAULT_SOURCE
#include "syncTimeout.h"
#include "syncElection.h"
#include "syncRaftCfg.h"
#include "syncRaftLog.h"
#include "syncReplication.h"
#include "syncRespMgr.h"
static void syncNodeCleanConfigIndex(SSyncNode* ths) {
int32_t newArrIndex = 0;
SyncIndex newConfigIndexArr[MAX_CONFIG_INDEX_COUNT];
memset(newConfigIndexArr, 0, sizeof(newConfigIndexArr));
SyncIndex newConfigIndexArr[MAX_CONFIG_INDEX_COUNT] = {0};
SSnapshot snapshot = {0};
if (ths->pFsm != NULL && ths->pFsm->FpGetSnapshotInfo != NULL) {
ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &snapshot);
}
ths->pFsm->FpGetSnapshotInfo(ths->pFsm, &snapshot);
if (snapshot.lastApplyIndex != SYNC_INDEX_INVALID) {
for (int i = 0; i < ths->pRaftCfg->configIndexCount; ++i) {
for (int32_t i = 0; i < ths->pRaftCfg->configIndexCount; ++i) {
if (ths->pRaftCfg->configIndexArr[i] < snapshot.lastConfigIndex) {
// pass
;
} else {
// save
newConfigIndexArr[newArrIndex] = ths->pRaftCfg->configIndexArr[i];
@ -47,13 +42,15 @@ static void syncNodeCleanConfigIndex(SSyncNode* ths) {
memcpy(ths->pRaftCfg->configIndexArr, newConfigIndexArr, sizeof(newConfigIndexArr));
int32_t code = raftCfgPersist(ths->pRaftCfg);
ASSERT(code == 0);
sNTrace(ths, "clean config index arr, old-cnt:%d, new-cnt:%d", oldCnt, ths->pRaftCfg->configIndexCount);
if (code != 0) {
sNFatal(ths, "failed to persist cfg");
} else {
sNTrace(ths, "clean config index arr, old-cnt:%d, new-cnt:%d", oldCnt, ths->pRaftCfg->configIndexCount);
}
}
}
int32_t syncNodeTimerRoutine(SSyncNode* ths) {
static int32_t syncNodeTimerRoutine(SSyncNode* ths) {
sNTrace(ths, "timer routines");
// timer replicate
@ -71,7 +68,7 @@ int32_t syncNodeTimerRoutine(SSyncNode* ths) {
SSyncLogStoreData* pData = ths->pLogStore->data;
int32_t code = walEndSnapshot(pData->pWal);
if (code != 0) {
sError("vgId:%d, timer wal snapshot end error since:%s", ths->vgId, terrstr());
sNError(ths, "timer wal snapshot end error since:%s", terrstr());
return -1;
} else {
sNTrace(ths, "wal snapshot end, index:%" PRId64, atomic_load_64(&ths->snapshottingIndex));

32
source/libs/sync/src/syncUtil.c
View File

@ -41,15 +41,15 @@ void syncUtilU642Addr(uint64_t u64, char* host, int64_t len, uint16_t* port) {
*port = (uint16_t)((u64 & 0x00000000FFFF0000) >> 16);
}
void syncUtilnodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet) {
void syncUtilNodeInfo2EpSet(const SNodeInfo* pInfo, SEpSet* pEpSet) {
pEpSet->inUse = 0;
pEpSet->numOfEps = 0;
addEpIntoEpSet(pEpSet, pInfo->nodeFqdn, pInfo->nodePort);
}
void syncUtilraftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet) {
void syncUtilRaftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet) {
char host[TSDB_FQDN_LEN] = {0};
uint16_t port;
uint16_t port = 0;
syncUtilU642Addr(raftId->addr, host, sizeof(host), &port);
pEpSet->inUse = 0;
@ -57,7 +57,7 @@ void syncUtilraftId2EpSet(const SRaftId* raftId, SEpSet* pEpSet) {
addEpIntoEpSet(pEpSet, host, port);
}
bool syncUtilnodeInfo2raftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId* raftId) {
bool syncUtilNodeInfo2RaftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId* raftId) {
uint32_t ipv4 = taosGetIpv4FromFqdn(pInfo->nodeFqdn);
if (ipv4 == 0xFFFFFFFF || ipv4 == 1) {
sError("failed to resolve ipv4 addr, fqdn: %s", pInfo->nodeFqdn);
@ -73,8 +73,7 @@ bool syncUtilnodeInfo2raftId(const SNodeInfo* pInfo, SyncGroupId vgId, SRaftId*
}
bool syncUtilSameId(const SRaftId* pId1, const SRaftId* pId2) {
bool ret = pId1->addr == pId2->addr && pId1->vgId == pId2->vgId;
return ret;
return pId1->addr == pId2->addr && pId1->vgId == pId2->vgId;
}
bool syncUtilEmptyId(const SRaftId* pId) { return (pId->addr == 0 && pId->vgId == 0); }
@ -90,18 +89,6 @@ int32_t syncUtilElectRandomMS(int32_t min, int32_t max) {
int32_t syncUtilQuorum(int32_t replicaNum) { return replicaNum / 2 + 1; }
cJSON* syncUtilNodeInfo2Json(const SNodeInfo* p) {
char u64buf[128] = {0};
cJSON* pRoot = cJSON_CreateObject();
cJSON_AddStringToObject(pRoot, "nodeFqdn", p->nodeFqdn);
cJSON_AddNumberToObject(pRoot, "nodePort", p->nodePort);
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SNodeInfo", pRoot);
return pJson;
}
cJSON* syncUtilRaftId2Json(const SRaftId* p) {
char u64buf[128] = {0};
cJSON* pRoot = cJSON_CreateObject();
@ -120,13 +107,6 @@ cJSON* syncUtilRaftId2Json(const SRaftId* p) {
return pJson;
}
char* syncUtilRaftId2Str(const SRaftId* p) {
cJSON* pJson = syncUtilRaftId2Json(p);
char* serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
static inline bool syncUtilCanPrint(char c) {
if (c >= 32 && c <= 126) {
return true;
@ -165,14 +145,12 @@ char* syncUtilPrintBin2(char* ptr, uint32_t len) {
}
void syncUtilMsgHtoN(void* msg) {
// htonl
SMsgHead* pHead = msg;
pHead->contLen = htonl(pHead->contLen);
pHead->vgId = htonl(pHead->vgId);
}
void syncUtilMsgNtoH(void* msg) {
// ntohl
SMsgHead* pHead = msg;
pHead->contLen = ntohl(pHead->contLen);
pHead->vgId = ntohl(pHead->vgId);

140
source/libs/sync/src/syncVoteMgr.c
View File

@ -23,12 +23,11 @@ static void voteGrantedClearVotes(SVotesGranted *pVotesGranted) {
}
SVotesGranted *voteGrantedCreate(SSyncNode *pSyncNode) {
SVotesGranted *pVotesGranted = taosMemoryMalloc(sizeof(SVotesGranted));
SVotesGranted *pVotesGranted = taosMemoryCalloc(1, sizeof(SVotesGranted));
if (pVotesGranted == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
memset(pVotesGranted, 0, sizeof(SVotesGranted));
pVotesGranted->replicas = &(pSyncNode->replicasId);
pVotesGranted->replicaNum = pSyncNode->replicaNum;
@ -59,20 +58,24 @@ void voteGrantedUpdate(SVotesGranted *pVotesGranted, SSyncNode *pSyncNode) {
pVotesGranted->pSyncNode = pSyncNode;
}
bool voteGrantedMajority(SVotesGranted *pVotesGranted) {
bool ret = pVotesGranted->votes >= pVotesGranted->quorum;
return ret;
}
bool voteGrantedMajority(SVotesGranted *pVotesGranted) { return pVotesGranted->votes >= pVotesGranted->quorum; }
void voteGrantedVote(SVotesGranted *pVotesGranted, SyncRequestVoteReply *pMsg) {
ASSERT(pMsg->voteGranted == true);
if (pMsg->term != pVotesGranted->term) {
sNTrace(pVotesGranted->pSyncNode, "vote grant vnode error");
if (!pMsg->voteGranted) {
sNFatal(pVotesGranted->pSyncNode, "vote granted should be true");
return;
}
ASSERT(syncUtilSameId(&pVotesGranted->pSyncNode->myRaftId, &pMsg->destId));
if (pMsg->term != pVotesGranted->term) {
sNTrace(pVotesGranted->pSyncNode, "vote grant term:%" PRId64 " not matched with msg term:%" PRId64,
pVotesGranted->term, pMsg->term);
return;
}
if (!syncUtilSameId(&pVotesGranted->pSyncNode->myRaftId, &pMsg->destId)) {
sNFatal(pVotesGranted->pSyncNode, "vote granted raftId not matched with msg");
return;
}
int32_t j = -1;
for (int32_t i = 0; i < pVotesGranted->replicaNum; ++i) {
@ -81,14 +84,21 @@ void voteGrantedVote(SVotesGranted *pVotesGranted, SyncRequestVoteReply *pMsg) {
break;
}
}
ASSERT(j != -1);
ASSERT(j >= 0 && j < pVotesGranted->replicaNum);
if ((j == -1) || !(j >= 0 && j < pVotesGranted->replicaNum)) {
sNFatal(pVotesGranted->pSyncNode, "invalid msg srcId, index:%d", j);
return;
}
if (pVotesGranted->isGranted[j] != true) {
++(pVotesGranted->votes);
pVotesGranted->isGranted[j] = true;
}
ASSERT(pVotesGranted->votes <= pVotesGranted->replicaNum);
if (pVotesGranted->votes > pVotesGranted->replicaNum) {
sNFatal(pVotesGranted->pSyncNode, "votes:%d not matched with replicaNum:%d", pVotesGranted->votes,
pVotesGranted->replicaNum);
return;
}
}
void voteGrantedReset(SVotesGranted *pVotesGranted, SyncTerm term) {
@ -97,53 +107,12 @@ void voteGrantedReset(SVotesGranted *pVotesGranted, SyncTerm term) {
pVotesGranted->toLeader = false;
}
cJSON *voteGranted2Json(SVotesGranted *pVotesGranted) {
char u64buf[128] = {0};
cJSON *pRoot = cJSON_CreateObject();
if (pVotesGranted != NULL) {
cJSON_AddNumberToObject(pRoot, "replicaNum", pVotesGranted->replicaNum);
cJSON *pReplicas = cJSON_CreateArray();
cJSON_AddItemToObject(pRoot, "replicas", pReplicas);
for (int32_t i = 0; i < pVotesGranted->replicaNum; ++i) {
cJSON_AddItemToArray(pReplicas, syncUtilRaftId2Json(&(*(pVotesGranted->replicas))[i]));
}
int32_t *arr = (int32_t *)taosMemoryMalloc(sizeof(int32_t) * pVotesGranted->replicaNum);
for (int32_t i = 0; i < pVotesGranted->replicaNum; ++i) {
arr[i] = pVotesGranted->isGranted[i];
}
cJSON *pIsGranted = cJSON_CreateIntArray(arr, pVotesGranted->replicaNum);
taosMemoryFree(arr);
cJSON_AddItemToObject(pRoot, "isGranted", pIsGranted);
cJSON_AddNumberToObject(pRoot, "votes", pVotesGranted->votes);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pVotesGranted->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
cJSON_AddNumberToObject(pRoot, "quorum", pVotesGranted->quorum);
cJSON_AddNumberToObject(pRoot, "toLeader", pVotesGranted->toLeader);
snprintf(u64buf, sizeof(u64buf), "%p", pVotesGranted->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
bool majority = voteGrantedMajority(pVotesGranted);
cJSON_AddNumberToObject(pRoot, "majority", majority);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SVotesGranted", pRoot);
return pJson;
}
char *voteGranted2Str(SVotesGranted *pVotesGranted) {
cJSON *pJson = voteGranted2Json(pVotesGranted);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
SVotesRespond *votesRespondCreate(SSyncNode *pSyncNode) {
SVotesRespond *pVotesRespond = taosMemoryMalloc(sizeof(SVotesRespond));
ASSERT(pVotesRespond != NULL);
memset(pVotesRespond, 0, sizeof(SVotesRespond));
SVotesRespond *pVotesRespond = taosMemoryCalloc(1, sizeof(SVotesRespond));
if (pVotesRespond == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
}
pVotesRespond->replicas = &(pSyncNode->replicasId);
pVotesRespond->replicaNum = pSyncNode->replicaNum;
@ -185,62 +154,15 @@ void votesRespondAdd(SVotesRespond *pVotesRespond, const SyncRequestVoteReply *p
for (int32_t i = 0; i < pVotesRespond->replicaNum; ++i) {
if (syncUtilSameId(&((*(pVotesRespond->replicas))[i]), &pMsg->srcId)) {
// ASSERT(pVotesRespond->isRespond[i] == false);
pVotesRespond->isRespond[i] = true;
return;
}
}
ASSERT(0);
sNFatal(pVotesRespond->pSyncNode, "votes respond not found");
}
void votesRespondReset(SVotesRespond *pVotesRespond, SyncTerm term) {
pVotesRespond->term = term;
memset(pVotesRespond->isRespond, 0, sizeof(pVotesRespond->isRespond));
/*
for (int32_t i = 0; i < pVotesRespond->replicaNum; ++i) {
pVotesRespond->isRespond[i] = false;
}
*/
}
cJSON *votesRespond2Json(SVotesRespond *pVotesRespond) {
char u64buf[128] = {0};
cJSON *pRoot = cJSON_CreateObject();
if (pVotesRespond != NULL) {
cJSON_AddNumberToObject(pRoot, "replicaNum", pVotesRespond->replicaNum);
cJSON *pReplicas = cJSON_CreateArray();
cJSON_AddItemToObject(pRoot, "replicas", pReplicas);
for (int32_t i = 0; i < pVotesRespond->replicaNum; ++i) {
cJSON_AddItemToArray(pReplicas, syncUtilRaftId2Json(&(*(pVotesRespond->replicas))[i]));
}
int32_t respondNum = 0;
int32_t *arr = (int32_t *)taosMemoryMalloc(sizeof(int32_t) * pVotesRespond->replicaNum);
for (int32_t i = 0; i < pVotesRespond->replicaNum; ++i) {
arr[i] = pVotesRespond->isRespond[i];
if (pVotesRespond->isRespond[i]) {
respondNum++;
}
}
cJSON *pIsRespond = cJSON_CreateIntArray(arr, pVotesRespond->replicaNum);
taosMemoryFree(arr);
cJSON_AddItemToObject(pRoot, "isRespond", pIsRespond);
cJSON_AddNumberToObject(pRoot, "respondNum", respondNum);
snprintf(u64buf, sizeof(u64buf), "%" PRIu64, pVotesRespond->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pVotesRespond->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SVotesRespond", pRoot);
return pJson;
}
char *votesRespond2Str(SVotesRespond *pVotesRespond) {
cJSON *pJson = votesRespond2Json(pVotesRespond);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}

7
source/libs/sync/test/syncAppendEntriesBatchTest.cpp
View File

@ -1,11 +1,4 @@
//#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncRaftEntry.h"
#include "syncUtil.h"
#include "trpc.h"
#include "syncTest.h"
void logTest() {

6
source/libs/sync/test/syncAppendEntriesReplyTest.cpp
View File

@ -1,9 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

6
source/libs/sync/test/syncAppendEntriesTest.cpp
View File

@ -1,9 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

6
source/libs/sync/test/syncApplyMsgTest.cpp
View File

@ -1,9 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

8
source/libs/sync/test/syncConfigChangeSnapshotTest.cpp
View File

@ -1,11 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "os.h"
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncUtil.h"
#include "wal.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

8
source/libs/sync/test/syncConfigChangeTest.cpp
View File

@ -1,11 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "os.h"
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncUtil.h"
#include "wal.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

7
source/libs/sync/test/syncElectTest.cpp
View File

@ -1,10 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncUtil.h"
#include "wal.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

10
source/libs/sync/test/syncEncodeTest.cpp
View File

@ -1,15 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncRaftEntry.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncTest.h"
#include "syncUtil.h"
#include "wal.h"
#if 0
void logTest() {

7
source/libs/sync/test/syncEnqTest.cpp
View File

@ -1,10 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

9
source/libs/sync/test/syncEntryCacheTest.cpp
View File

@ -1,12 +1,3 @@
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "tref.h"
#include "tskiplist.h"
#include "syncTest.h"
void logTest() {

8
source/libs/sync/test/syncEntryTest.cpp
View File

@ -1,11 +1,3 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {

7
source/libs/sync/test/syncEnvTest.cpp
View File

@ -1,9 +1,4 @@
#include "syncEnv.h"
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "ttime.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

8
source/libs/sync/test/syncHashCacheTest.cpp
View File

@ -1,12 +1,4 @@
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncTest.h"
#include "syncUtil.h"
#include "tskiplist.h"
void logTest() {
sTrace("--- sync log test: trace");

6
source/libs/sync/test/syncHeartbeatReplyTest.cpp
View File

@ -1,9 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

6
source/libs/sync/test/syncHeartbeatTest.cpp
View File

@ -1,9 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

8
source/libs/sync/test/syncIOClientTest.cpp
View File

@ -1,10 +1,4 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.h"
#include "tdatablock.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

10
source/libs/sync/test/syncIOSendMsgTest.cpp
View File

@ -1,10 +1,4 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");
@ -101,7 +95,7 @@ int main(int argc, char** argv) {
syncPingReply2RpcMsg(pSyncMsg, &rpcMsg);
SEpSet epSet;
syncUtilnodeInfo2EpSet(&pSyncNode->myNodeInfo, &epSet);
syncUtilNodeInfo2EpSet(&pSyncNode->myNodeInfo, &epSet);
rpcMsg.info.noResp = 1;
pSyncNode->syncSendMSg(&epSet, &rpcMsg);

6
source/libs/sync/test/syncIOServerTest.cpp
View File

@ -1,8 +1,4 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

5
source/libs/sync/test/syncIOTickPingTest.cpp
View File

@ -1,8 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

5
source/libs/sync/test/syncIOTickQTest.cpp
View File

@ -1,8 +1,5 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

10
source/libs/sync/test/syncIndexMgrTest.cpp
View File

@ -1,12 +1,4 @@
#include "syncIndexMgr.h"
//#include <gtest/gtest.h>
#include <stdio.h>
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
#include "syncTest.h"
void logTest() {
sTrace("--- sync log test: trace");

Some files were not shown because too many files have changed in this diff Show More