use of org.apache.hadoop.hive.ql.plan.FileSinkDesc in project hive by apache.
the class GenMapRedUtils method createMoveTask.
/**
* Create and add any dependent move tasks
*
* @param currTask
* @param chDir
* @param fsOp
* @param parseCtx
* @param mvTasks
* @param hconf
* @param dependencyTask
* @return
*/
public static Path createMoveTask(Task<? extends Serializable> currTask, boolean chDir, FileSinkOperator fsOp, ParseContext parseCtx, List<Task<MoveWork>> mvTasks, HiveConf hconf, DependencyCollectionTask dependencyTask) {
Path dest = null;
if (chDir) {
FileSinkDesc fileSinkDesc = fsOp.getConf();
dest = fileSinkDesc.getFinalDirName();
// generate the temporary file
// it must be on the same file system as the current destination
Context baseCtx = parseCtx.getContext();
// Create the required temporary file in the HDFS location if the destination
// path of the FileSinkOperator table is a blobstore path.
Path tmpDir = baseCtx.getTempDirForPath(fileSinkDesc.getDestPath(), true);
// Change all the linked file sink descriptors
if (fileSinkDesc.isLinkedFileSink()) {
for (FileSinkDesc fsConf : fileSinkDesc.getLinkedFileSinkDesc()) {
fsConf.setParentDir(tmpDir);
fsConf.setDirName(new Path(tmpDir, fsConf.getDirName().getName()));
}
} else {
fileSinkDesc.setDirName(tmpDir);
}
}
Task<MoveWork> mvTask = null;
if (!chDir) {
mvTask = GenMapRedUtils.findMoveTask(mvTasks, fsOp);
}
// Set the move task to be dependent on the current task
if (mvTask != null) {
GenMapRedUtils.addDependentMoveTasks(mvTask, hconf, currTask, dependencyTask);
}
return dest;
}
use of org.apache.hadoop.hive.ql.plan.FileSinkDesc in project hive by apache.
the class SemanticAnalyzer method genFileSinkPlan.
@SuppressWarnings("nls")
protected Operator genFileSinkPlan(String dest, QB qb, Operator input) throws SemanticException {
RowResolver inputRR = opParseCtx.get(input).getRowResolver();
QBMetaData qbm = qb.getMetaData();
Integer dest_type = qbm.getDestTypeForAlias(dest);
// destination table if any
Table dest_tab = null;
// should the destination table be written to using ACID
boolean destTableIsAcid = false;
boolean destTableIsTemporary = false;
boolean destTableIsMaterialization = false;
// destination partition if any
Partition dest_part = null;
// the intermediate destination directory
Path queryTmpdir = null;
// the final destination directory
Path dest_path = null;
TableDesc table_desc = null;
int currentTableId = 0;
boolean isLocal = false;
SortBucketRSCtx rsCtx = new SortBucketRSCtx();
DynamicPartitionCtx dpCtx = null;
LoadTableDesc ltd = null;
ListBucketingCtx lbCtx = null;
Map<String, String> partSpec = null;
switch(dest_type.intValue()) {
case QBMetaData.DEST_TABLE:
{
dest_tab = qbm.getDestTableForAlias(dest);
destTableIsAcid = AcidUtils.isAcidTable(dest_tab);
destTableIsTemporary = dest_tab.isTemporary();
// Is the user trying to insert into a external tables
if ((!conf.getBoolVar(HiveConf.ConfVars.HIVE_INSERT_INTO_EXTERNAL_TABLES)) && (dest_tab.getTableType().equals(TableType.EXTERNAL_TABLE))) {
throw new SemanticException(ErrorMsg.INSERT_EXTERNAL_TABLE.getMsg(dest_tab.getTableName()));
}
partSpec = qbm.getPartSpecForAlias(dest);
dest_path = dest_tab.getPath();
// verify that our destination is empty before proceeding
if (dest_tab.isImmutable() && qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName())) {
try {
FileSystem fs = dest_path.getFileSystem(conf);
if (!MetaStoreUtils.isDirEmpty(fs, dest_path)) {
LOG.warn("Attempted write into an immutable table : " + dest_tab.getTableName() + " : " + dest_path);
throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(dest_tab.getTableName()));
}
} catch (IOException ioe) {
LOG.warn("Error while trying to determine if immutable table has any data : " + dest_tab.getTableName() + " : " + dest_path);
throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(ioe.getMessage()));
}
}
// check for partition
List<FieldSchema> parts = dest_tab.getPartitionKeys();
if (parts != null && parts.size() > 0) {
// table is partitioned
if (partSpec == null || partSpec.size() == 0) {
// user did NOT specify partition
throw new SemanticException(generateErrorMessage(qb.getParseInfo().getDestForClause(dest), ErrorMsg.NEED_PARTITION_ERROR.getMsg()));
}
dpCtx = qbm.getDPCtx(dest);
if (dpCtx == null) {
dest_tab.validatePartColumnNames(partSpec, false);
dpCtx = new DynamicPartitionCtx(dest_tab, partSpec, conf.getVar(HiveConf.ConfVars.DEFAULTPARTITIONNAME), conf.getIntVar(HiveConf.ConfVars.DYNAMICPARTITIONMAXPARTSPERNODE));
qbm.setDPCtx(dest, dpCtx);
}
if (!HiveConf.getBoolVar(conf, HiveConf.ConfVars.DYNAMICPARTITIONING)) {
// allow DP
throw new SemanticException(generateErrorMessage(qb.getParseInfo().getDestForClause(dest), ErrorMsg.DYNAMIC_PARTITION_DISABLED.getMsg()));
}
if (dpCtx.getSPPath() != null) {
dest_path = new Path(dest_tab.getPath(), dpCtx.getSPPath());
}
if ((dest_tab.getNumBuckets() > 0)) {
dpCtx.setNumBuckets(dest_tab.getNumBuckets());
}
}
boolean isNonNativeTable = dest_tab.isNonNative();
if (isNonNativeTable) {
queryTmpdir = dest_path;
} else {
queryTmpdir = ctx.getTempDirForPath(dest_path, true);
}
if (dpCtx != null) {
// set the root of the temporary path where dynamic partition columns will populate
dpCtx.setRootPath(queryTmpdir);
}
// this table_desc does not contain the partitioning columns
table_desc = Utilities.getTableDesc(dest_tab);
// Add sorting/bucketing if needed
input = genBucketingSortingDest(dest, input, qb, table_desc, dest_tab, rsCtx);
idToTableNameMap.put(String.valueOf(destTableId), dest_tab.getTableName());
currentTableId = destTableId;
destTableId++;
lbCtx = constructListBucketingCtx(dest_tab.getSkewedColNames(), dest_tab.getSkewedColValues(), dest_tab.getSkewedColValueLocationMaps(), dest_tab.isStoredAsSubDirectories(), conf);
// NOTE: specify Dynamic partitions in dest_tab for WriteEntity
if (!isNonNativeTable) {
AcidUtils.Operation acidOp = AcidUtils.Operation.NOT_ACID;
if (destTableIsAcid) {
acidOp = getAcidType(table_desc.getOutputFileFormatClass(), dest);
checkAcidConstraints(qb, table_desc, dest_tab);
}
ltd = new LoadTableDesc(queryTmpdir, table_desc, dpCtx, acidOp);
ltd.setReplace(!qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
ltd.setLbCtx(lbCtx);
loadTableWork.add(ltd);
} else {
// This is a non-native table.
// We need to set stats as inaccurate.
setStatsForNonNativeTable(dest_tab);
// true if it is insert overwrite.
boolean overwrite = !qb.getParseInfo().isInsertIntoTable(String.format("%s.%s", dest_tab.getDbName(), dest_tab.getTableName()));
createInsertDesc(dest_tab, overwrite);
}
WriteEntity output = null;
// list of dynamically created partitions are known.
if ((dpCtx == null || dpCtx.getNumDPCols() == 0)) {
output = new WriteEntity(dest_tab, determineWriteType(ltd, isNonNativeTable, dest));
if (!outputs.add(output)) {
throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_tab.getTableName()));
}
}
if ((dpCtx != null) && (dpCtx.getNumDPCols() >= 0)) {
// No static partition specified
if (dpCtx.getNumSPCols() == 0) {
output = new WriteEntity(dest_tab, determineWriteType(ltd, isNonNativeTable, dest), false);
outputs.add(output);
output.setDynamicPartitionWrite(true);
} else // part of the partition specified
// Create a DummyPartition in this case. Since, the metastore does not store partial
// partitions currently, we need to store dummy partitions
{
try {
String ppath = dpCtx.getSPPath();
ppath = ppath.substring(0, ppath.length() - 1);
DummyPartition p = new DummyPartition(dest_tab, dest_tab.getDbName() + "@" + dest_tab.getTableName() + "@" + ppath, partSpec);
output = new WriteEntity(p, getWriteType(dest), false);
output.setDynamicPartitionWrite(true);
outputs.add(output);
} catch (HiveException e) {
throw new SemanticException(e.getMessage(), e);
}
}
}
ctx.getLoadTableOutputMap().put(ltd, output);
break;
}
case QBMetaData.DEST_PARTITION:
{
dest_part = qbm.getDestPartitionForAlias(dest);
dest_tab = dest_part.getTable();
destTableIsAcid = AcidUtils.isAcidTable(dest_tab);
if ((!conf.getBoolVar(HiveConf.ConfVars.HIVE_INSERT_INTO_EXTERNAL_TABLES)) && dest_tab.getTableType().equals(TableType.EXTERNAL_TABLE)) {
throw new SemanticException(ErrorMsg.INSERT_EXTERNAL_TABLE.getMsg(dest_tab.getTableName()));
}
Path tabPath = dest_tab.getPath();
Path partPath = dest_part.getDataLocation();
// verify that our destination is empty before proceeding
if (dest_tab.isImmutable() && qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName())) {
try {
FileSystem fs = partPath.getFileSystem(conf);
if (!MetaStoreUtils.isDirEmpty(fs, partPath)) {
LOG.warn("Attempted write into an immutable table partition : " + dest_tab.getTableName() + " : " + partPath);
throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(dest_tab.getTableName()));
}
} catch (IOException ioe) {
LOG.warn("Error while trying to determine if immutable table partition has any data : " + dest_tab.getTableName() + " : " + partPath);
throw new SemanticException(ErrorMsg.INSERT_INTO_IMMUTABLE_TABLE.getMsg(ioe.getMessage()));
}
}
// if the table is in a different dfs than the partition,
// replace the partition's dfs with the table's dfs.
dest_path = new Path(tabPath.toUri().getScheme(), tabPath.toUri().getAuthority(), partPath.toUri().getPath());
queryTmpdir = ctx.getTempDirForPath(dest_path, true);
table_desc = Utilities.getTableDesc(dest_tab);
// Add sorting/bucketing if needed
input = genBucketingSortingDest(dest, input, qb, table_desc, dest_tab, rsCtx);
idToTableNameMap.put(String.valueOf(destTableId), dest_tab.getTableName());
currentTableId = destTableId;
destTableId++;
lbCtx = constructListBucketingCtx(dest_part.getSkewedColNames(), dest_part.getSkewedColValues(), dest_part.getSkewedColValueLocationMaps(), dest_part.isStoredAsSubDirectories(), conf);
AcidUtils.Operation acidOp = AcidUtils.Operation.NOT_ACID;
if (destTableIsAcid) {
acidOp = getAcidType(table_desc.getOutputFileFormatClass(), dest);
checkAcidConstraints(qb, table_desc, dest_tab);
}
ltd = new LoadTableDesc(queryTmpdir, table_desc, dest_part.getSpec(), acidOp);
ltd.setReplace(!qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
ltd.setLbCtx(lbCtx);
loadTableWork.add(ltd);
if (!outputs.add(new WriteEntity(dest_part, determineWriteType(ltd, dest_tab.isNonNative(), dest)))) {
throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_tab.getTableName() + "@" + dest_part.getName()));
}
break;
}
case QBMetaData.DEST_LOCAL_FILE:
isLocal = true;
// fall through
case QBMetaData.DEST_DFS_FILE:
{
dest_path = new Path(qbm.getDestFileForAlias(dest));
if (isLocal) {
// for local directory - we always write to map-red intermediate
// store and then copy to local fs
queryTmpdir = ctx.getMRTmpPath();
} else {
try {
Path qPath = FileUtils.makeQualified(dest_path, conf);
queryTmpdir = ctx.getTempDirForPath(qPath, true);
} catch (Exception e) {
throw new SemanticException("Error creating temporary folder on: " + dest_path, e);
}
}
String cols = "";
String colTypes = "";
ArrayList<ColumnInfo> colInfos = inputRR.getColumnInfos();
// CTAS case: the file output format and serde are defined by the create
// table command rather than taking the default value
List<FieldSchema> field_schemas = null;
CreateTableDesc tblDesc = qb.getTableDesc();
CreateViewDesc viewDesc = qb.getViewDesc();
if (tblDesc != null) {
field_schemas = new ArrayList<FieldSchema>();
destTableIsTemporary = tblDesc.isTemporary();
destTableIsMaterialization = tblDesc.isMaterialization();
} else if (viewDesc != null) {
field_schemas = new ArrayList<FieldSchema>();
destTableIsTemporary = false;
}
boolean first = true;
for (ColumnInfo colInfo : colInfos) {
String[] nm = inputRR.reverseLookup(colInfo.getInternalName());
if (nm[1] != null) {
// non-null column alias
colInfo.setAlias(nm[1]);
}
//default column name
String colName = colInfo.getInternalName();
if (field_schemas != null) {
FieldSchema col = new FieldSchema();
if (!("".equals(nm[0])) && nm[1] != null) {
// remove ``
colName = unescapeIdentifier(colInfo.getAlias()).toLowerCase();
}
colName = fixCtasColumnName(colName);
col.setName(colName);
String typeName = colInfo.getType().getTypeName();
// CTAS should NOT create a VOID type
if (typeName.equals(serdeConstants.VOID_TYPE_NAME)) {
throw new SemanticException(ErrorMsg.CTAS_CREATES_VOID_TYPE.getMsg(colName));
}
col.setType(typeName);
field_schemas.add(col);
}
if (!first) {
cols = cols.concat(",");
colTypes = colTypes.concat(":");
}
first = false;
cols = cols.concat(colName);
// Replace VOID type with string when the output is a temp table or
// local files.
// A VOID type can be generated under the query:
//
// select NULL from tt;
// or
// insert overwrite local directory "abc" select NULL from tt;
//
// where there is no column type to which the NULL value should be
// converted.
//
String tName = colInfo.getType().getTypeName();
if (tName.equals(serdeConstants.VOID_TYPE_NAME)) {
colTypes = colTypes.concat(serdeConstants.STRING_TYPE_NAME);
} else {
colTypes = colTypes.concat(tName);
}
}
// update the create table descriptor with the resulting schema.
if (tblDesc != null) {
tblDesc.setCols(new ArrayList<FieldSchema>(field_schemas));
} else if (viewDesc != null) {
viewDesc.setSchema(new ArrayList<FieldSchema>(field_schemas));
}
boolean isDestTempFile = true;
if (!ctx.isMRTmpFileURI(dest_path.toUri().toString())) {
idToTableNameMap.put(String.valueOf(destTableId), dest_path.toUri().toString());
currentTableId = destTableId;
destTableId++;
isDestTempFile = false;
}
boolean isDfsDir = (dest_type.intValue() == QBMetaData.DEST_DFS_FILE);
loadFileWork.add(new LoadFileDesc(tblDesc, viewDesc, queryTmpdir, dest_path, isDfsDir, cols, colTypes));
if (tblDesc == null) {
if (viewDesc != null) {
table_desc = PlanUtils.getTableDesc(viewDesc, cols, colTypes);
} else if (qb.getIsQuery()) {
String fileFormat;
if (SessionState.get().getIsUsingThriftJDBCBinarySerDe()) {
fileFormat = "SequenceFile";
HiveConf.setVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT, fileFormat);
table_desc = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, fileFormat, ThriftJDBCBinarySerDe.class);
// Set the fetch formatter to be a no-op for the ListSinkOperator, since we'll
// write out formatted thrift objects to SequenceFile
conf.set(SerDeUtils.LIST_SINK_OUTPUT_FORMATTER, NoOpFetchFormatter.class.getName());
} else {
fileFormat = HiveConf.getVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT);
table_desc = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, fileFormat, LazySimpleSerDe.class);
}
} else {
table_desc = PlanUtils.getDefaultTableDesc(qb.getDirectoryDesc(), cols, colTypes);
}
} else {
table_desc = PlanUtils.getTableDesc(tblDesc, cols, colTypes);
}
if (!outputs.add(new WriteEntity(dest_path, !isDfsDir, isDestTempFile))) {
throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES.getMsg(dest_path.toUri().toString()));
}
break;
}
default:
throw new SemanticException("Unknown destination type: " + dest_type);
}
input = genConversionSelectOperator(dest, qb, input, table_desc, dpCtx);
inputRR = opParseCtx.get(input).getRowResolver();
ArrayList<ColumnInfo> vecCol = new ArrayList<ColumnInfo>();
if (updating(dest) || deleting(dest)) {
vecCol.add(new ColumnInfo(VirtualColumn.ROWID.getName(), VirtualColumn.ROWID.getTypeInfo(), "", true));
} else {
try {
StructObjectInspector rowObjectInspector = (StructObjectInspector) table_desc.getDeserializer(conf).getObjectInspector();
List<? extends StructField> fields = rowObjectInspector.getAllStructFieldRefs();
for (int i = 0; i < fields.size(); i++) {
vecCol.add(new ColumnInfo(fields.get(i).getFieldName(), TypeInfoUtils.getTypeInfoFromObjectInspector(fields.get(i).getFieldObjectInspector()), "", false));
}
} catch (Exception e) {
throw new SemanticException(e.getMessage(), e);
}
}
RowSchema fsRS = new RowSchema(vecCol);
// The output files of a FileSink can be merged if they are either not being written to a table
// or are being written to a table which is not bucketed
// and table the table is not sorted
boolean canBeMerged = (dest_tab == null || !((dest_tab.getNumBuckets() > 0) || (dest_tab.getSortCols() != null && dest_tab.getSortCols().size() > 0)));
// If this table is working with ACID semantics, turn off merging
canBeMerged &= !destTableIsAcid;
// Generate the partition columns from the parent input
if (dest_type.intValue() == QBMetaData.DEST_TABLE || dest_type.intValue() == QBMetaData.DEST_PARTITION) {
genPartnCols(dest, input, qb, table_desc, dest_tab, rsCtx);
}
FileSinkDesc fileSinkDesc = new FileSinkDesc(queryTmpdir, table_desc, conf.getBoolVar(HiveConf.ConfVars.COMPRESSRESULT), currentTableId, rsCtx.isMultiFileSpray(), canBeMerged, rsCtx.getNumFiles(), rsCtx.getTotalFiles(), rsCtx.getPartnCols(), dpCtx, dest_path);
boolean isHiveServerQuery = SessionState.get().isHiveServerQuery();
fileSinkDesc.setHiveServerQuery(isHiveServerQuery);
// FileSinkOperator knows how to properly write to it.
if (destTableIsAcid) {
AcidUtils.Operation wt = updating(dest) ? AcidUtils.Operation.UPDATE : (deleting(dest) ? AcidUtils.Operation.DELETE : AcidUtils.Operation.INSERT);
fileSinkDesc.setWriteType(wt);
acidFileSinks.add(fileSinkDesc);
}
fileSinkDesc.setTemporary(destTableIsTemporary);
fileSinkDesc.setMaterialization(destTableIsMaterialization);
/* Set List Bucketing context. */
if (lbCtx != null) {
lbCtx.processRowSkewedIndex(fsRS);
lbCtx.calculateSkewedValueSubDirList();
}
fileSinkDesc.setLbCtx(lbCtx);
// set the stats publishing/aggregating key prefix
// the same as directory name. The directory name
// can be changed in the optimizer but the key should not be changed
// it should be the same as the MoveWork's sourceDir.
fileSinkDesc.setStatsAggPrefix(fileSinkDesc.getDirName().toString());
if (!destTableIsMaterialization && HiveConf.getVar(conf, HIVESTATSDBCLASS).equalsIgnoreCase(StatDB.fs.name())) {
String statsTmpLoc = ctx.getTempDirForPath(dest_path).toString();
fileSinkDesc.setStatsTmpDir(statsTmpLoc);
LOG.debug("Set stats collection dir : " + statsTmpLoc);
}
if (dest_part != null) {
try {
String staticSpec = Warehouse.makePartPath(dest_part.getSpec());
fileSinkDesc.setStaticSpec(staticSpec);
} catch (MetaException e) {
throw new SemanticException(e);
}
} else if (dpCtx != null) {
fileSinkDesc.setStaticSpec(dpCtx.getSPPath());
}
if (isHiveServerQuery && null != table_desc && table_desc.getSerdeClassName().equalsIgnoreCase(ThriftJDBCBinarySerDe.class.getName()) && HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_SERVER2_THRIFT_RESULTSET_SERIALIZE_IN_TASKS)) {
fileSinkDesc.setIsUsingThriftJDBCBinarySerDe(true);
} else {
fileSinkDesc.setIsUsingThriftJDBCBinarySerDe(false);
}
Operator output = putOpInsertMap(OperatorFactory.getAndMakeChild(fileSinkDesc, fsRS, input), inputRR);
if (ltd != null && SessionState.get() != null) {
SessionState.get().getLineageState().mapDirToOp(ltd.getSourcePath(), (FileSinkOperator) output);
} else if (queryState.getCommandType().equals(HiveOperation.CREATETABLE_AS_SELECT.getOperationName())) {
Path tlocation = null;
String tName = Utilities.getDbTableName(tableDesc.getTableName())[1];
try {
Warehouse wh = new Warehouse(conf);
tlocation = wh.getTablePath(db.getDatabase(tableDesc.getDatabaseName()), tName);
} catch (MetaException | HiveException e) {
throw new SemanticException(e);
}
SessionState.get().getLineageState().mapDirToOp(tlocation, (FileSinkOperator) output);
}
if (LOG.isDebugEnabled()) {
LOG.debug("Created FileSink Plan for clause: " + dest + "dest_path: " + dest_path + " row schema: " + inputRR.toString());
}
FileSinkOperator fso = (FileSinkOperator) output;
fso.getConf().setTable(dest_tab);
fsopToTable.put(fso, dest_tab);
// and it is an insert overwrite or insert into table
if (dest_tab != null && conf.getBoolVar(ConfVars.HIVESTATSAUTOGATHER) && conf.getBoolVar(ConfVars.HIVESTATSCOLAUTOGATHER) && ColumnStatsAutoGatherContext.canRunAutogatherStats(fso)) {
if (dest_type.intValue() == QBMetaData.DEST_TABLE) {
genAutoColumnStatsGatheringPipeline(qb, table_desc, partSpec, input, qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
} else if (dest_type.intValue() == QBMetaData.DEST_PARTITION) {
genAutoColumnStatsGatheringPipeline(qb, table_desc, dest_part.getSpec(), input, qb.getParseInfo().isInsertIntoTable(dest_tab.getDbName(), dest_tab.getTableName()));
}
}
return output;
}
use of org.apache.hadoop.hive.ql.plan.FileSinkDesc in project hive by apache.
the class GenMapRedUtils method createTemporaryFile.
/**
* Break the pipeline between parent and child, and then
* output data generated by parent to a temporary file stored in taskTmpDir.
* A FileSinkOperator is added after parent to output the data.
* Before child, we add a TableScanOperator to load data stored in the temporary
* file back.
* @param parent
* @param child
* @param taskTmpDir
* @param tt_desc
* @param parseCtx
* @return The TableScanOperator inserted before child.
*/
public static TableScanOperator createTemporaryFile(Operator<? extends OperatorDesc> parent, Operator<? extends OperatorDesc> child, Path taskTmpDir, TableDesc tt_desc, ParseContext parseCtx) {
// Create a FileSinkOperator for the file name of taskTmpDir
boolean compressIntermediate = parseCtx.getConf().getBoolVar(HiveConf.ConfVars.COMPRESSINTERMEDIATE);
FileSinkDesc desc = new FileSinkDesc(taskTmpDir, tt_desc, compressIntermediate);
if (compressIntermediate) {
desc.setCompressCodec(parseCtx.getConf().getVar(HiveConf.ConfVars.COMPRESSINTERMEDIATECODEC));
desc.setCompressType(parseCtx.getConf().getVar(HiveConf.ConfVars.COMPRESSINTERMEDIATETYPE));
}
Operator<? extends OperatorDesc> fileSinkOp = OperatorFactory.get(parent.getCompilationOpContext(), desc, parent.getSchema());
// Connect parent to fileSinkOp
parent.replaceChild(child, fileSinkOp);
fileSinkOp.setParentOperators(Utilities.makeList(parent));
// Create a dummy TableScanOperator for the file generated through fileSinkOp
TableScanOperator tableScanOp = createTemporaryTableScanOperator(parent.getCompilationOpContext(), parent.getSchema());
// Connect this TableScanOperator to child.
tableScanOp.setChildOperators(Utilities.makeList(child));
child.replaceParent(parent, tableScanOp);
return tableScanOp;
}
use of org.apache.hadoop.hive.ql.plan.FileSinkDesc in project hive by apache.
the class Vectorizer method vectorizeOperator.
public Operator<? extends OperatorDesc> vectorizeOperator(Operator<? extends OperatorDesc> op, VectorizationContext vContext, boolean isTezOrSpark, VectorTaskColumnInfo vectorTaskColumnInfo) throws HiveException {
Operator<? extends OperatorDesc> vectorOp = null;
boolean isNative;
switch(op.getType()) {
case TABLESCAN:
vectorOp = vectorizeTableScanOperator(op, vContext);
isNative = true;
break;
case MAPJOIN:
{
if (op instanceof MapJoinOperator) {
VectorMapJoinInfo vectorMapJoinInfo = new VectorMapJoinInfo();
MapJoinDesc desc = (MapJoinDesc) op.getConf();
boolean specialize = canSpecializeMapJoin(op, desc, isTezOrSpark, vContext, vectorMapJoinInfo);
if (!specialize) {
Class<? extends Operator<?>> opClass = null;
// *NON-NATIVE* vector map differences for LEFT OUTER JOIN and Filtered...
List<ExprNodeDesc> bigTableFilters = desc.getFilters().get((byte) desc.getPosBigTable());
boolean isOuterAndFiltered = (!desc.isNoOuterJoin() && bigTableFilters.size() > 0);
if (!isOuterAndFiltered) {
opClass = VectorMapJoinOperator.class;
} else {
opClass = VectorMapJoinOuterFilteredOperator.class;
}
vectorOp = OperatorFactory.getVectorOperator(opClass, op.getCompilationOpContext(), op.getConf(), vContext);
isNative = false;
} else {
// TEMPORARY Until Native Vector Map Join with Hybrid passes tests...
// HiveConf.setBoolVar(physicalContext.getConf(),
// HiveConf.ConfVars.HIVEUSEHYBRIDGRACEHASHJOIN, false);
vectorOp = specializeMapJoinOperator(op, vContext, desc, vectorMapJoinInfo);
isNative = true;
if (vectorTaskColumnInfo != null) {
if (usesVectorUDFAdaptor(vectorMapJoinInfo.getBigTableKeyExpressions())) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
if (usesVectorUDFAdaptor(vectorMapJoinInfo.getBigTableValueExpressions())) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
}
}
} else {
Preconditions.checkState(op instanceof SMBMapJoinOperator);
SMBJoinDesc smbJoinSinkDesc = (SMBJoinDesc) op.getConf();
VectorSMBJoinDesc vectorSMBJoinDesc = new VectorSMBJoinDesc();
smbJoinSinkDesc.setVectorDesc(vectorSMBJoinDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), smbJoinSinkDesc, vContext);
isNative = false;
}
}
break;
case REDUCESINK:
{
VectorReduceSinkInfo vectorReduceSinkInfo = new VectorReduceSinkInfo();
ReduceSinkDesc desc = (ReduceSinkDesc) op.getConf();
boolean specialize = canSpecializeReduceSink(desc, isTezOrSpark, vContext, vectorReduceSinkInfo);
if (!specialize) {
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), op.getConf(), vContext);
isNative = false;
} else {
vectorOp = specializeReduceSinkOperator(op, vContext, desc, vectorReduceSinkInfo);
isNative = true;
if (vectorTaskColumnInfo != null) {
if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkKeyExpressions())) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
if (usesVectorUDFAdaptor(vectorReduceSinkInfo.getReduceSinkValueExpressions())) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
}
}
}
break;
case FILTER:
{
vectorOp = vectorizeFilterOperator(op, vContext);
isNative = true;
if (vectorTaskColumnInfo != null) {
VectorFilterDesc vectorFilterDesc = (VectorFilterDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
VectorExpression vectorPredicateExpr = vectorFilterDesc.getPredicateExpression();
if (usesVectorUDFAdaptor(vectorPredicateExpr)) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
}
}
break;
case SELECT:
{
vectorOp = vectorizeSelectOperator(op, vContext);
isNative = true;
if (vectorTaskColumnInfo != null) {
VectorSelectDesc vectorSelectDesc = (VectorSelectDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
VectorExpression[] vectorSelectExprs = vectorSelectDesc.getSelectExpressions();
if (usesVectorUDFAdaptor(vectorSelectExprs)) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
}
}
break;
case GROUPBY:
{
vectorOp = vectorizeGroupByOperator(op, vContext);
isNative = false;
if (vectorTaskColumnInfo != null) {
VectorGroupByDesc vectorGroupByDesc = (VectorGroupByDesc) ((AbstractOperatorDesc) vectorOp.getConf()).getVectorDesc();
if (!vectorGroupByDesc.isVectorOutput()) {
vectorTaskColumnInfo.setGroupByVectorOutput(false);
}
VectorExpression[] vecKeyExpressions = vectorGroupByDesc.getKeyExpressions();
if (usesVectorUDFAdaptor(vecKeyExpressions)) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
VectorAggregateExpression[] vecAggregators = vectorGroupByDesc.getAggregators();
for (VectorAggregateExpression vecAggr : vecAggregators) {
if (usesVectorUDFAdaptor(vecAggr.inputExpression())) {
vectorTaskColumnInfo.setUsesVectorUDFAdaptor(true);
}
}
}
}
break;
case FILESINK:
{
FileSinkDesc fileSinkDesc = (FileSinkDesc) op.getConf();
VectorFileSinkDesc vectorFileSinkDesc = new VectorFileSinkDesc();
fileSinkDesc.setVectorDesc(vectorFileSinkDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), fileSinkDesc, vContext);
isNative = false;
}
break;
case LIMIT:
{
LimitDesc limitDesc = (LimitDesc) op.getConf();
VectorLimitDesc vectorLimitDesc = new VectorLimitDesc();
limitDesc.setVectorDesc(vectorLimitDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), limitDesc, vContext);
isNative = true;
}
break;
case EVENT:
{
AppMasterEventDesc eventDesc = (AppMasterEventDesc) op.getConf();
VectorAppMasterEventDesc vectorEventDesc = new VectorAppMasterEventDesc();
eventDesc.setVectorDesc(vectorEventDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), eventDesc, vContext);
isNative = true;
}
break;
case HASHTABLESINK:
{
SparkHashTableSinkDesc sparkHashTableSinkDesc = (SparkHashTableSinkDesc) op.getConf();
VectorSparkHashTableSinkDesc vectorSparkHashTableSinkDesc = new VectorSparkHashTableSinkDesc();
sparkHashTableSinkDesc.setVectorDesc(vectorSparkHashTableSinkDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkHashTableSinkDesc, vContext);
isNative = true;
}
break;
case SPARKPRUNINGSINK:
{
SparkPartitionPruningSinkDesc sparkPartitionPruningSinkDesc = (SparkPartitionPruningSinkDesc) op.getConf();
VectorSparkPartitionPruningSinkDesc vectorSparkPartitionPruningSinkDesc = new VectorSparkPartitionPruningSinkDesc();
sparkPartitionPruningSinkDesc.setVectorDesc(vectorSparkPartitionPruningSinkDesc);
vectorOp = OperatorFactory.getVectorOperator(op.getCompilationOpContext(), sparkPartitionPruningSinkDesc, vContext);
isNative = true;
}
break;
default:
// These are children of GROUP BY operators with non-vector outputs.
isNative = false;
vectorOp = op;
break;
}
Preconditions.checkState(vectorOp != null);
if (vectorTaskColumnInfo != null && !isNative) {
vectorTaskColumnInfo.setAllNative(false);
}
LOG.debug("vectorizeOperator " + vectorOp.getClass().getName());
LOG.debug("vectorizeOperator " + vectorOp.getConf().getClass().getName());
if (vectorOp != op) {
fixupParentChildOperators(op, vectorOp);
((AbstractOperatorDesc) vectorOp.getConf()).setVectorMode(true);
}
return vectorOp;
}
use of org.apache.hadoop.hive.ql.plan.FileSinkDesc in project hive by apache.
the class Utilities method getStatsTmpDirs.
public static List<String> getStatsTmpDirs(BaseWork work, Configuration conf) {
List<String> statsTmpDirs = new ArrayList<>();
if (!StatsSetupConst.StatDB.fs.name().equalsIgnoreCase(HiveConf.getVar(conf, ConfVars.HIVESTATSDBCLASS))) {
// no-op for non-fs stats collection
return statsTmpDirs;
}
// if its auto-stats gather for inserts or CTAS, stats dir will be in FileSink
Set<Operator<? extends OperatorDesc>> ops = work.getAllLeafOperators();
if (work instanceof MapWork) {
// if its an anlayze statement, stats dir will be in TableScan
ops.addAll(work.getAllRootOperators());
}
for (Operator<? extends OperatorDesc> op : ops) {
OperatorDesc desc = op.getConf();
String statsTmpDir = null;
if (desc instanceof FileSinkDesc) {
statsTmpDir = ((FileSinkDesc) desc).getStatsTmpDir();
} else if (desc instanceof TableScanDesc) {
statsTmpDir = ((TableScanDesc) desc).getTmpStatsDir();
}
if (statsTmpDir != null && !statsTmpDir.isEmpty()) {
statsTmpDirs.add(statsTmpDir);
}
}
return statsTmpDirs;
}
Aggregations