Examples with ListBucketingCtx org.apache.hadoop.hive.ql.plan.ListBucketingCtx used on opensource projects

Example 1 with ListBucketingCtx

use of org.apache.hadoop.hive.ql.plan.ListBucketingCtx in project hive by apache.

the class DDLSemanticAnalyzer method analyzeAlterTablePartMergeFiles.

private void analyzeAlterTablePartMergeFiles(ASTNode ast, String tableName, HashMap<String, String> partSpec) throws SemanticException {
    AlterTablePartMergeFilesDesc mergeDesc = new AlterTablePartMergeFilesDesc(tableName, partSpec);
    List<Path> inputDir = new ArrayList<Path>();
    Path oldTblPartLoc = null;
    Path newTblPartLoc = null;
    Table tblObj = null;
    ListBucketingCtx lbCtx = null;
    try {
        tblObj = getTable(tableName);
        List<String> bucketCols = null;
        Class<? extends InputFormat> inputFormatClass = null;
        boolean isArchived = false;
        boolean checkIndex = HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVE_CONCATENATE_CHECK_INDEX);
        if (checkIndex) {
            List<Index> indexes = db.getIndexes(tblObj.getDbName(), tblObj.getTableName(), Short.MAX_VALUE);
            if (indexes != null && indexes.size() > 0) {
                throw new SemanticException("can not do merge because source table " + tableName + " is indexed.");
            }
        }
        if (tblObj.isPartitioned()) {
            if (partSpec == null) {
                throw new SemanticException("source table " + tableName + " is partitioned but no partition desc found.");
            } else {
                Partition part = getPartition(tblObj, partSpec, false);
                if (part == null) {
                    throw new SemanticException("source table " + tableName + " is partitioned but partition not found.");
                }
                bucketCols = part.getBucketCols();
                inputFormatClass = part.getInputFormatClass();
                isArchived = ArchiveUtils.isArchived(part);
                Path tabPath = tblObj.getPath();
                Path partPath = part.getDataLocation();
                // if the table is in a different dfs than the partition,
                // replace the partition's dfs with the table's dfs.
                newTblPartLoc = new Path(tabPath.toUri().getScheme(), tabPath.toUri().getAuthority(), partPath.toUri().getPath());
                oldTblPartLoc = partPath;
                lbCtx = constructListBucketingCtx(part.getSkewedColNames(), part.getSkewedColValues(), part.getSkewedColValueLocationMaps(), part.isStoredAsSubDirectories(), conf);
            }
        } else {
            inputFormatClass = tblObj.getInputFormatClass();
            bucketCols = tblObj.getBucketCols();
            // input and output are the same
            oldTblPartLoc = tblObj.getPath();
            newTblPartLoc = tblObj.getPath();
            lbCtx = constructListBucketingCtx(tblObj.getSkewedColNames(), tblObj.getSkewedColValues(), tblObj.getSkewedColValueLocationMaps(), tblObj.isStoredAsSubDirectories(), conf);
        }
        // throw a HiveException for other than rcfile and orcfile.
        if (!((inputFormatClass.equals(RCFileInputFormat.class) || (inputFormatClass.equals(OrcInputFormat.class))))) {
            throw new SemanticException("Only RCFile and ORCFile Formats are supported right now.");
        }
        mergeDesc.setInputFormatClass(inputFormatClass);
        // throw a HiveException if the table/partition is bucketized
        if (bucketCols != null && bucketCols.size() > 0) {
            throw new SemanticException("Merge can not perform on bucketized partition/table.");
        }
        // throw a HiveException if the table/partition is archived
        if (isArchived) {
            throw new SemanticException("Merge can not perform on archived partitions.");
        }
        inputDir.add(oldTblPartLoc);
        mergeDesc.setInputDir(inputDir);
        mergeDesc.setLbCtx(lbCtx);
        addInputsOutputsAlterTable(tableName, partSpec, AlterTableTypes.MERGEFILES);
        DDLWork ddlWork = new DDLWork(getInputs(), getOutputs(), mergeDesc);
        ddlWork.setNeedLock(true);
        Task<? extends Serializable> mergeTask = TaskFactory.get(ddlWork, conf);
        TableDesc tblDesc = Utilities.getTableDesc(tblObj);
        Path queryTmpdir = ctx.getExternalTmpPath(newTblPartLoc);
        mergeDesc.setOutputDir(queryTmpdir);
        LoadTableDesc ltd = new LoadTableDesc(queryTmpdir, tblDesc, partSpec == null ? new HashMap<String, String>() : partSpec);
        ltd.setLbCtx(lbCtx);
        Task<MoveWork> moveTsk = TaskFactory.get(new MoveWork(null, null, ltd, null, false), conf);
        mergeTask.addDependentTask(moveTsk);
        if (conf.getBoolVar(HiveConf.ConfVars.HIVESTATSAUTOGATHER)) {
            StatsWork statDesc;
            if (oldTblPartLoc.equals(newTblPartLoc)) {
                // If we're merging to the same location, we can avoid some metastore calls
                TableSpec tablepart = new TableSpec(db, conf, tableName, partSpec);
                statDesc = new StatsWork(tablepart);
            } else {
                statDesc = new StatsWork(ltd);
            }
            statDesc.setNoStatsAggregator(true);
            statDesc.setClearAggregatorStats(true);
            statDesc.setStatsReliable(conf.getBoolVar(HiveConf.ConfVars.HIVE_STATS_RELIABLE));
            Task<? extends Serializable> statTask = TaskFactory.get(statDesc, conf);
            moveTsk.addDependentTask(statTask);
        }
        rootTasks.add(mergeTask);
    } catch (Exception e) {
        throw new SemanticException(e);
    }
}

Example 2 with ListBucketingCtx

use of org.apache.hadoop.hive.ql.plan.ListBucketingCtx in project hive by apache.

the class BaseSemanticAnalyzer method constructListBucketingCtx.

/**
   * Construct list bucketing context.
   *
   * @param skewedColNames
   * @param skewedValues
   * @param skewedColValueLocationMaps
   * @param isStoredAsSubDirectories
   * @return
   */
protected ListBucketingCtx constructListBucketingCtx(List<String> skewedColNames, List<List<String>> skewedValues, Map<List<String>, String> skewedColValueLocationMaps, boolean isStoredAsSubDirectories, HiveConf conf) {
    ListBucketingCtx lbCtx = new ListBucketingCtx();
    lbCtx.setSkewedColNames(skewedColNames);
    lbCtx.setSkewedColValues(skewedValues);
    lbCtx.setLbLocationMap(skewedColValueLocationMaps);
    lbCtx.setStoredAsSubDirectories(isStoredAsSubDirectories);
    lbCtx.setDefaultKey(ListBucketingPrunerUtils.HIVE_LIST_BUCKETING_DEFAULT_KEY);
    lbCtx.setDefaultDirName(ListBucketingPrunerUtils.HIVE_LIST_BUCKETING_DEFAULT_DIR_NAME);
    return lbCtx;
}

Example 3 with ListBucketingCtx

use of org.apache.hadoop.hive.ql.plan.ListBucketingCtx 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;
}

Example 4 with ListBucketingCtx

use of org.apache.hadoop.hive.ql.plan.ListBucketingCtx in project hive by apache.

the class DDLSemanticAnalyzer method analyzeTruncateTable.

private void analyzeTruncateTable(ASTNode ast) throws SemanticException {
    // TOK_TABLE_PARTITION
    ASTNode root = (ASTNode) ast.getChild(0);
    String tableName = getUnescapedName((ASTNode) root.getChild(0));
    Table table = getTable(tableName, true);
    if (table.getTableType() != TableType.MANAGED_TABLE) {
        throw new SemanticException(ErrorMsg.TRUNCATE_FOR_NON_MANAGED_TABLE.format(tableName));
    }
    if (table.isNonNative()) {
        //TODO
        throw new SemanticException(ErrorMsg.TRUNCATE_FOR_NON_NATIVE_TABLE.format(tableName));
    }
    if (!table.isPartitioned() && root.getChildCount() > 1) {
        throw new SemanticException(ErrorMsg.PARTSPEC_FOR_NON_PARTITIONED_TABLE.format(tableName));
    }
    Map<String, String> partSpec = getPartSpec((ASTNode) root.getChild(1));
    if (partSpec == null) {
        if (!table.isPartitioned()) {
            outputs.add(new WriteEntity(table, WriteEntity.WriteType.DDL_EXCLUSIVE));
        } else {
            for (Partition partition : getPartitions(table, null, false)) {
                outputs.add(new WriteEntity(partition, WriteEntity.WriteType.DDL_EXCLUSIVE));
            }
        }
    } else {
        if (isFullSpec(table, partSpec)) {
            validatePartSpec(table, partSpec, (ASTNode) root.getChild(1), conf, true);
            Partition partition = getPartition(table, partSpec, true);
            outputs.add(new WriteEntity(partition, WriteEntity.WriteType.DDL_EXCLUSIVE));
        } else {
            validatePartSpec(table, partSpec, (ASTNode) root.getChild(1), conf, false);
            for (Partition partition : getPartitions(table, partSpec, false)) {
                outputs.add(new WriteEntity(partition, WriteEntity.WriteType.DDL_EXCLUSIVE));
            }
        }
    }
    TruncateTableDesc truncateTblDesc = new TruncateTableDesc(tableName, partSpec);
    DDLWork ddlWork = new DDLWork(getInputs(), getOutputs(), truncateTblDesc);
    Task<? extends Serializable> truncateTask = TaskFactory.get(ddlWork, conf);
    // Is this a truncate column command
    List<String> columnNames = null;
    if (ast.getChildCount() == 2) {
        try {
            columnNames = getColumnNames((ASTNode) ast.getChild(1));
            // Throw an error if the table is indexed
            List<Index> indexes = db.getIndexes(table.getDbName(), tableName, (short) 1);
            if (indexes != null && indexes.size() > 0) {
                throw new SemanticException(ErrorMsg.TRUNCATE_COLUMN_INDEXED_TABLE.getMsg());
            }
            List<String> bucketCols = null;
            Class<? extends InputFormat> inputFormatClass = null;
            boolean isArchived = false;
            Path newTblPartLoc = null;
            Path oldTblPartLoc = null;
            List<FieldSchema> cols = null;
            ListBucketingCtx lbCtx = null;
            boolean isListBucketed = false;
            List<String> listBucketColNames = null;
            if (table.isPartitioned()) {
                Partition part = db.getPartition(table, partSpec, false);
                Path tabPath = table.getPath();
                Path partPath = part.getDataLocation();
                // if the table is in a different dfs than the partition,
                // replace the partition's dfs with the table's dfs.
                newTblPartLoc = new Path(tabPath.toUri().getScheme(), tabPath.toUri().getAuthority(), partPath.toUri().getPath());
                oldTblPartLoc = partPath;
                cols = part.getCols();
                bucketCols = part.getBucketCols();
                inputFormatClass = part.getInputFormatClass();
                isArchived = ArchiveUtils.isArchived(part);
                lbCtx = constructListBucketingCtx(part.getSkewedColNames(), part.getSkewedColValues(), part.getSkewedColValueLocationMaps(), part.isStoredAsSubDirectories(), conf);
                isListBucketed = part.isStoredAsSubDirectories();
                listBucketColNames = part.getSkewedColNames();
            } else {
                // input and output are the same
                oldTblPartLoc = table.getPath();
                newTblPartLoc = table.getPath();
                cols = table.getCols();
                bucketCols = table.getBucketCols();
                inputFormatClass = table.getInputFormatClass();
                lbCtx = constructListBucketingCtx(table.getSkewedColNames(), table.getSkewedColValues(), table.getSkewedColValueLocationMaps(), table.isStoredAsSubDirectories(), conf);
                isListBucketed = table.isStoredAsSubDirectories();
                listBucketColNames = table.getSkewedColNames();
            }
            // throw a HiveException for non-rcfile.
            if (!inputFormatClass.equals(RCFileInputFormat.class)) {
                throw new SemanticException(ErrorMsg.TRUNCATE_COLUMN_NOT_RC.getMsg());
            }
            // throw a HiveException if the table/partition is archived
            if (isArchived) {
                throw new SemanticException(ErrorMsg.TRUNCATE_COLUMN_ARCHIVED.getMsg());
            }
            Set<Integer> columnIndexes = new HashSet<Integer>();
            for (String columnName : columnNames) {
                boolean found = false;
                for (int columnIndex = 0; columnIndex < cols.size(); columnIndex++) {
                    if (columnName.equalsIgnoreCase(cols.get(columnIndex).getName())) {
                        columnIndexes.add(columnIndex);
                        found = true;
                        break;
                    }
                }
                // Throw an exception if the user is trying to truncate a column which doesn't exist
                if (!found) {
                    throw new SemanticException(ErrorMsg.INVALID_COLUMN.getMsg(columnName));
                }
                // Throw an exception if the table/partition is bucketed on one of the columns
                for (String bucketCol : bucketCols) {
                    if (bucketCol.equalsIgnoreCase(columnName)) {
                        throw new SemanticException(ErrorMsg.TRUNCATE_BUCKETED_COLUMN.getMsg(columnName));
                    }
                }
                if (isListBucketed) {
                    for (String listBucketCol : listBucketColNames) {
                        if (listBucketCol.equalsIgnoreCase(columnName)) {
                            throw new SemanticException(ErrorMsg.TRUNCATE_LIST_BUCKETED_COLUMN.getMsg(columnName));
                        }
                    }
                }
            }
            truncateTblDesc.setColumnIndexes(new ArrayList<Integer>(columnIndexes));
            truncateTblDesc.setInputDir(oldTblPartLoc);
            truncateTblDesc.setLbCtx(lbCtx);
            addInputsOutputsAlterTable(tableName, partSpec, AlterTableTypes.TRUNCATE);
            ddlWork.setNeedLock(true);
            TableDesc tblDesc = Utilities.getTableDesc(table);
            // Write the output to temporary directory and move it to the final location at the end
            // so the operation is atomic.
            Path queryTmpdir = ctx.getExternalTmpPath(newTblPartLoc);
            truncateTblDesc.setOutputDir(queryTmpdir);
            LoadTableDesc ltd = new LoadTableDesc(queryTmpdir, tblDesc, partSpec == null ? new HashMap<String, String>() : partSpec);
            ltd.setLbCtx(lbCtx);
            Task<MoveWork> moveTsk = TaskFactory.get(new MoveWork(null, null, ltd, null, false), conf);
            truncateTask.addDependentTask(moveTsk);
            // Recalculate the HDFS stats if auto gather stats is set
            if (conf.getBoolVar(HiveConf.ConfVars.HIVESTATSAUTOGATHER)) {
                StatsWork statDesc;
                if (oldTblPartLoc.equals(newTblPartLoc)) {
                    // If we're merging to the same location, we can avoid some metastore calls
                    TableSpec tablepart = new TableSpec(this.db, conf, root);
                    statDesc = new StatsWork(tablepart);
                } else {
                    statDesc = new StatsWork(ltd);
                }
                statDesc.setNoStatsAggregator(true);
                statDesc.setClearAggregatorStats(true);
                statDesc.setStatsReliable(conf.getBoolVar(HiveConf.ConfVars.HIVE_STATS_RELIABLE));
                Task<? extends Serializable> statTask = TaskFactory.get(statDesc, conf);
                moveTsk.addDependentTask(statTask);
            }
        } catch (HiveException e) {
            throw new SemanticException(e);
        }
    }
    rootTasks.add(truncateTask);
}

Example 5 with ListBucketingCtx

use of org.apache.hadoop.hive.ql.plan.ListBucketingCtx in project hive by apache.

the class DDLTask method mergeFiles.

/**
   * First, make sure the source table/partition is not
   * archived/indexes/non-rcfile. If either of these is true, throw an
   * exception.
   *
   * The way how it does the merge is to create a BlockMergeTask from the
   * mergeFilesDesc.
   *
   * @param db
   * @param mergeFilesDesc
   * @return
   * @throws HiveException
   */
private int mergeFiles(Hive db, AlterTablePartMergeFilesDesc mergeFilesDesc, DriverContext driverContext) throws HiveException {
    ListBucketingCtx lbCtx = mergeFilesDesc.getLbCtx();
    boolean lbatc = lbCtx == null ? false : lbCtx.isSkewedStoredAsDir();
    int lbd = lbCtx == null ? 0 : lbCtx.calculateListBucketingLevel();
    // merge work only needs input and output.
    MergeFileWork mergeWork = new MergeFileWork(mergeFilesDesc.getInputDir(), mergeFilesDesc.getOutputDir(), mergeFilesDesc.getInputFormatClass().getName());
    LinkedHashMap<Path, ArrayList<String>> pathToAliases = new LinkedHashMap<>();
    ArrayList<String> inputDirstr = new ArrayList<String>(1);
    inputDirstr.add(mergeFilesDesc.getInputDir().toString());
    pathToAliases.put(mergeFilesDesc.getInputDir().get(0), inputDirstr);
    mergeWork.setPathToAliases(pathToAliases);
    mergeWork.setListBucketingCtx(mergeFilesDesc.getLbCtx());
    mergeWork.resolveConcatenateMerge(db.getConf());
    mergeWork.setMapperCannotSpanPartns(true);
    mergeWork.setSourceTableInputFormat(mergeFilesDesc.getInputFormatClass().getName());
    final FileMergeDesc fmd;
    if (mergeFilesDesc.getInputFormatClass().equals(RCFileInputFormat.class)) {
        fmd = new RCFileMergeDesc();
    } else {
        // safe to assume else is ORC as semantic analyzer will check for RC/ORC
        fmd = new OrcFileMergeDesc();
    }
    fmd.setDpCtx(null);
    fmd.setHasDynamicPartitions(false);
    fmd.setListBucketingAlterTableConcatenate(lbatc);
    fmd.setListBucketingDepth(lbd);
    fmd.setOutputPath(mergeFilesDesc.getOutputDir());
    CompilationOpContext opContext = driverContext.getCtx().getOpContext();
    Operator<? extends OperatorDesc> mergeOp = OperatorFactory.get(opContext, fmd);
    LinkedHashMap<String, Operator<? extends OperatorDesc>> aliasToWork = new LinkedHashMap<String, Operator<? extends OperatorDesc>>();
    aliasToWork.put(mergeFilesDesc.getInputDir().toString(), mergeOp);
    mergeWork.setAliasToWork(aliasToWork);
    DriverContext driverCxt = new DriverContext();
    Task task;
    if (conf.getVar(ConfVars.HIVE_EXECUTION_ENGINE).equals("tez")) {
        TezWork tezWork = new TezWork(conf.getVar(HiveConf.ConfVars.HIVEQUERYID), conf);
        mergeWork.setName("File Merge");
        tezWork.add(mergeWork);
        task = new TezTask();
        task.setWork(tezWork);
    } else {
        task = new MergeFileTask();
        task.setWork(mergeWork);
    }
    // initialize the task and execute
    task.initialize(queryState, getQueryPlan(), driverCxt, opContext);
    subtask = task;
    int ret = task.execute(driverCxt);
    if (subtask.getException() != null) {
        setException(subtask.getException());
    }
    return ret;
}