Package org.apache.hadoop.hive.ql.plan

Examples of org.apache.hadoop.hive.ql.plan.TableDesc


      }
    }
  }

  public static TableDesc getTableDesc(Table tbl) {
    return (new TableDesc(tbl.getDeserializer().getClass(), tbl.getInputFormatClass(), tbl
        .getOutputFormatClass(), tbl.getSchema()));
  }
View Full Code Here


        .getOutputFormatClass(), tbl.getSchema()));
  }

  // column names and column types are all delimited by comma
  public static TableDesc getTableDesc(String cols, String colTypes) {
    return (new TableDesc(LazySimpleSerDe.class, SequenceFileInputFormat.class,
        HiveSequenceFileOutputFormat.class, Utilities.makeProperties(
            org.apache.hadoop.hive.serde.Constants.SERIALIZATION_FORMAT, "" + Utilities.ctrlaCode,
            org.apache.hadoop.hive.serde.Constants.LIST_COLUMNS, cols,
            org.apache.hadoop.hive.serde.Constants.LIST_COLUMN_TYPES, colTypes)));
  }
View Full Code Here

    Operator<ExtractDesc> extract = OperatorFactory.getAndMakeChild(new ExtractDesc(
        new ExprNodeColumnDesc(TypeInfoFactory.stringTypeInfo,
            Utilities.ReduceField.VALUE.toString(), "", false)),
            new RowSchema(out_rwsch.getColumnInfos()));

    TableDesc ts = (TableDesc) fsConf.getTableInfo().clone();
    fsConf.getTableInfo().getProperties().remove(
        org.apache.hadoop.hive.metastore.api.Constants.META_TABLE_PARTITION_COLUMNS);

    FileSinkDesc newFSD = new FileSinkDesc(finalName, ts, parseCtx.getConf()
        .getBoolVar(HiveConf.ConfVars.COMPRESSRESULT));
View Full Code Here

    // Create a TableScan operator
    RowSchema inputRS = fsInput.getSchema();
    Operator<? extends Serializable> tsMerge = OperatorFactory.get(TableScanDesc.class, inputRS);

    // Create a FileSink operator
    TableDesc ts = (TableDesc) fsInputDesc.getTableInfo().clone();
    FileSinkDesc fsOutputDesc =  new FileSinkDesc(finalName, ts,
        parseCtx.getConf().getBoolVar(HiveConf.ConfVars.COMPRESSRESULT));
    FileSinkOperator fsOutput = (FileSinkOperator) OperatorFactory.getAndMakeChild(
        fsOutputDesc,  inputRS, tsMerge);
View Full Code Here

  private MapredWork createMergeTask(HiveConf conf, Operator<? extends Serializable> topOp,
      FileSinkDesc fsDesc) {

    ArrayList<String> aliases = new ArrayList<String>();
    String inputDir = fsDesc.getDirName();
    TableDesc tblDesc = fsDesc.getTableInfo();
    aliases.add(inputDir); // dummy alias: just use the input path

    // constructing the default MapredWork
    MapredWork cplan = GenMapRedUtils.getMapRedWork(conf);
    cplan.getPathToAliases().put(inputDir, aliases);
View Full Code Here

      opTaskMap.put(null, currTask);
      GenMRMapJoinCtx mjCtx = ctx.getMapJoinCtx(currMapJoinOp);
      MapredWork plan = (MapredWork) currTask.getWork();

      String taskTmpDir = mjCtx.getTaskTmpDir();
      TableDesc tt_desc = mjCtx.getTTDesc();
      assert plan.getPathToAliases().get(taskTmpDir) == null;
      plan.getPathToAliases().put(taskTmpDir, new ArrayList<String>());
      plan.getPathToAliases().get(taskTmpDir).add(taskTmpDir);
      plan.getPathToPartitionInfo().put(taskTmpDir,
          new PartitionDesc(tt_desc, null));
View Full Code Here

    } else if (sem.getResultSchema() != null) {
      List<FieldSchema> lst = sem.getResultSchema();
      schema = new Schema(lst, null);
    } else if (sem.getFetchTask() != null) {
      FetchTask ft = sem.getFetchTask();
      TableDesc td = ft.getTblDesc();
      // partitioned tables don't have tableDesc set on the FetchTask. Instead
      // they have a list of PartitionDesc objects, each with a table desc.
      // Let's
      // try to fetch the desc for the first partition and use it's
      // deserializer.
      if (td == null && ft.getWork() != null && ft.getWork().getPartDesc() != null) {
        if (ft.getWork().getPartDesc().size() > 0) {
          td = ft.getWork().getPartDesc().get(0).getTableDesc();
        }
      }

      if (td == null) {
        LOG.info("No returning schema.");
      } else {
        String tableName = "result";
        List<FieldSchema> lst = null;
        try {
          lst = MetaStoreUtils.getFieldsFromDeserializer(tableName, td.getDeserializer());
        } catch (Exception e) {
          LOG.warn("Error getting schema: "
              + org.apache.hadoop.util.StringUtils.stringifyException(e));
        }
        if (lst != null) {
View Full Code Here

            true, JavaUtils.getClassLoader());
      } catch (ClassNotFoundException e) {
        throw new SemanticException(e);
      }

      TableDesc tblDesc = PlanUtils.getTableDesc(serdeClass, Integer
          .toString(Utilities.tabCode), cols, colTypes, defaultCols);
      // copy all the properties
      if (child.getChildCount() == 2) {
        ASTNode prop = (ASTNode) ((ASTNode) child.getChild(1)).getChild(0);
        for (int propChild = 0; propChild < prop.getChildCount(); propChild++) {
          String key = unescapeSQLString(prop.getChild(propChild).getChild(0)
              .getText());
          String value = unescapeSQLString(prop.getChild(propChild).getChild(1)
              .getText());
          tblDesc.getProperties().setProperty(key, value);
        }
      }
      return tblDesc;
    } else if (child.getType() == HiveParser.TOK_SERDEPROPS) {
      TableDesc tblDesc = PlanUtils.getDefaultTableDesc(Integer
          .toString(Utilities.ctrlaCode), cols, colTypes, defaultCols);
      int numChildRowFormat = child.getChildCount();
      for (int numC = 0; numC < numChildRowFormat; numC++) {
        ASTNode rowChild = (ASTNode) child.getChild(numC);
        switch (rowChild.getToken().getType()) {
        case HiveParser.TOK_TABLEROWFORMATFIELD:
          String fieldDelim = unescapeSQLString(rowChild.getChild(0).getText());
          tblDesc.getProperties()
              .setProperty(Constants.FIELD_DELIM, fieldDelim);
          tblDesc.getProperties().setProperty(Constants.SERIALIZATION_FORMAT,
              fieldDelim);

          if (rowChild.getChildCount() >= 2) {
            String fieldEscape = unescapeSQLString(rowChild.getChild(1)
                .getText());
            tblDesc.getProperties().setProperty(Constants.ESCAPE_CHAR,
                fieldEscape);
          }
          break;
        case HiveParser.TOK_TABLEROWFORMATCOLLITEMS:
          tblDesc.getProperties().setProperty(Constants.COLLECTION_DELIM,
              unescapeSQLString(rowChild.getChild(0).getText()));
          break;
        case HiveParser.TOK_TABLEROWFORMATMAPKEYS:
          tblDesc.getProperties().setProperty(Constants.MAPKEY_DELIM,
              unescapeSQLString(rowChild.getChild(0).getText()));
          break;
        case HiveParser.TOK_TABLEROWFORMATLINES:
          String lineDelim = unescapeSQLString(rowChild.getChild(0).getText());
          tblDesc.getProperties().setProperty(Constants.LINE_DELIM, lineDelim);
          if (!lineDelim.equals("\n") && !lineDelim.equals("10")) {
            throw new SemanticException(
                ErrorMsg.LINES_TERMINATED_BY_NON_NEWLINE.getMsg());
          }
          break;
View Full Code Here

      inpColumns.append(inputSchema.get(i).getInternalName());
      inpColumnTypes.append(inputSchema.get(i).getType().getTypeName());
    }

    TableDesc outInfo;
    TableDesc errInfo;
    TableDesc inInfo;
    String defaultSerdeName = conf.getVar(HiveConf.ConfVars.HIVESCRIPTSERDE);
    Class<? extends Deserializer> serde;

    try {
      serde = (Class<? extends Deserializer>) Class.forName(defaultSerdeName,
View Full Code Here

    Table dest_tab = null;     // destination table if any
    Partition dest_part = null;// destination partition if any
    String queryTmpdir = null; // the intermediate destination directory
    Path dest_path = null; // the final destination directory
    TableDesc table_desc = null;
    int currentTableId = 0;
    boolean isLocal = false;
    SortBucketRSCtx rsCtx = new SortBucketRSCtx();
    DynamicPartitionCtx dpCtx = null;
    LoadTableDesc ltd = null;
    boolean holdDDLTime = checkHoldDDLTime(qb);

    switch (dest_type.intValue()) {
    case QBMetaData.DEST_TABLE: {

      dest_tab = qbm.getDestTableForAlias(dest);

      Map<String, String> partSpec = qbm.getPartSpecForAlias(dest);
      dest_path = dest_tab.getPath();

      // 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(ErrorMsg.NEED_PARTITION_ERROR.getMsg());
        }
        // the HOLD_DDLTIIME hint should not be used with dynamic partition since the
        // newly generated partitions should always update their DDLTIME
        if (holdDDLTime) {
          throw new SemanticException(ErrorMsg.HOLD_DDLTIME_ON_NONEXIST_PARTITIONS.getMsg());
        }
        dpCtx = qbm.getDPCtx(dest);
        if (dpCtx == null) {
          Utilities.validatePartSpec(dest_tab, partSpec);
          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
          if (dpCtx.getNumDPCols() > 0 &&
              (HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVEMERGEMAPFILES) ||
               HiveConf.getBoolVar(conf, HiveConf.ConfVars.HIVEMERGEMAPREDFILES)) &&
               Utilities.supportCombineFileInputFormat() == false) {
            // Do not support merge for Hadoop versions (pre-0.20) that do not
            // support CombineHiveInputFormat
            HiveConf.setBoolVar(conf, HiveConf.ConfVars.HIVEMERGEMAPFILES, false);
            HiveConf.setBoolVar(conf, HiveConf.ConfVars.HIVEMERGEMAPREDFILES, false);
          }
          // turn on hive.task.progress to update # of partitions created to the JT
          HiveConf.setBoolVar(conf, HiveConf.ConfVars.HIVEJOBPROGRESS, true);

        } else { // QBMetaData.DEST_PARTITION capture the all-SP case
          throw new SemanticException(ErrorMsg.DYNAMIC_PARTITION_DISABLED.getMsg());
        }
        if (dpCtx.getSPPath() != null) {
          dest_path = new Path(dest_tab.getPath(), dpCtx.getSPPath());
        }
        if ((dest_tab.getNumBuckets() > 0) &&
            (conf.getBoolVar(HiveConf.ConfVars.HIVEENFORCEBUCKETING))) {
          dpCtx.setNumBuckets(dest_tab.getNumBuckets());
        }
      }

      boolean isNonNativeTable = dest_tab.isNonNative();
      if (isNonNativeTable) {
        queryTmpdir = dest_path.toUri().getPath();
      } else {
        queryTmpdir = ctx.getExternalTmpFileURI(dest_path.toUri());
      }
      if (dpCtx != null) {
        // set the root of the temporay 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++;

      // Create the work for moving the table
      // NOTE: specify Dynamic partitions in dest_tab for WriteEntity
      if (!isNonNativeTable) {
        ltd = new LoadTableDesc(queryTmpdir, ctx.getExternalTmpFileURI(dest_path.toUri()),
            table_desc, dpCtx);
        if (holdDDLTime) {
          LOG.info("this query will not update transient_lastDdlTime!");
          ltd.setHoldDDLTime(true);
        }
        loadTableWork.add(ltd);
      }

      // Here only register the whole table for post-exec hook if no DP present
      // in the case of DP, we will register WriteEntity in MoveTask when the
      // list of dynamically created partitions are known.
      if ((dpCtx == null || dpCtx.getNumDPCols() == 0) &&
          !outputs.add(new WriteEntity(dest_tab))) {
        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) {
          outputs.add(new WriteEntity(dest_tab, false));
        }
        // 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
        else {
          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);
            outputs.add(new WriteEntity(p, false));
          } catch (HiveException e) {
            throw new SemanticException(e.getMessage());
          }
        }
      }

      break;
    }
    case QBMetaData.DEST_PARTITION: {

      dest_part = qbm.getDestPartitionForAlias(dest);
      dest_tab = dest_part.getTable();
      Path tabPath = dest_tab.getPath();
      Path partPath = dest_part.getPartitionPath();

        // 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());

      if ("har".equalsIgnoreCase(dest_path.toUri().getScheme())) {
        throw new SemanticException(ErrorMsg.OVERWRITE_ARCHIVED_PART
            .getMsg());
      }
      queryTmpdir = ctx.getExternalTmpFileURI(dest_path.toUri());
      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++;

      ltd = new LoadTableDesc(queryTmpdir, ctx.getExternalTmpFileURI(dest_path.toUri()),
          table_desc, dest_part.getSpec());
      if (holdDDLTime) {
        try {
          Partition part = db.getPartition(dest_tab, dest_part.getSpec(), false);
          if (part == null) {
            throw new SemanticException(ErrorMsg.HOLD_DDLTIME_ON_NONEXIST_PARTITIONS.getMsg());
          }
        } catch (HiveException e) {
          throw new SemanticException(e);
        }
        LOG.info("this query will not update transient_lastDdlTime!");
        ltd.setHoldDDLTime(true);
      }
      loadTableWork.add(ltd);
      if (!outputs.add(new WriteEntity(dest_part))) {
        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));
      String destStr = dest_path.toString();

      if (isLocal) {
        // for local directory - we always write to map-red intermediate
        // store and then copy to local fs
        queryTmpdir = ctx.getMRTmpFileURI();
      } else {
        // otherwise write to the file system implied by the directory
        // no copy is required. we may want to revisit this policy in future

        try {
          Path qPath = FileUtils.makeQualified(dest_path, conf);
          queryTmpdir = ctx.getExternalTmpFileURI(qPath.toUri());
        } catch (Exception e) {
          throw new SemanticException("Error creating temporary folder on: "
              + dest_path, e);
        }
      }
      String cols = new String();
      String colTypes = new String();
      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();
      if (tblDesc != null) {
        field_schemas = new ArrayList<FieldSchema>();
      }

      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]);
        }

        if (field_schemas != null) {
          FieldSchema col = new FieldSchema();
          if (nm[1] != null) {
            col.setName(unescapeIdentifier(colInfo.getAlias()).toLowerCase()); // remove ``
          } else {
            col.setName(colInfo.getInternalName());
          }
          col.setType(colInfo.getType().getTypeName());
          field_schemas.add(col);
        }

        if (!first) {
          cols = cols.concat(",");
          colTypes = colTypes.concat(":");
        }

        first = false;
        cols = cols.concat(colInfo.getInternalName());

        // 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(Constants.VOID_TYPE_NAME)) {
          colTypes = colTypes.concat(Constants.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));
      }

      if (!ctx.isMRTmpFileURI(destStr)) {
        idToTableNameMap.put(String.valueOf(destTableId), destStr);
        currentTableId = destTableId;
        destTableId++;
      }

      boolean isDfsDir = (dest_type.intValue() == QBMetaData.DEST_DFS_FILE);
      loadFileWork.add(new LoadFileDesc(queryTmpdir, destStr, isDfsDir, cols,
          colTypes));

      if (tblDesc == null) {
        if (qb.getIsQuery()) {
          String fileFormat = HiveConf.getVar(conf, HiveConf.ConfVars.HIVEQUERYRESULTFILEFORMAT);
          table_desc = PlanUtils.getDefaultQueryOutputTableDesc(cols, colTypes, fileFormat);
        } else {
          table_desc = PlanUtils.getDefaultTableDesc(Integer
              .toString(Utilities.ctrlaCode), cols, colTypes, false);
        }
      } else {
        table_desc = PlanUtils.getTableDesc(tblDesc, cols, colTypes);
      }

      if (!outputs.add(new WriteEntity(destStr, !isDfsDir))) {
        throw new SemanticException(ErrorMsg.OUTPUT_SPECIFIED_MULTIPLE_TIMES
            .getMsg(destStr));
      }
      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>();

    try {
      StructObjectInspector rowObjectInspector = (StructObjectInspector) table_desc
          .getDeserializer().getObjectInspector();
      List<? extends StructField> fields = rowObjectInspector
          .getAllStructFieldRefs();
      for (int i = 0; i < fields.size(); i++) {
        vecCol.add(new ColumnInfo(fields.get(i).getFieldName(), TypeInfoUtils
View Full Code Here

TOP

Related Classes of org.apache.hadoop.hive.ql.plan.TableDesc

Copyright © 2018 www.massapicom. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.