Package org.apache.hadoop.hive.ql.optimizer.GenMRProcContext

Examples of org.apache.hadoop.hive.ql.optimizer.GenMRProcContext.GenMapRedCtx

64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
      return (new GenMRRedSink1()).process(nd, stack, opProcCtx, nodeOutputs);

    // union consisted on a bunch of map-reduce jobs, and it has been split at the union
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask    = mapredCtx.getCurrTask();
    mapredWork plan = (mapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx.getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);
   
    ctx.setCurrTask(currTask);

    // If the plan for this reducer does not exist, initialize the plan
    if (opMapTask == null) {
      // When the reducer is encountered for the first time
      if (plan.getReducer() == null)
        GenMapRedUtils.initUnionPlan(op, ctx);
      // When union is followed by a multi-table insert
      else
        GenMapRedUtils.splitPlan(op, ctx);
    }
    // The union is already initialized. However, the union is walked from another input
    // initUnionPlan is idempotent
    else if (plan.getReducer() == reducer)
      GenMapRedUtils.initUnionPlan(op, ctx);
    // There is a join after union. One of the branches of union has already been initialized.
    // Initialize the current branch, and join with the original plan.
    else {
      GenMapRedUtils.initUnionPlan(ctx, currTask, false);
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, true, false, false);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
   
    // the union operator has been processed
    ctx.setCurrUnionOp(null);
    return null;
  }
View Full Code Here
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
    // Map-only subqueries can be optimized in future to not write to a file in future
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx.getMapCurrCtx();

    // The plan needs to be broken only if one of the sub-queries involve a map-reduce job
    if (uCtx.isMapOnlySubq()) {
      mapCurrCtx.put((Operator<? extends Serializable>)nd, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
      return null;
    }

    ctx.setCurrUnionOp(union);

    UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
    assert uPrsCtx != null;

    Task<? extends Serializable> currTask = ctx.getCurrTask();
    int pos = UnionProcFactory.getPositionParent(union, stack);

    // is the current task a root task
    if (uPrsCtx.getRootTask(pos) && (!ctx.getRootTasks().contains(currTask)))
      ctx.getRootTasks().add(currTask);
   
    GenMRUnionCtx uCtxTask = ctx.getUnionTask(union);
    Task<? extends Serializable> uTask = null;

    Operator<? extends Serializable> parent = union.getParentOperators().get(pos);  
    mapredWork uPlan = null;

    // union is encountered for the first time
    if (uCtxTask == null) {
      uCtxTask = new GenMRUnionCtx();
      uPlan = GenMapRedUtils.getMapRedWork();
      uTask = TaskFactory.get(uPlan, parseCtx.getConf());
      uCtxTask.setUTask(uTask);
      ctx.setUnionTask(union, uCtxTask);
    }
    else {
      uTask = uCtxTask.getUTask();
      uPlan = (mapredWork)uTask.getWork();
    }

    tableDesc tt_desc = PlanUtils.getIntermediateFileTableDesc(
          PlanUtils.getFieldSchemasFromRowSchema(parent.getSchema(), "temporarycol"));
   
    // generate the temporary file
    Context baseCtx = parseCtx.getContext();
    String taskTmpDir = baseCtx.getMRTmpFileURI();
   
    // Add the path to alias mapping
    uCtxTask.addTaskTmpDir(taskTmpDir);
    uCtxTask.addTTDesc(tt_desc);

    // The union task is empty. The files created for all the inputs are assembled in the
    // union context and later used to initialize the union plan
   
    // Create a file sink operator for this file name
    Operator<? extends Serializable> fs_op =
      OperatorFactory.get
      (new fileSinkDesc(taskTmpDir, tt_desc,
                        parseCtx.getConf().getBoolVar(HiveConf.ConfVars.COMPRESSINTERMEDIATE)),
       parent.getSchema());
   
    assert parent.getChildOperators().size() == 1;
    parent.getChildOperators().set(0, fs_op);

    List<Operator<? extends Serializable>> parentOpList = new ArrayList<Operator<? extends Serializable>>();
    parentOpList.add(parent);
    fs_op.setParentOperators(parentOpList);

    currTask.addDependentTask(uTask);

    // If it is map-only task, add the files to be processed
    if (uPrsCtx.getMapOnlySubq(pos) && uPrsCtx.getRootTask(pos))
      GenMapRedUtils.setTaskPlan(ctx.getCurrAliasId(), ctx.getCurrTopOp(), (mapredWork) currTask.getWork(), false, ctx);

    ctx.setCurrTask(uTask);
    ctx.setCurrAliasId(null);
    ctx.setCurrTopOp(null);

    mapCurrCtx.put((Operator<? extends Serializable>)nd, new GenMapRedCtx(ctx.getCurrTask(), null, null));
   
    return null;
  }
View Full Code Here
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
    ParseContext parseCtx = ctx.getParseCtx();

    // map-join consisted on a bunch of map-only jobs, and it has been split after the mapjoin
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask    = mapredCtx.getCurrTask();
    mapredWork plan = (mapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx.getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);
   
    ctx.setCurrTask(currTask);

    // If the plan for this reducer does not exist, initialize the plan
    if (opMapTask == null) {
      // When the reducer is encountered for the first time
      if (plan.getReducer() == null)
        GenMapRedUtils.initMapJoinPlan(op, ctx, true, false, true, -1);
      // When mapjoin is followed by a multi-table insert
      else
        GenMapRedUtils.splitPlan(op, ctx);
    }
    // There is a join after mapjoin. One of the branches of mapjoin has already been initialized.
    // Initialize the current branch, and join with the original plan.
    else {
      assert plan.getReducer() != reducer;
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, false, true, false);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
   
    // the mapjoin operator has been processed
    ctx.setCurrMapJoinOp(null);
    return null;
  }
View Full Code Here
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
  public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx opProcCtx, Object... nodeOutputs) throws SemanticException {
    ReduceSinkOperator op = (ReduceSinkOperator)nd;
    GenMRProcContext ctx = (GenMRProcContext)opProcCtx;

    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx.getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask      = mapredCtx.getCurrTask();
    Operator<? extends Serializable> currTopOp = mapredCtx.getCurrTopOp();
    String currAliasId = mapredCtx.getCurrAliasId();
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx.getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);

    ctx.setCurrTopOp(currTopOp);
    ctx.setCurrAliasId(currAliasId);
    ctx.setCurrTask(currTask);

    if (opMapTask == null)
      GenMapRedUtils.splitPlan(op, ctx);
    else {
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, true, false, false);
      currTask = opMapTask;
      ctx.setCurrTask(currTask);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
    return null;
  }
View Full Code Here
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
    // union consisted on a bunch of map-reduce jobs, and it has been split at
    // the union
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork plan = (MapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx
        .getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);

    ctx.setCurrTask(currTask);

    // If the plan for this reducer does not exist, initialize the plan
    if (opMapTask == null) {
      // When the reducer is encountered for the first time
      if (plan.getReducer() == null) {
        GenMapRedUtils.initUnionPlan(op, ctx);
        // When union is followed by a multi-table insert
      } else {
        GenMapRedUtils.splitPlan(op, ctx);
      }
    } else if (plan.getReducer() == reducer) {
      // The union is already initialized. However, the union is walked from
      // another input
      // initUnionPlan is idempotent
      GenMapRedUtils.initUnionPlan(op, ctx);
    } else {
      GenMapRedUtils.initUnionPlan(ctx, currTask, false);
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, true, false,
          false);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(),
        ctx.getCurrAliasId()));

    // the union operator has been processed
    ctx.setCurrUnionOp(null);
    return null;
View Full Code Here
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
      UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
      if ((uPrsCtx != null) && (uPrsCtx.getMapJoinQuery())) {
        GenMapRedUtils.mergeMapJoinUnion(union, ctx, UnionProcFactory
            .getPositionParent(union, stack));
      } else {
        mapCurrCtx.put((Operator<? extends Serializable>) nd, new GenMapRedCtx(
            ctx.getCurrTask(), ctx.getCurrTopOp(), ctx.getCurrAliasId()));
      }
      return null;
    }

    ctx.setCurrUnionOp(union);

    UnionParseContext uPrsCtx = uCtx.getUnionParseContext(union);
    assert uPrsCtx != null;

    Task<? extends Serializable> currTask = ctx.getCurrTask();
    int pos = UnionProcFactory.getPositionParent(union, stack);

    // is the current task a root task
    if (uPrsCtx.getRootTask(pos) && (!ctx.getRootTasks().contains(currTask))) {
      ctx.getRootTasks().add(currTask);
    }

    GenMRUnionCtx uCtxTask = ctx.getUnionTask(union);
    Task<? extends Serializable> uTask = null;

    Operator<? extends Serializable> parent = union.getParentOperators().get(
        pos);
    MapredWork uPlan = null;

    // union is encountered for the first time
    if (uCtxTask == null) {
      uCtxTask = new GenMRUnionCtx();
      uPlan = GenMapRedUtils.getMapRedWork(parseCtx);
      uTask = TaskFactory.get(uPlan, parseCtx.getConf());
      uCtxTask.setUTask(uTask);
      ctx.setUnionTask(union, uCtxTask);
    } else {
      uTask = uCtxTask.getUTask();
    }

    // If there is a mapjoin at position 'pos'
    if (uPrsCtx.getMapJoinSubq(pos)) {
      AbstractMapJoinOperator<? extends MapJoinDesc> mjOp = ctx.getCurrMapJoinOp();
      assert mjOp != null;
      GenMRMapJoinCtx mjCtx = ctx.getMapJoinCtx(mjOp);
      assert mjCtx != null;
      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));
      plan.getAliasToWork().put(taskTmpDir, mjCtx.getRootMapJoinOp());
    }

    TableDesc tt_desc = PlanUtils.getIntermediateFileTableDesc(PlanUtils
        .getFieldSchemasFromRowSchema(parent.getSchema(), "temporarycol"));

    // generate the temporary file
    Context baseCtx = parseCtx.getContext();
    String taskTmpDir = baseCtx.getMRTmpFileURI();

    // Add the path to alias mapping
    uCtxTask.addTaskTmpDir(taskTmpDir);
    uCtxTask.addTTDesc(tt_desc);

    // The union task is empty. The files created for all the inputs are
    // assembled in the
    // union context and later used to initialize the union plan

    // Create a file sink operator for this file name
    Operator<? extends Serializable> fs_op = OperatorFactory.get(
        new FileSinkDesc(taskTmpDir, tt_desc, parseCtx.getConf().getBoolVar(
        HiveConf.ConfVars.COMPRESSINTERMEDIATE)), parent.getSchema());

    assert parent.getChildOperators().size() == 1;
    parent.getChildOperators().set(0, fs_op);

    List<Operator<? extends Serializable>> parentOpList =
      new ArrayList<Operator<? extends Serializable>>();
    parentOpList.add(parent);
    fs_op.setParentOperators(parentOpList);

    currTask.addDependentTask(uTask);

    // If it is map-only task, add the files to be processed
    if (uPrsCtx.getMapOnlySubq(pos) && uPrsCtx.getRootTask(pos)) {
      GenMapRedUtils.setTaskPlan(ctx.getCurrAliasId(), ctx.getCurrTopOp(),
          (MapredWork) currTask.getWork(), false, ctx);
    }

    ctx.setCurrTask(uTask);
    ctx.setCurrAliasId(null);
    ctx.setCurrTopOp(null);

    mapCurrCtx.put((Operator<? extends Serializable>) nd, new GenMapRedCtx(ctx
        .getCurrTask(), null, null));

    return null;
  }
View Full Code Here
49
50
51
52
53
54
55
56
57
58
59
      Object... nodeOutputs) throws SemanticException {
    GenMRProcContext ctx = (GenMRProcContext) procCtx;

    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(stack.get(stack.size() - 2));
    mapCurrCtx.put((Operator<? extends Serializable>) nd, new GenMapRedCtx(
        mapredCtx.getCurrTask(), mapredCtx.getCurrTopOp(), mapredCtx
        .getCurrAliasId()));
    return null;
  }
View Full Code Here
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
    // map-join consisted on a bunch of map-only jobs, and it has been split
    // after the mapjoin
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork plan = (MapredWork) currTask.getWork();
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx
        .getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);

    ctx.setCurrTask(currTask);

    // If the plan for this reducer does not exist, initialize the plan
    if (opMapTask == null) {
      // When the reducer is encountered for the first time
      if (plan.getReducer() == null) {
        GenMapRedUtils.initMapJoinPlan(op, ctx, true, false, true, -1);
        // When mapjoin is followed by a multi-table insert
      } else {
        GenMapRedUtils.splitPlan(op, ctx);
      }
    } else {
      // There is a join after mapjoin. One of the branches of mapjoin has already
      // been initialized.
      // Initialize the current branch, and join with the original plan.
      assert plan.getReducer() != reducer;
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, false, true,
          false);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(),
        ctx.getCurrAliasId()));

    // the mapjoin operator has been processed
    ctx.setCurrMapJoinOp(null);
    return null;
View Full Code Here
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
    ReduceSinkOperator op = (ReduceSinkOperator) nd;
    GenMRProcContext ctx = (GenMRProcContext) opProcCtx;

    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(op.getParentOperators().get(0));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    Operator<? extends Serializable> currTopOp = mapredCtx.getCurrTopOp();
    String currAliasId = mapredCtx.getCurrAliasId();
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    Map<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx
        .getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);

    ctx.setCurrTopOp(currTopOp);
    ctx.setCurrAliasId(currAliasId);
    ctx.setCurrTask(currTask);

    if (opMapTask == null) {
      GenMapRedUtils.splitPlan(op, ctx);
    } else {
      GenMapRedUtils.joinPlan(op, currTask, opMapTask, ctx, -1, true, false,
          false);
      currTask = opMapTask;
      ctx.setCurrTask(currTask);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(),
        ctx.getCurrAliasId()));
    return null;
  }
View Full Code Here
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
    ReduceSinkOperator op = (ReduceSinkOperator) nd;
    GenMRProcContext ctx = (GenMRProcContext) opProcCtx;

    Map<Operator<? extends Serializable>, GenMapRedCtx> mapCurrCtx = ctx
        .getMapCurrCtx();
    GenMapRedCtx mapredCtx = mapCurrCtx.get(stack.get(stack.size() - 2));
    Task<? extends Serializable> currTask = mapredCtx.getCurrTask();
    MapredWork currPlan = (MapredWork) currTask.getWork();
    Operator<? extends Serializable> currTopOp = mapredCtx.getCurrTopOp();
    String currAliasId = mapredCtx.getCurrAliasId();
    Operator<? extends Serializable> reducer = op.getChildOperators().get(0);
    HashMap<Operator<? extends Serializable>, Task<? extends Serializable>> opTaskMap = ctx
        .getOpTaskMap();
    Task<? extends Serializable> opMapTask = opTaskMap.get(reducer);

    ctx.setCurrTopOp(currTopOp);
    ctx.setCurrAliasId(currAliasId);
    ctx.setCurrTask(currTask);

    // If the plan for this reducer does not exist, initialize the plan
    if (opMapTask == null) {
      if (currPlan.getReducer() == null) {
        GenMapRedUtils.initPlan(op, ctx);
      } else {
        GenMapRedUtils.splitPlan(op, ctx);
      }
    } else {
      // This will happen in case of joins. The current plan can be thrown away
      // after being merged with the original plan
      GenMapRedUtils.joinPlan(op, null, opMapTask, ctx, -1, false, false, false);
      currTask = opMapTask;
      ctx.setCurrTask(currTask);
    }

    mapCurrCtx.put(op, new GenMapRedCtx(ctx.getCurrTask(), ctx.getCurrTopOp(),
        ctx.getCurrAliasId()));
    return null;
  }
View Full Code Here
TOP

Related Classes of org.apache.hadoop.hive.ql.optimizer.GenMRProcContext.GenMapRedCtx

  • org.apache.hadoop.hive.ql.optimizer.GenMapRedUtils
  • org.apache.hadoop.hive.ql.optimizer.GenMROperator
  • org.apache.hadoop.hive.ql.optimizer.GenMRRedSink1
  • org.apache.hadoop.hive.ql.optimizer.GenMRRedSink2
  • org.apache.hadoop.hive.ql.optimizer.GenMRRedSink3
  • org.apache.hadoop.hive.ql.optimizer.GenMRRedSink4
  • org.apache.hadoop.hive.ql.optimizer.GenMRTableScan1
  • org.apache.hadoop.hive.ql.optimizer.GenMRUnion1
  • org.apache.hadoop.hive.ql.optimizer.MapJoinFactory$MapJoin
  • org.apache.hadoop.hive.ql.optimizer.MapJoinFactory$MapJoinMapJoin
    • 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.