Package org.apache.hadoop.hive.ql.exec

Examples of org.apache.hadoop.hive.ql.exec.HashTableSinkOperator

440
441
442
443
444
445
446
447
448
449
450
451
    // generate empty hashtable for this (byte)tag
    String tmpURI = this.getWork().getTmpFileURI();

    String fileName = work.getBucketFileName(bigBucketFileName);

    HashTableSinkOperator htso = (HashTableSinkOperator)childOp;
    String tmpURIPath = Utilities.generatePath(tmpURI, htso.getConf().getDumpFilePrefix(),
        tag, fileName);
    console.printInfo(Utilities.now() + "\tDump the hashtable into file: " + tmpURIPath);
    Path path = new Path(tmpURIPath);
    FileSystem fs = path.getFileSystem(job);
    ObjectOutputStream out = new ObjectOutputStream(fs.create(path));
View Full Code Here
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
      } catch (Exception e) {
        e.printStackTrace();
      }

      HashTableSinkDesc hashTableSinkDesc = new HashTableSinkDesc(mapJoinOp.getConf());
      HashTableSinkOperator hashTableSinkOp = (HashTableSinkOperator) OperatorFactory
          .get(hashTableSinkDesc);

      // set hashtable memory usage
      float hashtableMemoryUsage;
      if (context.isFollowedByGroupBy()) {
        hashtableMemoryUsage = context.getParseCtx().getConf().getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
      } else {
        hashtableMemoryUsage = context.getParseCtx().getConf().getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEMAXMEMORYUSAGE);
      }
      hashTableSinkOp.getConf().setHashtableMemoryUsage(hashtableMemoryUsage);

      // get the last operator for processing big tables
      int bigTable = mapJoinOp.getConf().getPosBigTable();
      Byte[] order = mapJoinOp.getConf().getTagOrder();
      int bigTableAlias = (int) order[bigTable];

      // the parent ops for hashTableSinkOp
      List<Operator<? extends OperatorDesc>> smallTablesParentOp =
        new ArrayList<Operator<? extends OperatorDesc>>();
      List<Operator<? extends OperatorDesc>> dummyOperators =
        new ArrayList<Operator<? extends OperatorDesc>>();
      // get all parents
      List<Operator<? extends OperatorDesc>> parentsOp = mapJoinOp.getParentOperators();
      for (int i = 0; i < parentsOp.size(); i++) {
        if (i == bigTableAlias) {
          smallTablesParentOp.add(null);
          continue;
        }
        Operator<? extends OperatorDesc> parent = parentsOp.get(i);
        // let hashtable Op be the child of this parent
        parent.replaceChild(mapJoinOp, hashTableSinkOp);
        // keep the parent id correct
        smallTablesParentOp.add(parent);

        // create an new operator: HashTable DummyOpeator, which share the table desc
        HashTableDummyDesc desc = new HashTableDummyDesc();
        HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(desc);
        TableDesc tbl;

        if (parent.getSchema() == null) {
          if (parent instanceof TableScanOperator) {
            tbl = ((TableScanOperator) parent).getTableDesc();
          } else {
            throw new SemanticException();
          }
        } else {
          // get parent schema
          RowSchema rowSchema = parent.getSchema();
          tbl = PlanUtils.getIntermediateFileTableDesc(PlanUtils.getFieldSchemasFromRowSchema(
              rowSchema, ""));
        }
        dummyOp.getConf().setTbl(tbl);
        // let the dummy op be the parent of mapjoin op
        mapJoinOp.replaceParent(parent, dummyOp);
        List<Operator<? extends OperatorDesc>> dummyChildren =
          new ArrayList<Operator<? extends OperatorDesc>>();
        dummyChildren.add(mapJoinOp);
        dummyOp.setChildOperators(dummyChildren);
        // add this dummy op to the dummp operator list
        dummyOperators.add(dummyOp);
      }
      hashTableSinkOp.setParentOperators(smallTablesParentOp);
      for (Operator<? extends OperatorDesc> op : dummyOperators) {
        context.addDummyParentOp(op);
      }
      return null;
    }
View Full Code Here
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
      } catch (Exception e) {
        e.printStackTrace();
      }

      HashTableSinkDesc hashTableSinkDesc = new HashTableSinkDesc(mapJoinOp.getConf());
      HashTableSinkOperator hashTableSinkOp = (HashTableSinkOperator) OperatorFactory
          .get(hashTableSinkDesc);

      // set hashtable memory usage
      float hashtableMemoryUsage;
      if (context.isFollowedByGroupBy()) {
        hashtableMemoryUsage = context.getParseCtx().getConf().getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
      } else {
        hashtableMemoryUsage = context.getParseCtx().getConf().getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEMAXMEMORYUSAGE);
      }
      hashTableSinkOp.getConf().setHashtableMemoryUsage(hashtableMemoryUsage);

      // get the last operator for processing big tables
      int bigTable = mapJoinOp.getConf().getPosBigTable();
      Byte[] order = mapJoinOp.getConf().getTagOrder();
      int bigTableAlias = (int) order[bigTable];

      // the parent ops for hashTableSinkOp
      List<Operator<? extends Serializable>> smallTablesParentOp = new ArrayList<Operator<? extends Serializable>>();
      List<Operator<? extends Serializable>> dummyOperators = new ArrayList<Operator<? extends Serializable>>();
      // get all parents
      List<Operator<? extends Serializable>> parentsOp = mapJoinOp.getParentOperators();
      for (int i = 0; i < parentsOp.size(); i++) {
        if (i == bigTableAlias) {
          smallTablesParentOp.add(null);
          continue;
        }
        Operator<? extends Serializable> parent = parentsOp.get(i);
        // let hashtable Op be the child of this parent
        parent.replaceChild(mapJoinOp, hashTableSinkOp);
        // keep the parent id correct
        smallTablesParentOp.add(parent);

        // create an new operator: HashTable DummyOpeator, which share the table desc
        HashTableDummyDesc desc = new HashTableDummyDesc();
        HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(desc);
        TableDesc tbl;

        if (parent.getSchema() == null) {
          if (parent instanceof TableScanOperator) {
            tbl = ((TableScanOperator) parent).getTableDesc();
          } else {
            throw new SemanticException();
          }
        } else {
          // get parent schema
          RowSchema rowSchema = parent.getSchema();
          tbl = PlanUtils.getIntermediateFileTableDesc(PlanUtils.getFieldSchemasFromRowSchema(
              rowSchema, ""));
        }
        dummyOp.getConf().setTbl(tbl);
        // let the dummy op be the parent of mapjoin op
        mapJoinOp.replaceParent(parent, dummyOp);
        List<Operator<? extends Serializable>> dummyChildren = new ArrayList<Operator<? extends Serializable>>();
        dummyChildren.add(mapJoinOp);
        dummyOp.setChildOperators(dummyChildren);
        // add this dummy op to the dummp operator list
        dummyOperators.add(dummyOp);
      }
      hashTableSinkOp.setParentOperators(smallTablesParentOp);
      for (Operator<? extends Serializable> op : dummyOperators) {
        context.addDummyParentOp(op);
      }
      return null;
    }
View Full Code Here
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
      return;
    }
    JobConf job = new JobConf(hconf);
    MapredLocalTask localTask = new MapredLocalTask(localWork, job, false);

    HashTableSinkOperator sink = new TemporaryHashSinkOperator(desc);
    sink.setParentOperators(new ArrayList<Operator<? extends OperatorDesc>>(directWorks));

    for (Operator<?> operator : directWorks) {
      if (operator != null) {
        operator.setChildOperators(Arrays.<Operator<? extends OperatorDesc>>asList(sink));
      }
    }
    localTask.setExecContext(context);
    localTask.startForward(inputFileName);

    MapJoinTableContainer[] tables = sink.getMapJoinTables();
    for (int i = 0; i < sink.getNumParent(); i++) {
      if (sink.getParentOperators().get(i) != null) {
        mapJoinTables[i] = tables[i];
      }
    }

    Arrays.fill(tables, null);
View Full Code Here
436
437
438
439
440
441
442
443
444
445
446
447
    // generate empty hashtable for this (byte)tag
    Path tmpPath = this.getWork().getTmpPath();

    String fileName = work.getBucketFileName(bigBucketFileName);

    HashTableSinkOperator htso = (HashTableSinkOperator)childOp;
    Path path = Utilities.generatePath(tmpPath, htso.getConf().getDumpFilePrefix(),
        tag, fileName);
    console.printInfo(Utilities.now() + "\tDump the hashtable into file: " + path);
    FileSystem fs = path.getFileSystem(job);
    ObjectOutputStream out = new ObjectOutputStream(fs.create(path));
    try {
View Full Code Here
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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
      mapJoinOp.getConf().resetOrder();

      HiveConf conf = context.getParseCtx().getConf();

      HashTableSinkDesc hashTableSinkDesc = new HashTableSinkDesc(mapJoinOp.getConf());
      HashTableSinkOperator hashTableSinkOp = (HashTableSinkOperator) OperatorFactory
          .get(hashTableSinkDesc);

      // set hashtable memory usage
      float hashtableMemoryUsage;
      if (context.isFollowedByGroupBy()) {
        hashtableMemoryUsage = conf.getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
      } else {
        hashtableMemoryUsage = conf.getFloatVar(
            HiveConf.ConfVars.HIVEHASHTABLEMAXMEMORYUSAGE);
      }
      mapJoinOp.getConf().setHashTableMemoryUsage(hashtableMemoryUsage);
      LOG.info("Setting max memory usage to " + hashtableMemoryUsage + " for table sink "
          + (context.isFollowedByGroupBy() ? "" : "not") + " followed by group by");
      hashTableSinkOp.getConf().setHashtableMemoryUsage(hashtableMemoryUsage);

      // get the last operator for processing big tables
      int bigTable = mapJoinOp.getConf().getPosBigTable();
      Byte[] orders = mapJoinOp.getConf().getTagOrder();

      // todo: support tez/vectorization
      boolean useNontaged = conf.getBoolVar(
          HiveConf.ConfVars.HIVECONVERTJOINUSENONSTAGED) &&
          conf.getVar(HiveConf.ConfVars.HIVE_EXECUTION_ENGINE).equals("mr") &&
          !conf.getBoolVar(HiveConf.ConfVars.HIVE_VECTORIZATION_ENABLED);

      // the parent ops for hashTableSinkOp
      List<Operator<? extends OperatorDesc>> smallTablesParentOp =
        new ArrayList<Operator<? extends OperatorDesc>>();
      List<Operator<? extends OperatorDesc>> dummyOperators =
        new ArrayList<Operator<? extends OperatorDesc>>();
      List<Operator<? extends OperatorDesc>> directOperators =
          new ArrayList<Operator<? extends OperatorDesc>>();
      // get all parents
      List<Operator<? extends OperatorDesc>> parentsOp = mapJoinOp.getParentOperators();
      for (int i = 0; i < parentsOp.size(); i++) {
        if (i == bigTable) {
          smallTablesParentOp.add(null);
          directOperators.add(null);
          continue;
        }
        Operator<? extends OperatorDesc> parent = parentsOp.get(i);
        boolean directFetchable = useNontaged &&
            (parent instanceof TableScanOperator || parent instanceof MapJoinOperator);
        if (directFetchable) {
          // no filter, no projection. no need to stage
          smallTablesParentOp.add(null);
          directOperators.add(parent);
          hashTableSinkDesc.getKeys().put(orders[i], null);
          hashTableSinkDesc.getExprs().put(orders[i], null);
          hashTableSinkDesc.getFilters().put(orders[i], null);
        } else {
          // keep the parent id correct
          smallTablesParentOp.add(parent);
          directOperators.add(null);
        }
        // let hashtable Op be the child of this parent
        parent.replaceChild(mapJoinOp, hashTableSinkOp);
        if (directFetchable) {
          parent.setChildOperators(null);
        }

        // create an new operator: HashTable DummyOpeator, which share the table desc
        HashTableDummyDesc desc = new HashTableDummyDesc();
        HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(desc);
        TableDesc tbl;

        if (parent.getSchema() == null) {
          if (parent instanceof TableScanOperator) {
            tbl = ((TableScanOperator) parent).getTableDesc();
          } else {
            throw new SemanticException("Expected parent operator of type TableScanOperator." +
              "Found " + parent.getClass().getName() + " instead.");
          }
        } else {
          // get parent schema
          RowSchema rowSchema = parent.getSchema();
          tbl = PlanUtils.getIntermediateFileTableDesc(PlanUtils.getFieldSchemasFromRowSchema(
              rowSchema, ""));
        }
        dummyOp.getConf().setTbl(tbl);
        // let the dummy op be the parent of mapjoin op
        mapJoinOp.replaceParent(parent, dummyOp);
        List<Operator<? extends OperatorDesc>> dummyChildren =
          new ArrayList<Operator<? extends OperatorDesc>>();
        dummyChildren.add(mapJoinOp);
        dummyOp.setChildOperators(dummyChildren);
        // add this dummy op to the dummp operator list
        dummyOperators.add(dummyOp);
      }
      hashTableSinkOp.setParentOperators(smallTablesParentOp);
      for (Operator<? extends OperatorDesc> op : dummyOperators) {
        context.addDummyParentOp(op);
      }
      if (hasAnyDirectFetch(directOperators)) {
        context.addDirectWorks(mapJoinOp, directOperators);
View Full Code Here
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
      return;
    }
    JobConf job = new JobConf(hconf);
    MapredLocalTask localTask = new MapredLocalTask(localWork, job, false);

    HashTableSinkOperator sink = new TemporaryHashSinkOperator(desc);
    sink.setParentOperators(new ArrayList<Operator<? extends OperatorDesc>>(directWorks));

    for (Operator<?> operator : directWorks) {
      if (operator != null) {
        operator.setChildOperators(Arrays.<Operator<? extends OperatorDesc>>asList(sink));
      }
    }
    localTask.setExecContext(context);
    localTask.startForward(inputFileName);

    MapJoinTableContainer[] tables = sink.getMapJoinTables();
    for (int i = 0; i < sink.getNumParent(); i++) {
      if (sink.getParentOperators().get(i) != null) {
        mapJoinTables[i] = tables[i];
      }
    }

    Arrays.fill(tables, null);
View Full Code Here
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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
      mapJoinDesc.setHashTableMemoryUsage(hashtableMemoryUsage);
      LOG.info("Setting max memory usage to " + hashtableMemoryUsage + " for table sink "
          + (context.isFollowedByGroupBy() ? "" : "not") + " followed by group by");

      HashTableSinkDesc hashTableSinkDesc = new HashTableSinkDesc(mapJoinDesc);
      HashTableSinkOperator hashTableSinkOp = (HashTableSinkOperator) OperatorFactory
          .get(hashTableSinkDesc);

      // get the last operator for processing big tables
      int bigTable = mapJoinDesc.getPosBigTable();

      // todo: support tez/vectorization
      boolean useNontaged = conf.getBoolVar(
          HiveConf.ConfVars.HIVECONVERTJOINUSENONSTAGED) &&
          conf.getVar(HiveConf.ConfVars.HIVE_EXECUTION_ENGINE).equals("mr") &&
          !conf.getBoolVar(HiveConf.ConfVars.HIVE_VECTORIZATION_ENABLED);

      // the parent ops for hashTableSinkOp
      List<Operator<? extends OperatorDesc>> smallTablesParentOp =
        new ArrayList<Operator<? extends OperatorDesc>>();
      List<Operator<? extends OperatorDesc>> dummyOperators =
        new ArrayList<Operator<? extends OperatorDesc>>();
      List<Operator<? extends OperatorDesc>> directOperators =
          new ArrayList<Operator<? extends OperatorDesc>>();
      // get all parents
      List<Operator<? extends OperatorDesc>> parentsOp = mapJoinOp.getParentOperators();
      for (byte i = 0; i < parentsOp.size(); i++) {
        if (i == bigTable) {
          smallTablesParentOp.add(null);
          directOperators.add(null);
          continue;
        }
        Operator<? extends OperatorDesc> parent = parentsOp.get(i);
        boolean directFetchable = useNontaged &&
            (parent instanceof TableScanOperator || parent instanceof MapJoinOperator);
        if (directFetchable) {
          // no filter, no projection. no need to stage
          smallTablesParentOp.add(null);
          directOperators.add(parent);
          hashTableSinkDesc.getKeys().put(i, null);
          hashTableSinkDesc.getExprs().put(i, null);
          hashTableSinkDesc.getFilters().put(i, null);
        } else {
          // keep the parent id correct
          smallTablesParentOp.add(parent);
          directOperators.add(null);
          int[] valueIndex = mapJoinDesc.getValueIndex(i);
          if (valueIndex != null) {
            // remove values in key exprs
            // schema for value is already fixed in MapJoinProcessor#convertJoinOpMapJoinOp
            List<ExprNodeDesc> newValues = new ArrayList<ExprNodeDesc>();
            List<ExprNodeDesc> values = hashTableSinkDesc.getExprs().get(i);
            for (int index = 0; index < values.size(); index++) {
              if (valueIndex[index] < 0) {
                newValues.add(values.get(index));
              }
            }
            hashTableSinkDesc.getExprs().put(i, newValues);
          }
        }
        // let hashtable Op be the child of this parent
        parent.replaceChild(mapJoinOp, hashTableSinkOp);
        if (directFetchable) {
          parent.setChildOperators(null);
        }

        // create new operator: HashTable DummyOperator, which share the table desc
        HashTableDummyDesc desc = new HashTableDummyDesc();
        HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(desc);
        TableDesc tbl;

        if (parent.getSchema() == null) {
          if (parent instanceof TableScanOperator) {
            tbl = ((TableScanOperator) parent).getTableDesc();
          } else {
            throw new SemanticException("Expected parent operator of type TableScanOperator." +
              "Found " + parent.getClass().getName() + " instead.");
          }
        } else {
          // get parent schema
          RowSchema rowSchema = parent.getSchema();
          tbl = PlanUtils.getIntermediateFileTableDesc(PlanUtils.getFieldSchemasFromRowSchema(
              rowSchema, ""));
        }
        dummyOp.getConf().setTbl(tbl);
        // let the dummy op be the parent of mapjoin op
        mapJoinOp.replaceParent(parent, dummyOp);
        List<Operator<? extends OperatorDesc>> dummyChildren =
          new ArrayList<Operator<? extends OperatorDesc>>();
        dummyChildren.add(mapJoinOp);
        dummyOp.setChildOperators(dummyChildren);
        // add this dummy op to the dummp operator list
        dummyOperators.add(dummyOp);
      }
      hashTableSinkOp.setParentOperators(smallTablesParentOp);
      for (Operator<? extends OperatorDesc> op : dummyOperators) {
        context.addDummyParentOp(op);
      }
      if (hasAnyDirectFetch(directOperators)) {
        context.addDirectWorks(mapJoinOp, directOperators);
View Full Code Here
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
      return;
    }
    JobConf job = new JobConf(hconf);
    MapredLocalTask localTask = new MapredLocalTask(localWork, job, false);

    HashTableSinkOperator sink = new TemporaryHashSinkOperator(desc);
    sink.setParentOperators(new ArrayList<Operator<? extends OperatorDesc>>(directWorks));

    for (Operator<?> operator : directWorks) {
      if (operator != null) {
        operator.setChildOperators(Arrays.<Operator<? extends OperatorDesc>>asList(sink));
      }
    }
    localTask.setExecContext(context);
    localTask.startForward(inputFileName);

    MapJoinTableContainer[] tables = sink.getMapJoinTables();
    for (int i = 0; i < sink.getNumParent(); i++) {
      if (sink.getParentOperators().get(i) != null) {
        mapJoinTables[i] = tables[i];
      }
    }

    Arrays.fill(tables, null);
View Full Code Here
436
437
438
439
440
441
442
443
444
445
446
447
    // generate empty hashtable for this (byte)tag
    Path tmpPath = this.getWork().getTmpPath();

    String fileName = work.getBucketFileName(bigBucketFileName);

    HashTableSinkOperator htso = (HashTableSinkOperator)childOp;
    Path path = Utilities.generatePath(tmpPath, htso.getConf().getDumpFilePrefix(),
        tag, fileName);
    console.printInfo(Utilities.now() + "\tDump the hashtable into file: " + path);
    FileSystem fs = path.getFileSystem(job);
    ObjectOutputStream out = new ObjectOutputStream(fs.create(path));
    try {
View Full Code Here
TOP

Related Classes of org.apache.hadoop.hive.ql.exec.HashTableSinkOperator

  • org.apache.hadoop.fs.FileStatus
  • org.apache.hadoop.fs.FileSystem
  • org.apache.hadoop.fs.Path
  • org.apache.hadoop.hive.ql.exec.mapjoin.MapJoinMemoryExhaustionHandler
  • org.apache.hadoop.hive.ql.exec.mr.HashTableLoader
  • org.apache.hadoop.hive.ql.exec.mr.MapredLocalTask
  • org.apache.hadoop.hive.ql.exec.persistence.AbstractMapJoinKey
  • org.apache.hadoop.hive.ql.exec.persistence.HashMapWrapper
  • org.apache.hadoop.hive.ql.exec.persistence.MapJoinEagerRowContainer
  • org.apache.hadoop.hive.ql.exec.persistence.MapJoinKey
    • 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.