use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.
the class PartitionPruner method prune.
/**
* Get the partition list for the table that satisfies the partition pruner
* condition.
*
* @param tab
* the table object for the alias
* @param prunerExpr
* the pruner expression for the alias
* @param conf
* for checking whether "strict" mode is on.
* @param alias
* for generating error message only.
* @param prunedPartitionsMap
* cached result for the table
* @return the partition list for the table that satisfies the partition
* pruner condition.
* @throws SemanticException
*/
public static PrunedPartitionList prune(Table tab, ExprNodeDesc prunerExpr, HiveConf conf, String alias, Map<String, PrunedPartitionList> prunedPartitionsMap) throws SemanticException {
if (LOG.isTraceEnabled()) {
LOG.trace("Started pruning partition");
LOG.trace("dbname = " + tab.getDbName());
LOG.trace("tabname = " + tab.getTableName());
LOG.trace("prune Expression = " + (prunerExpr == null ? "" : prunerExpr));
}
String key = tab.getFullyQualifiedName() + ";";
if (!tab.isPartitioned()) {
// If the table is not partitioned, return empty list.
return getAllPartsFromCacheOrServer(tab, key, false, prunedPartitionsMap);
}
if (!hasColumnExpr(prunerExpr)) {
// If the "strict" mode is on, we have to provide partition pruner for each table.
String error = StrictChecks.checkNoPartitionFilter(conf);
if (error != null) {
throw new SemanticException(error + " No partition predicate for Alias \"" + alias + "\" Table \"" + tab.getTableName() + "\"");
}
}
if (prunerExpr == null) {
// In non-strict mode and there is no predicates at all - get everything.
return getAllPartsFromCacheOrServer(tab, key, false, prunedPartitionsMap);
}
Set<String> partColsUsedInFilter = new LinkedHashSet<String>();
// Replace virtual columns with nulls. See javadoc for details.
prunerExpr = removeNonPartCols(prunerExpr, extractPartColNames(tab), partColsUsedInFilter);
// Remove all parts that are not partition columns. See javadoc for details.
ExprNodeDesc compactExpr = compactExpr(prunerExpr.clone());
String oldFilter = prunerExpr.getExprString();
if (compactExpr == null || isBooleanExpr(compactExpr)) {
if (isFalseExpr(compactExpr)) {
return new PrunedPartitionList(tab, key + compactExpr.getExprString(), new LinkedHashSet<Partition>(0), new ArrayList<String>(0), false);
}
// For null and true values, return every partition
return getAllPartsFromCacheOrServer(tab, key, true, prunedPartitionsMap);
}
String compactExprString = compactExpr.getExprString();
if (LOG.isDebugEnabled()) {
LOG.debug("Filter w/ compacting: " + compactExprString + "; filter w/o compacting: " + oldFilter);
}
key = key + compactExprString;
PrunedPartitionList ppList = prunedPartitionsMap.get(key);
if (ppList != null) {
return ppList;
}
ppList = getPartitionsFromServer(tab, key, (ExprNodeGenericFuncDesc) compactExpr, conf, alias, partColsUsedInFilter, oldFilter.equals(compactExpr.getExprString()));
prunedPartitionsMap.put(key, ppList);
return ppList;
}
use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.
the class PartitionPruner method getAllPartsFromCacheOrServer.
private static PrunedPartitionList getAllPartsFromCacheOrServer(Table tab, String key, boolean unknownPartitions, Map<String, PrunedPartitionList> partsCache) throws SemanticException {
PrunedPartitionList ppList = partsCache == null ? null : partsCache.get(key);
if (ppList != null) {
return ppList;
}
Set<Partition> parts;
try {
parts = getAllPartitions(tab);
} catch (HiveException e) {
throw new SemanticException(e);
}
ppList = new PrunedPartitionList(tab, key, parts, null, unknownPartitions);
if (partsCache != null) {
partsCache.put(key, ppList);
}
return ppList;
}
use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.
the class SparkReduceSinkMapJoinProc method process.
/* (non-Javadoc)
* This processor addresses the RS-MJ case that occurs in spark on the small/hash
* table side of things. The work that RS will be a part of must be connected
* to the MJ work via be a broadcast edge.
* We should not walk down the tree when we encounter this pattern because:
* the type of work (map work or reduce work) needs to be determined
* on the basis of the big table side because it may be a mapwork (no need for shuffle)
* or reduce work.
*/
@SuppressWarnings("unchecked")
@Override
public Object process(Node nd, Stack<Node> stack, NodeProcessorCtx procContext, Object... nodeOutputs) throws SemanticException {
GenSparkProcContext context = (GenSparkProcContext) procContext;
if (!nd.getClass().equals(MapJoinOperator.class)) {
return null;
}
MapJoinOperator mapJoinOp = (MapJoinOperator) nd;
if (stack.size() < 2 || !(stack.get(stack.size() - 2) instanceof ReduceSinkOperator)) {
context.currentMapJoinOperators.add(mapJoinOp);
return null;
}
context.preceedingWork = null;
context.currentRootOperator = null;
ReduceSinkOperator parentRS = (ReduceSinkOperator) stack.get(stack.size() - 2);
// remove the tag for in-memory side of mapjoin
parentRS.getConf().setSkipTag(true);
parentRS.setSkipTag(true);
// remember the original parent list before we start modifying it.
if (!context.mapJoinParentMap.containsKey(mapJoinOp)) {
List<Operator<?>> parents = new ArrayList<Operator<?>>(mapJoinOp.getParentOperators());
context.mapJoinParentMap.put(mapJoinOp, parents);
}
List<BaseWork> mapJoinWork;
/*
* If there was a pre-existing work generated for the big-table mapjoin side,
* we need to hook the work generated for the RS (associated with the RS-MJ pattern)
* with the pre-existing work.
*
* Otherwise, we need to associate that the mapjoin op
* to be linked to the RS work (associated with the RS-MJ pattern).
*
*/
mapJoinWork = context.mapJoinWorkMap.get(mapJoinOp);
int workMapSize = context.childToWorkMap.get(parentRS).size();
Preconditions.checkArgument(workMapSize == 1, "AssertionError: expected context.childToWorkMap.get(parentRS).size() to be 1, but was " + workMapSize);
BaseWork parentWork = context.childToWorkMap.get(parentRS).get(0);
// set the link between mapjoin and parent vertex
int pos = context.mapJoinParentMap.get(mapJoinOp).indexOf(parentRS);
if (pos == -1) {
throw new SemanticException("Cannot find position of parent in mapjoin");
}
LOG.debug("Mapjoin " + mapJoinOp + ", pos: " + pos + " --> " + parentWork.getName());
mapJoinOp.getConf().getParentToInput().put(pos, parentWork.getName());
SparkEdgeProperty edgeProp = new SparkEdgeProperty(SparkEdgeProperty.SHUFFLE_NONE);
if (mapJoinWork != null) {
for (BaseWork myWork : mapJoinWork) {
// link the work with the work associated with the reduce sink that triggered this rule
SparkWork sparkWork = context.currentTask.getWork();
LOG.debug("connecting " + parentWork.getName() + " with " + myWork.getName());
sparkWork.connect(parentWork, myWork, edgeProp);
}
}
// remember in case we need to connect additional work later
Map<BaseWork, SparkEdgeProperty> linkWorkMap = null;
if (context.linkOpWithWorkMap.containsKey(mapJoinOp)) {
linkWorkMap = context.linkOpWithWorkMap.get(mapJoinOp);
} else {
linkWorkMap = new HashMap<BaseWork, SparkEdgeProperty>();
}
linkWorkMap.put(parentWork, edgeProp);
context.linkOpWithWorkMap.put(mapJoinOp, linkWorkMap);
List<ReduceSinkOperator> reduceSinks = context.linkWorkWithReduceSinkMap.get(parentWork);
if (reduceSinks == null) {
reduceSinks = new ArrayList<ReduceSinkOperator>();
}
reduceSinks.add(parentRS);
context.linkWorkWithReduceSinkMap.put(parentWork, reduceSinks);
// create the dummy operators
List<Operator<?>> dummyOperators = new ArrayList<Operator<?>>();
// create an new operator: HashTableDummyOperator, which share the table desc
HashTableDummyDesc desc = new HashTableDummyDesc();
HashTableDummyOperator dummyOp = (HashTableDummyOperator) OperatorFactory.get(mapJoinOp.getCompilationOpContext(), desc);
TableDesc tbl;
// need to create the correct table descriptor for key/value
RowSchema rowSchema = parentRS.getParentOperators().get(0).getSchema();
tbl = PlanUtils.getReduceValueTableDesc(PlanUtils.getFieldSchemasFromRowSchema(rowSchema, ""));
dummyOp.getConf().setTbl(tbl);
Map<Byte, List<ExprNodeDesc>> keyExprMap = mapJoinOp.getConf().getKeys();
List<ExprNodeDesc> keyCols = keyExprMap.get(Byte.valueOf((byte) 0));
StringBuilder keyOrder = new StringBuilder();
StringBuilder keyNullOrder = new StringBuilder();
for (int i = 0; i < keyCols.size(); i++) {
keyOrder.append("+");
keyNullOrder.append("a");
}
TableDesc keyTableDesc = PlanUtils.getReduceKeyTableDesc(PlanUtils.getFieldSchemasFromColumnList(keyCols, "mapjoinkey"), keyOrder.toString(), keyNullOrder.toString());
mapJoinOp.getConf().setKeyTableDesc(keyTableDesc);
// let the dummy op be the parent of mapjoin op
mapJoinOp.replaceParent(parentRS, dummyOp);
List<Operator<? extends OperatorDesc>> dummyChildren = new ArrayList<Operator<? extends OperatorDesc>>();
dummyChildren.add(mapJoinOp);
dummyOp.setChildOperators(dummyChildren);
dummyOperators.add(dummyOp);
// cut the operator tree so as to not retain connections from the parent RS downstream
List<Operator<? extends OperatorDesc>> childOperators = parentRS.getChildOperators();
int childIndex = childOperators.indexOf(mapJoinOp);
childOperators.remove(childIndex);
// at task startup
if (mapJoinWork != null) {
for (BaseWork myWork : mapJoinWork) {
myWork.addDummyOp(dummyOp);
}
}
if (context.linkChildOpWithDummyOp.containsKey(mapJoinOp)) {
for (Operator<?> op : context.linkChildOpWithDummyOp.get(mapJoinOp)) {
dummyOperators.add(op);
}
}
context.linkChildOpWithDummyOp.put(mapJoinOp, dummyOperators);
// replace ReduceSinkOp with HashTableSinkOp for the RSops which are parents of MJop
MapJoinDesc mjDesc = mapJoinOp.getConf();
HiveConf conf = context.conf;
// Unlike in MR, we may call this method multiple times, for each
// small table HTS. But, since it's idempotent, it should be OK.
mjDesc.resetOrder();
float hashtableMemoryUsage;
if (hasGroupBy(mapJoinOp, context)) {
hashtableMemoryUsage = conf.getFloatVar(HiveConf.ConfVars.HIVEHASHTABLEFOLLOWBYGBYMAXMEMORYUSAGE);
} else {
hashtableMemoryUsage = conf.getFloatVar(HiveConf.ConfVars.HIVEHASHTABLEMAXMEMORYUSAGE);
}
mjDesc.setHashTableMemoryUsage(hashtableMemoryUsage);
SparkHashTableSinkDesc hashTableSinkDesc = new SparkHashTableSinkDesc(mjDesc);
SparkHashTableSinkOperator hashTableSinkOp = (SparkHashTableSinkOperator) OperatorFactory.get(mapJoinOp.getCompilationOpContext(), hashTableSinkDesc);
byte tag = (byte) pos;
int[] valueIndex = mjDesc.getValueIndex(tag);
if (valueIndex != null) {
List<ExprNodeDesc> newValues = new ArrayList<ExprNodeDesc>();
List<ExprNodeDesc> values = hashTableSinkDesc.getExprs().get(tag);
for (int index = 0; index < values.size(); index++) {
if (valueIndex[index] < 0) {
newValues.add(values.get(index));
}
}
hashTableSinkDesc.getExprs().put(tag, newValues);
}
// get all parents of reduce sink
List<Operator<? extends OperatorDesc>> rsParentOps = parentRS.getParentOperators();
for (Operator<? extends OperatorDesc> parent : rsParentOps) {
parent.replaceChild(parentRS, hashTableSinkOp);
}
hashTableSinkOp.setParentOperators(rsParentOps);
hashTableSinkOp.getConf().setTag(tag);
return true;
}
use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.
the class HiveOpConverter method translateJoin.
private OpAttr translateJoin(RelNode joinRel) throws SemanticException {
// 0. Additional data structures needed for the join optimization
// through Hive
String[] baseSrc = new String[joinRel.getInputs().size()];
String tabAlias = getHiveDerivedTableAlias();
// 1. Convert inputs
OpAttr[] inputs = new OpAttr[joinRel.getInputs().size()];
List<Operator<?>> children = new ArrayList<Operator<?>>(joinRel.getInputs().size());
for (int i = 0; i < inputs.length; i++) {
inputs[i] = dispatch(joinRel.getInput(i));
children.add(inputs[i].inputs.get(0));
baseSrc[i] = inputs[i].tabAlias;
}
// 2. Generate tags
for (int tag = 0; tag < children.size(); tag++) {
ReduceSinkOperator reduceSinkOp = (ReduceSinkOperator) children.get(tag);
reduceSinkOp.getConf().setTag(tag);
}
// 3. Virtual columns
Set<Integer> newVcolsInCalcite = new HashSet<Integer>();
newVcolsInCalcite.addAll(inputs[0].vcolsInCalcite);
if (joinRel instanceof HiveMultiJoin || !(joinRel instanceof SemiJoin)) {
int shift = inputs[0].inputs.get(0).getSchema().getSignature().size();
for (int i = 1; i < inputs.length; i++) {
newVcolsInCalcite.addAll(HiveCalciteUtil.shiftVColsSet(inputs[i].vcolsInCalcite, shift));
shift += inputs[i].inputs.get(0).getSchema().getSignature().size();
}
}
if (LOG.isDebugEnabled()) {
LOG.debug("Translating operator rel#" + joinRel.getId() + ":" + joinRel.getRelTypeName() + " with row type: [" + joinRel.getRowType() + "]");
}
// 4. Extract join key expressions from HiveSortExchange
ExprNodeDesc[][] joinExpressions = new ExprNodeDesc[inputs.length][];
for (int i = 0; i < inputs.length; i++) {
joinExpressions[i] = ((HiveSortExchange) joinRel.getInput(i)).getJoinExpressions();
}
// 5. Extract rest of join predicate info. We infer the rest of join condition
// that will be added to the filters (join conditions that are not part of
// the join key)
List<RexNode> joinFilters;
if (joinRel instanceof HiveJoin) {
joinFilters = ImmutableList.of(((HiveJoin) joinRel).getJoinFilter());
} else if (joinRel instanceof HiveMultiJoin) {
joinFilters = ((HiveMultiJoin) joinRel).getJoinFilters();
} else if (joinRel instanceof HiveSemiJoin) {
joinFilters = ImmutableList.of(((HiveSemiJoin) joinRel).getJoinFilter());
} else {
throw new SemanticException("Can't handle join type: " + joinRel.getClass().getName());
}
List<List<ExprNodeDesc>> filterExpressions = Lists.newArrayList();
for (int i = 0; i < joinFilters.size(); i++) {
List<ExprNodeDesc> filterExpressionsForInput = new ArrayList<ExprNodeDesc>();
if (joinFilters.get(i) != null) {
for (RexNode conj : RelOptUtil.conjunctions(joinFilters.get(i))) {
ExprNodeDesc expr = convertToExprNode(conj, joinRel, null, newVcolsInCalcite);
filterExpressionsForInput.add(expr);
}
}
filterExpressions.add(filterExpressionsForInput);
}
// 6. Generate Join operator
JoinOperator joinOp = genJoin(joinRel, joinExpressions, filterExpressions, children, baseSrc, tabAlias);
// 7. Return result
return new OpAttr(tabAlias, newVcolsInCalcite, joinOp);
}
use of org.apache.hadoop.hive.ql.parse.SemanticException in project hive by apache.
the class DumpMetaData method loadDumpFromFile.
private void loadDumpFromFile() throws SemanticException {
BufferedReader br = null;
try {
// read from dumpfile and instantiate self
FileSystem fs = dumpFile.getFileSystem(hiveConf);
br = new BufferedReader(new InputStreamReader(fs.open(dumpFile)));
String line = null;
if ((line = br.readLine()) != null) {
String[] lineContents = line.split("\t", 5);
setDump(DumpType.valueOf(lineContents[0]), Long.valueOf(lineContents[1]), Long.valueOf(lineContents[2]), new Path(lineContents[3]));
setPayload(lineContents[4].equals(Utilities.nullStringOutput) ? null : lineContents[4]);
ReplChangeManager.setCmRoot(cmRoot);
} else {
throw new IOException("Unable to read valid values from dumpFile:" + dumpFile.toUri().toString());
}
} catch (IOException ioe) {
throw new SemanticException(ioe);
} finally {
if (br != null) {
try {
br.close();
} catch (IOException e) {
throw new SemanticException(e);
}
}
}
}
Aggregations