/**
* destination table : a table whose columns are getting queried intermediate table : a table
* which is only used as a link between cube and destination table
*
* @param alias
* @param tblName
* @param isOptional pass false when it's a destination table pass true when it's an intermediate
* table when join chain destination is being added, this will be false.
* @param isChainedDimension pass true when you're adding the dimension as a joinchain
* destination, pass false when this table is mentioned by name in the user query
* @return true if added
* @throws LensException
*/
private boolean addQueriedTable(
String alias, String tblName, boolean isOptional, boolean isChainedDimension)
throws LensException {
alias = alias.toLowerCase();
if (cubeTbls.containsKey(alias)) {
return true;
}
try {
if (metastoreClient.isCube(tblName)) {
if (cube != null) {
if (!cube.getName().equalsIgnoreCase(tblName)) {
throw new LensException(
LensCubeErrorCode.MORE_THAN_ONE_CUBE.getLensErrorInfo(), cube.getName(), tblName);
}
}
cube = metastoreClient.getCube(tblName);
cubeTbls.put(alias, (AbstractCubeTable) cube);
} else if (metastoreClient.isDimension(tblName)) {
Dimension dim = metastoreClient.getDimension(tblName);
if (!isOptional) {
dimensions.add(dim);
}
if (!isChainedDimension) {
nonChainedDimensions.add(dim);
}
cubeTbls.put(alias, dim);
} else {
return false;
}
} catch (HiveException e) {
return false;
}
return true;
}
public Set<String> getPartitionColumnsQueried() {
Set<String> partsQueried = Sets.newHashSet();
for (TimeRange range : getTimeRanges()) {
partsQueried.add(range.getPartitionColumn());
}
return partsQueried;
}
public void addColumnsQueried(String alias, String column) {
Set<String> cols = tblAliasToColumns.get(alias.toLowerCase());
if (cols == null) {
cols = new LinkedHashSet<String>();
tblAliasToColumns.put(alias.toLowerCase(), cols);
}
cols.add(column);
}
public void pruneCandidateFactWithCandidateSet(CandidateTablePruneCause pruneCause) {
// remove candidate facts that are not part of any covering set
Set<CandidateFact> allCoveringFacts = new HashSet<CandidateFact>();
for (Set<CandidateFact> set : candidateFactSets) {
allCoveringFacts.addAll(set);
}
for (Iterator<CandidateFact> i = candidateFacts.iterator(); i.hasNext(); ) {
CandidateFact cfact = i.next();
if (!allCoveringFacts.contains(cfact)) {
log.info("Not considering fact table:{} as {}", cfact, pruneCause);
addFactPruningMsgs(cfact.fact, pruneCause);
i.remove();
}
}
}
public CubeQueryContext(ASTNode ast, QB qb, Configuration queryConf, HiveConf metastoreConf)
throws LensException {
this.ast = ast;
this.qb = qb;
this.conf = queryConf;
this.clauseName = getClause();
this.timeRanges = new ArrayList<TimeRange>();
try {
metastoreClient = CubeMetastoreClient.getInstance(metastoreConf);
} catch (HiveException e) {
throw new LensException(e);
}
if (qb.getParseInfo().getWhrForClause(clauseName) != null) {
this.whereAST = qb.getParseInfo().getWhrForClause(clauseName);
}
if (qb.getParseInfo().getHavingForClause(clauseName) != null) {
this.havingAST = qb.getParseInfo().getHavingForClause(clauseName);
}
if (qb.getParseInfo().getOrderByForClause(clauseName) != null) {
this.orderByAST = qb.getParseInfo().getOrderByForClause(clauseName);
}
if (qb.getParseInfo().getGroupByForClause(clauseName) != null) {
this.groupByAST = qb.getParseInfo().getGroupByForClause(clauseName);
}
if (qb.getParseInfo().getSelForClause(clauseName) != null) {
this.selectAST = qb.getParseInfo().getSelForClause(clauseName);
}
for (ASTNode aggrTree : qb.getParseInfo().getAggregationExprsForClause(clauseName).values()) {
String aggr = HQLParser.getString(aggrTree);
aggregateExprs.add(aggr);
}
extractMetaTables();
}
private void extractMetaTables() throws LensException {
List<String> tabAliases = new ArrayList<String>(qb.getTabAliases());
Set<String> missing = new HashSet<String>();
for (String alias : tabAliases) {
boolean added = addQueriedTable(alias);
if (!added) {
missing.add(alias);
}
}
for (String alias : missing) {
// try adding them as joinchains
boolean added = addJoinChain(alias, false);
if (!added) {
log.info("Queried tables do not exist. Missing table:{}", alias);
throw new LensException(LensCubeErrorCode.NEITHER_CUBE_NOR_DIMENSION.getLensErrorInfo());
}
}
}
String getQBFromString(CandidateFact fact, Map<Dimension, CandidateDim> dimsToQuery)
throws LensException {
String fromString;
if (getJoinAST() == null) {
if (cube != null) {
if (dimensions.size() > 0) {
throw new LensException(LensCubeErrorCode.NO_JOIN_CONDITION_AVAILABLE.getLensErrorInfo());
}
fromString = fact.getStorageString(getAliasForTableName(cube.getName()));
} else {
if (dimensions.size() != 1) {
throw new LensException(LensCubeErrorCode.NO_JOIN_CONDITION_AVAILABLE.getLensErrorInfo());
}
Dimension dim = dimensions.iterator().next();
fromString = dimsToQuery.get(dim).getStorageString(getAliasForTableName(dim.getName()));
}
} else {
StringBuilder builder = new StringBuilder();
getQLString(qb.getQbJoinTree(), builder, fact, dimsToQuery);
fromString = builder.toString();
}
return fromString;
}
/**
* Prune candidate fact sets with respect to available candidate facts.
*
* <p>Prune a candidate set, if any of the fact is missing.
*
* @param pruneCause
*/
public void pruneCandidateFactSet(CandidateTablePruneCode pruneCause) {
// remove candidate fact sets that have missing facts
for (Iterator<Set<CandidateFact>> i = candidateFactSets.iterator(); i.hasNext(); ) {
Set<CandidateFact> cfacts = i.next();
if (!candidateFacts.containsAll(cfacts)) {
log.info(
"Not considering fact table set:{} as they have non candidate tables and facts missing because of {}",
cfacts,
pruneCause);
i.remove();
}
}
// prune candidate facts
pruneCandidateFactWithCandidateSet(CandidateTablePruneCode.ELEMENT_IN_SET_PRUNED);
}
public boolean isCubeMeasure(String col) {
if (col == null) {
return false;
}
col = col.trim();
// Take care of brackets added around col names in HQLParsrer.getString
if (col.startsWith("(") && col.endsWith(")") && col.length() > 2) {
col = col.substring(1, col.length() - 1);
}
String[] split = StringUtils.split(col, ".");
if (split.length <= 1) {
col = col.trim().toLowerCase();
if (queriedExprs.contains(col)) {
return exprCtx
.getExpressionContext(col, getAliasForTableName(cube.getName()))
.hasMeasures();
} else {
return cube.getMeasureNames().contains(col);
}
} else {
String cubeName = split[0].trim().toLowerCase();
String colName = split[1].trim().toLowerCase();
if (cubeName.equalsIgnoreCase(cube.getName())
|| cubeName.equals(getAliasForTableName(cube.getName()))) {
if (queriedExprs.contains(colName)) {
return exprCtx.getExpressionContext(colName, cubeName).hasMeasures();
} else {
return cube.getMeasureNames().contains(colName.toLowerCase());
}
} else {
return false;
}
}
}
private HQLContextInterface createHQLContext(
Set<CandidateFact> facts,
Map<Dimension, CandidateDim> dimsToQuery,
Map<CandidateFact, Set<Dimension>> factDimMap)
throws LensException {
if (facts == null || facts.size() == 0) {
return new DimOnlyHQLContext(dimsToQuery, this, this);
} else if (facts.size() == 1 && facts.iterator().next().getStorageTables().size() > 1) {
// create single fact with multiple storage context
return new SingleFactMultiStorageHQLContext(facts.iterator().next(), dimsToQuery, this, this);
} else if (facts.size() == 1 && facts.iterator().next().getStorageTables().size() == 1) {
// create single fact context
return new SingleFactSingleStorageHQLContext(
facts.iterator().next(), dimsToQuery, this, this);
} else {
return new MultiFactHQLContext(facts, dimsToQuery, factDimMap, this);
}
}
private Map<Dimension, CandidateDim> pickCandidateDimsToQuery(Set<Dimension> dimensions)
throws LensException {
Map<Dimension, CandidateDim> dimsToQuery = new HashMap<Dimension, CandidateDim>();
if (!dimensions.isEmpty()) {
for (Dimension dim : dimensions) {
if (candidateDims.get(dim) != null && candidateDims.get(dim).size() > 0) {
CandidateDim cdim = candidateDims.get(dim).iterator().next();
log.info(
"Available candidate dims are:{}, picking up {} for querying.",
candidateDims.get(dim),
cdim.dimtable);
dimsToQuery.put(dim, cdim);
} else {
String reason = "";
if (dimPruningMsgs.get(dim) != null && !dimPruningMsgs.get(dim).isEmpty()) {
ByteArrayOutputStream out = null;
try {
ObjectMapper mapper = new ObjectMapper();
out = new ByteArrayOutputStream();
mapper.writeValue(out, dimPruningMsgs.get(dim).getJsonObject());
reason = out.toString("UTF-8");
} catch (Exception e) {
throw new LensException("Error writing dim pruning messages", e);
} finally {
if (out != null) {
try {
out.close();
} catch (IOException e) {
throw new LensException(e);
}
}
}
}
log.error(
"Query rewrite failed due to NO_CANDIDATE_DIM_AVAILABLE, Cause {}",
dimPruningMsgs.get(dim).toJsonObject());
throw new NoCandidateDimAvailableException(dimPruningMsgs.get(dim));
}
}
}
return dimsToQuery;
}
public String toHQL() throws LensException {
Set<CandidateFact> cfacts = pickCandidateFactToQuery();
Map<Dimension, CandidateDim> dimsToQuery = pickCandidateDimsToQuery(dimensions);
log.info("facts:{}, dimsToQuery: {}", cfacts, dimsToQuery);
if (autoJoinCtx != null) {
// prune join paths for picked fact and dimensions
autoJoinCtx.pruneAllPaths(cube, cfacts, dimsToQuery);
}
Map<CandidateFact, Set<Dimension>> factDimMap = new HashMap<>();
if (cfacts != null) {
if (cfacts.size() > 1) {
// copy ASTs for each fact
for (CandidateFact cfact : cfacts) {
cfact.copyASTs(this);
factDimMap.put(cfact, new HashSet<>(dimsToQuery.keySet()));
}
}
for (CandidateFact fact : cfacts) {
addRangeClauses(fact);
}
}
// pick dimension tables required during expression expansion for the picked fact and dimensions
Set<Dimension> exprDimensions = new HashSet<Dimension>();
if (cfacts != null) {
for (CandidateFact cfact : cfacts) {
Set<Dimension> factExprDimTables =
exprCtx.rewriteExprCtx(cfact, dimsToQuery, cfacts.size() > 1 ? cfact : this);
exprDimensions.addAll(factExprDimTables);
if (cfacts.size() > 1) {
factDimMap.get(cfact).addAll(factExprDimTables);
}
}
if (cfacts.size() > 1) {
havingAST = MultiFactHQLContext.pushDownHaving(havingAST, this, cfacts);
}
} else {
// dim only query
exprDimensions.addAll(exprCtx.rewriteExprCtx(null, dimsToQuery, this));
}
dimsToQuery.putAll(pickCandidateDimsToQuery(exprDimensions));
log.info("facts:{}, dimsToQuery: {}", cfacts, dimsToQuery);
// pick denorm tables for the picked fact and dimensions
Set<Dimension> denormTables = new HashSet<Dimension>();
if (cfacts != null) {
for (CandidateFact cfact : cfacts) {
Set<Dimension> factDenormTables =
deNormCtx.rewriteDenormctx(cfact, dimsToQuery, cfacts.size() > 1);
denormTables.addAll(factDenormTables);
if (cfacts.size() > 1) {
factDimMap.get(cfact).addAll(factDenormTables);
}
}
} else {
denormTables.addAll(deNormCtx.rewriteDenormctx(null, dimsToQuery, false));
}
dimsToQuery.putAll(pickCandidateDimsToQuery(denormTables));
log.info("facts:{}, dimsToQuery: {}", cfacts, dimsToQuery);
// Prune join paths once denorm tables are picked
if (autoJoinCtx != null) {
// prune join paths for picked fact and dimensions
autoJoinCtx.pruneAllPaths(cube, cfacts, dimsToQuery);
}
if (autoJoinCtx != null) {
// add optional dims from Join resolver
Set<Dimension> joiningTables = new HashSet<Dimension>();
if (cfacts != null && cfacts.size() > 1) {
for (CandidateFact cfact : cfacts) {
Set<Dimension> factJoiningTables =
autoJoinCtx.pickOptionalTables(cfact, factDimMap.get(cfact), this);
factDimMap.get(cfact).addAll(factJoiningTables);
joiningTables.addAll(factJoiningTables);
}
} else {
joiningTables.addAll(autoJoinCtx.pickOptionalTables(null, dimsToQuery.keySet(), this));
}
dimsToQuery.putAll(pickCandidateDimsToQuery(joiningTables));
}
log.info("Picked Fact:{} dimsToQuery: {}", cfacts, dimsToQuery);
pickedDimTables = dimsToQuery.values();
pickedFacts = cfacts;
if (cfacts != null) {
if (cfacts.size() > 1) {
// Update ASTs for each fact
for (CandidateFact cfact : cfacts) {
cfact.updateASTs(this);
}
whereAST =
MultiFactHQLContext.convertHavingToWhere(
havingAST, this, cfacts, new DefaultAliasDecider());
}
}
hqlContext = createHQLContext(cfacts, dimsToQuery, factDimMap);
return hqlContext.toHQL();
}
public boolean isAggregateExpr(String expr) {
return aggregateExprs.contains(expr == null ? null : expr.toLowerCase());
}
public boolean hasAggregates() {
return !aggregateExprs.isEmpty() || getExprCtx().hasAggregates();
}
public boolean hasDimensionInQuery() {
return dimensions != null && !dimensions.isEmpty();
}
public void addQueriedExprsWithMeasures(Set<String> exprs) {
queriedExprsWithMeasures.addAll(exprs);
}
public void addQueriedExprs(Set<String> exprs) {
queriedExprs.addAll(exprs);
}
public void addQueriedMsrs(Set<String> msrs) {
queriedMsrs.addAll(msrs);
}
public void addQueriedDimAttrs(Set<String> dimAttrs) {
queriedDimAttrs.addAll(dimAttrs);
}
public void addAggregateExpr(String expr) {
aggregateExprs.add(expr);
}