ExplainTuple *PhysicalFastExtract::addSpecificExplainInfo(ExplainTupleMaster *explainTuple, ComTdb *tdb,
Generator *generator)
{
NAString description = "Target_type: ";
if (getTargetType() == FILE)
{
if (isHiveInsert())
description += "hive table";
else
description += "file";
}
else if (getTargetType() == SOCKET)
description += "socket";
else
description += "none";
if (isHiveInsert())
{
NAString str = getTargetName();
size_t colonIndex = str.index(":", 1,0,NAString::ignoreCase);
while (colonIndex != NA_NPOS)
{
str = str.replace(colonIndex, 1, "_", 1);
colonIndex = str.index(":", 1,0,NAString::ignoreCase);
}
description += " location: ";
description += str;
}
if (isHiveInsert())
{
description += " table_name: ";
description += getHiveTableName();
}
else
{
description += " target_name: ";
description += getTargetName();
}
description += " delimiter: ";
description += getDelimiter();
if (isAppend())
description += " append: yes";
if ( !isHiveInsert() && includeHeader())
{
description += " header: ";
description += getHeader();
}
if (getCompressionType() != NONE)
{
description += " compression_type: ";
if (getCompressionType() == LZO)
description += "LZO";
else
description += "error";
}
description += " null_string: ";
description += getNullString();
description += " record_separator: ";
description += getRecordSeparator();
explainTuple->setDescription(description);
if (isHiveInsert())
explainTuple->setTableName(getHiveTableName());
return explainTuple;
}
Lng32 AddKeyGroups()
{
HSGlobalsClass *hs_globals = GetHSContext();
if (HSGlobalsClass::isHiveCat(hs_globals->objDef->getCatName()))
{
// HSHiveTableDef::getKeyList()/getIndexArray() not yet implemented.
*CmpCommon::diags() << DgSqlCode(-UERR_NO_ONEVERYKEY) << DgString0("hive");
return -1;
}
Lng32 retcode = 0;
Lng32 numColsInGroup = 0;
HSColumnStruct col;
NAString tempColList = "";
NAString tempCol;
NAString autoGroup;
ULng32 numKeys;
ULng32 i, j;
NATable* naTbl = hs_globals->objDef->getNATable();
HSLogMan *LM = HSLogMan::Instance();
// ----------------------------------------------------------
// Generate histograms for KEY
// ----------------------------------------------------------
// The clustering index is included in the list of indices returned by
// NATable::getIndexList(), so we store its pointer so we can skip it
// when the other indexes are processed below.
NAFileSet* clusteringIndex = naTbl->getClusteringIndex();
const NAColumnArray& keyCols = clusteringIndex->getIndexKeyColumns();
Lng32 colPos;
numKeys = keyCols.entries();
if (numKeys == 1) // SINGLE-COLUMN KEY
{
colPos = keyCols[0]->getPosition();
if (LM->LogNeeded())
{
sprintf(LM->msg, "\t\tKEY:\t\t(%s)", hs_globals->objDef->getColName(colPos));
LM->Log(LM->msg);
}
if (ColumnExists(colPos)) // avoid duplicates
{
LM->Log("\t\t** duplicate column group has been ignored.");
}
else // add to single-column group list
{
retcode = AddSingleColumn(colPos);
}
}
else if (numKeys > 1) // MULTI-COLUMN KEY
{
// Create multiple MC group(s) if numkeys > 1. Subset MC groups will
// also be created if numkeys > 2, E.g. If numkeys = 5, then
// MC groups with 5, 4, 3, and 2 columns will be created using
// the key columns. Note that if numkeys is larger than CQD
// USTAT_NUM_MC_GROUPS_FOR_KEYS (default = 5), then the number
// of groups created will be limited by this value. So, e.g. if
// numkeys = 10, then MC groups with 5, 4, 3, and 2 columns will
// be created (that is, 5 groups will be created - incl the single).
ULng32 minMCGroupSz = 2;
ULng32 maxMCGroups = (ULng32)
CmpCommon::getDefaultNumeric(USTAT_NUM_MC_GROUPS_FOR_KEYS);
// Generate no MCs with more cols than specified by the cqd.
if (numKeys > maxMCGroups)
numKeys = maxMCGroups;
// For salted table, generate only the longest MC for the key (subject
// to max cols determined above) unless a cqd is set to gen all MCs of
// allowable sizes.
if (CmpCommon::getDefault(USTAT_ADD_SALTED_KEY_PREFIXES_FOR_MC) == DF_OFF &&
hs_globals->objDef->getColNum("_SALT_", FALSE) >= 0)
minMCGroupSz = numKeys;
while (numKeys >= minMCGroupSz) // Create only MC groups not single cols
{
HSColSet colSet;
autoGroup = "(";
for (j = 0; j < numKeys; j++)
{
colPos = keyCols[j]->getPosition();
col = hs_globals->objDef->getColInfo(colPos);
col.colnum = colPos;
colSet.insert(col);
autoGroup += col.colname->data();
autoGroup += ",";
}
if (LM->LogNeeded())
{
autoGroup.replace(autoGroup.length()-1,1,")"); // replace comma with close parenthesis
sprintf(LM->msg, "\t\tKEY:\t\t%s", autoGroup.data());
LM->Log(LM->msg);
}
if (retcode = AddColumnSet(colSet))
{
HSHandleError(retcode);
}
numKeys--;
}
}
// ----------------------------------------------------------
// Generate histograms for all INDEXES
// ----------------------------------------------------------
const NAFileSetList& indexes = naTbl->getIndexList();
NAFileSet* index;
for (i = 0; i < indexes.entries(); i++ )
{
index = indexes[i];
if (index == clusteringIndex)
continue; // clustering index processed above already
const NAColumnArray& keyCols = index->getIndexKeyColumns();
numKeys = keyCols.entries();
if (numKeys == 1) // SINGLE-COLUMN INDEX
{
colPos = keyCols[0]->getPosition();
if (LM->LogNeeded())
{
sprintf(LM->msg, "\t\tINDEX[%d]\t(%s)", i,
hs_globals->objDef->getColName(colPos));
LM->Log(LM->msg);
}
if (ColumnExists(colPos)) // avoid duplicates
{
LM->Log("\t\t*** duplicate column group has been ignored.");
}
else // add to single-column group list
{
retcode = AddSingleColumn(colPos);
}
}
else // MULTI-COLUMN INDEX
{
// Create multiple MC group(s) if numkeys > 1. Subset MC groups will
// also be created if numkeys > 2, E.g. If numkeys = 5, then
// MC groups with 5, 4, 3, and 2 columns will be created using
// the key columns. Note that if numkeys is larger than CQD
// USTAT_NUM_MC_GROUPS_FOR_KEYS (default = 5), then the number
// of groups created will be limited by this value. So, e.g. if
// numkeys = 10, then MC groups with 10, 9, 8, 7, 6 columns will
// be created (that is, 5 groups will be created).
ULng32 minMCGroupSz = 2;
ULng32 maxMCGroups = (ULng32)
CmpCommon::getDefaultNumeric(USTAT_NUM_MC_GROUPS_FOR_KEYS);
if (numKeys > maxMCGroups)
minMCGroupSz = numKeys - maxMCGroups + 1;
while (numKeys >= minMCGroupSz) // MinMCGroupSz is greater than 1.
{
HSColSet colSet;
tempColList = "";
autoGroup = "(";
for (j = 0; j < numKeys; j++)
{
colPos = keyCols[j]->getPosition();
tempCol = ".";
tempCol += LongToNAString(colPos);
tempCol += ".";
// Eliminate duplicate columns in the index;
// They may have been introduced by appending the key to the specified index.
if (!tempColList.contains(tempCol))
{
col = hs_globals->objDef->getColInfo(colPos);
col.colnum = colPos;
colSet.insert((const struct HSColumnStruct) col);
tempColList += tempCol.data();
numColsInGroup++;
autoGroup += col.colname->data();
autoGroup += ",";
}
}
if (colSet.entries())
{
if (numColsInGroup > 1)
{
if (LM->LogNeeded())
{
autoGroup.replace(autoGroup.length()-1,1,")"); // replace comma with close parenthesis
sprintf(LM->msg, "\t\tINDEX[%d]\t%s", i, autoGroup.data());
LM->Log(LM->msg);
}
if (retcode = AddColumnSet(colSet))
{
HSHandleError(retcode);
}
}
numColsInGroup = 0;
}
numKeys--;
}
}
}
return retcode;
}