void StructuralStatsDatabase::display(OperationContext &context, ostream &out, const DictionaryDatabase *ddb) const
{
Cursor myCursor(const_cast<DbWrapper&>(db_), getTxn(context), CURSOR_WRITE, 0);
if(myCursor.error() != 0) throw XmlException(myCursor.error());
NameID id1;
NameID id2;
StructuralStats stats;
int err = 0;
while((err = myCursor.get(context.key(), context.data(), DB_NEXT)) == 0) {
id1.reset();
id2.reset();
unmarshalKey(context.key(), id1, id2);
stats.reset();
stats.unmarshal(context.data());
if(ddb) out << ddb->lookupName(context, id1);
else out << id1;
if(id2 != 0) {
if(ddb) out << " -> " << ddb->lookupName(context, id2);
else out << " -> " << id2;
}
out << ": ";
stats.display(out);
out << endl;
}
}
int StructuralStatsDatabase::getStats(OperationContext &context, StructuralStats &stats) const
{
Cursor myCursor(const_cast<DbWrapper&>(db_), getTxn(context), CURSOR_READ, 0);
if(myCursor.error() != 0) return myCursor.error();
StructuralStats current;
// Loop over every node and every node's descendant information
int err;
while((err = myCursor.get(context.key(), context.data(), DB_NEXT)) == 0) {
current.reset();
current.unmarshal(context.data());
stats.add(current);
}
if(err != DB_NOTFOUND && err != 0) return err;
if(stats.sumSize_ == 0 && stats.numberOfNodes_ != 0) {
// Fill in an estimate for the missing size values that you get with DLS
stats.sumSize_ = NODE_SIZE * stats.numberOfNodes_;
stats.sumChildSize_ = NODE_SIZE * stats.sumNumberOfChildren_;
stats.sumDescendantSize_ = NODE_SIZE * stats.sumNumberOfDescendants_;
}
return 0;
}
int DictionaryDatabase::lookupIDFromName(OperationContext &context,
const Name &name,
NameID &id, bool define)
{
int err = 0;
if (name == Name::dbxml_colon_name) {
id = nidName_;
} else if (name == Name::dbxml_colon_root) {
id = nidRoot_;
}
if (id == 0) {
MutexLock ml(mutex_);
u_int32_t flags = (getTxn(context) && !define) ? DB_READ_COMMITTED : 0;
name.setDbtFromThis_SecondaryKey(context.key());
if (stringCacheLookup(context, context.key(), id))
return 0;
err = secondary_->get(getTxn(context), &context.key(), &context.data(), flags);
if (err == 0) {
id.setThisFromDbt(context.data());
stringCache_.insert(&context.key(), id.raw());
} else if (err == DB_NOTFOUND && define) {
err = defineName(context, name, id); // define from name:uri
} else {
id.reset();
}
}
return err;
}
int DictionaryDatabase::lookupNameFromID(OperationContext &context,
const NameID &id, Name &name) const
{
int err = 0;
nameId_t raw = id.raw() - 1; // id space is 1-based, not 0
if (id == nidName_) {
name = Name::dbxml_colon_name;
return 0;
} else if (id == nidRoot_) {
name = Name::dbxml_colon_root;
return 0;
}
if ((raw < DICTIONARY_RESERVE_SIZE) && usePreloads_) {
name = preloadNames[raw];
} else {
err = lookupFromID(context, context.data(), id);
if (err == 0) {
name.setThisFromDbt(context.data());
} else {
name.reset();
}
}
return err;
}
void DictionaryDatabase::display(OperationContext &context, ostream &out) const
{
{
Cursor myCursor(const_cast<PrimaryDatabase&>(*primary_.get()), getTxn(context), CURSOR_READ, 0);
if(myCursor.error() != 0) throw XmlException(myCursor.error());
int err = 0;
NameID id;
while((err = myCursor.get(context.key(), context.data(), DB_NEXT)) == 0) {
id.setThisFromDbtAsId(context.key());
Buffer val(context.data().data, context.data().size, true);
out << id << " -> " << val.asString(true) << endl;
}
}
{
Cursor myCursor(const_cast<SecondaryDatabase&>(*secondary_.get()), getTxn(context), CURSOR_READ, 0);
if(myCursor.error() != 0) throw XmlException(myCursor.error());
int err = 0;
NameID id;
while((err = myCursor.get(context.key(), context.data(), DB_NEXT)) == 0) {
Buffer val(context.key().data, context.key().size, true);
id.setThisFromDbt(context.data());
out << val.asString(true) << " -> " << id << endl;
}
}
}
// We assume that key has the correct endianness.
int IndexDatabase::getIndexEntry(OperationContext &context, const DbXmlDbt &key, IndexEntry &ie) const
{
u_int32_t flags = (context.txn()) ? DB_READ_COMMITTED : 0;
int err = get(context.txn(), &(const_cast<DbXmlDbt&>(key)), &context.data(), flags);
if(err == 0) {
ie.setThisFromDbt(context.data());
}
return err;
}
int StructuralStatsDatabase::addStats(OperationContext &context, const StructuralStatsDatabase &sdb)
{
Cursor myCursor(db_, getTxn(context), CURSOR_WRITE, 0);
if(myCursor.error() != 0) return myCursor.error();
Cursor sdbCursor(const_cast<DbWrapper&>(sdb.db_), sdb.getTxn(context), CURSOR_READ, 0);
if(sdbCursor.error() != 0) return sdbCursor.error();
int err = 0;
StructuralStats stats;
StructuralStats current;
bool nodeStats;
while((err = sdbCursor.get(context.key(), context.data(), DB_NEXT)) == 0) {
stats.reset();
nodeStats = stats.unmarshal(context.data());
err = myCursor.get(context.key(), context.data(), DB_SET);
if(err == DB_NOTFOUND) {
stats.marshal(context.data(), nodeStats);
err = myCursor.put(context.key(), context.data(), DB_KEYLAST);
} else if(err == 0) {
current.unmarshal(context.data());
current.add(stats);
current.marshal(context.data(), nodeStats);
err = myCursor.put(context.key(), context.data(), DB_CURRENT);
} else break;
}
if(err != DB_NOTFOUND) return err;
return 0;
}
int DictionaryDatabase::lookupStringNameFromID(
OperationContext &context,
const NameID &id, const char **name) const
{
int err = 0;
nameId_t raw = id.raw() - 1; // id space is 1-based, not 0
if ((raw < DICTIONARY_RESERVE_SIZE) && usePreloads_) {
*name = preloadNames[raw];
} else {
err = lookupFromID(context, context.data(), id);
if (err == 0)
*name = (const char *)context.data().data;
}
return err;
}
int StructuralStatsDatabase::addStats(OperationContext &context, const StructuralStatsWriteCache &cache)
{
Cursor myCursor(db_, getTxn(context), CURSOR_WRITE, 0);
if(myCursor.error() != 0) return myCursor.error();
int err = 0;
StructuralStats current;
StructuralStatsWriteCache::Map::const_iterator it = cache.cache_.begin();
for(; err == 0 && it != cache.cache_.end(); ++it) {
StructuralStatsWriteCache::InnerMap::const_iterator it2 = it->second.begin();
for(; it2 != it->second.end(); ++it2) {
marshalKey(it->first, it2->first, context.key());
err = myCursor.get(context.key(), context.data(), DB_SET);
if(err == DB_NOTFOUND) {
it2->second.marshal(context.data(), it2->first == 0);
err = myCursor.put(context.key(), context.data(), DB_KEYLAST);
} else if(err == 0) {
current.unmarshal(context.data());
current.add(it2->second);
current.marshal(context.data(), it2->first == 0);
err = myCursor.put(context.key(), context.data(), DB_CURRENT);
} else break;
}
}
if(err != DB_NOTFOUND) return err;
return 0;
}
int StructuralStatsDatabase::subtractStats(OperationContext &context, const NameID &id1, const NameID &id2,
const StructuralStats &stats)
{
DBXML_ASSERT(id1 != 0 || id2 == 0);
Cursor myCursor(const_cast<DbWrapper&>(db_), getTxn(context), CURSOR_WRITE, 0);
if(myCursor.error() != 0) return myCursor.error();
marshalKey(id1, id2, context.key());
int err = myCursor.get(context.key(), context.data(), DB_SET);
if(err != DB_NOTFOUND && err != 0) return err;
StructuralStats current;
if(err == DB_NOTFOUND) {
current.subtract(stats);
current.marshal(context.data(), id2 == 0);
err = myCursor.put(context.key(), context.data(), DB_KEYLAST);
} else {
current.unmarshal(context.data());
current.subtract(stats);
current.marshal(context.data(), id2 == 0);
err = myCursor.put(context.key(), context.data(), DB_CURRENT);
}
return err;
}
int StructuralStatsDatabase::getStats(OperationContext &context, const NameID &id1, StructuralStats &stats) const
{
if(id1 == 0) return getStats(context, stats);
Cursor myCursor(const_cast<DbWrapper&>(db_), getTxn(context), CURSOR_READ, 0);
if(myCursor.error() != 0) return myCursor.error();
NameID cid1, cid2;
StructuralStats current;
// Find the node information
marshalKey(id1, 0, context.key());
int err = myCursor.get(context.key(), context.data(), DB_SET);
if(err == DB_NOTFOUND) return getStats(context, stats);
// Loop over the node information and all it's descendant information
while(err == 0) {
cid1.reset();
unmarshalKey(context.key(), cid1, cid2);
if(id1 != cid1) break;
current.reset();
current.unmarshal(context.data());
stats.add(current);
err = myCursor.get(context.key(), context.data(), DB_NEXT);
}
if(err != DB_NOTFOUND && err != 0) return err;
if(stats.sumSize_ == 0 && stats.numberOfNodes_ != 0) {
// Fill in an estimate for the missing size values that you get with DLS
stats.sumSize_ = NODE_SIZE * stats.numberOfNodes_;
stats.sumChildSize_ = NODE_SIZE * stats.sumNumberOfChildren_;
stats.sumDescendantSize_ = NODE_SIZE * stats.sumNumberOfDescendants_;
}
return 0;
}
int DictionaryDatabase::lookupIDFromStringNameInternal(
OperationContext &context,
DbXmlDbt &dbt,
NameID &id,
bool define) const
{
if (!dbt.size) {
id.reset();
return 0;
}
u_int32_t flags = (getTxn(context) && !define) ? DB_READ_COMMITTED : 0;
if (stringCacheLookup(context, dbt, id))
return 0;
int err = secondary_->get(getTxn(context), &dbt,
&context.data(), flags);
if (err == 0) {
id.setThisFromDbt(context.data());
stringCache_.insert(&dbt, id.raw());
} else
id.reset();
return err;
}
int StructuralStatsDatabase::getStats(OperationContext &context, const NameID &id1, const NameID &id2,
StructuralStats &stats) const
{
if(id1 == 0) return getStats(context, stats);
if(id2 == 0) return getStats(context, id1, stats);
Cursor myCursor(const_cast<DbWrapper&>(db_), getTxn(context), CURSOR_READ, 0);
if(myCursor.error() != 0) return myCursor.error();
// Lookup the descendant information
marshalKey(id1, id2, context.key());
int err = myCursor.get(context.key(), context.data(), DB_SET);
if(err == DB_NOTFOUND) return getStats(context, id1, stats);
if(err != 0) return err;
stats.unmarshal(context.data());
// Lookup the node information
marshalKey(id1, 0, context.key());
err = myCursor.get(context.key(), context.data(), DB_SET);
DBXML_ASSERT(err != DB_NOTFOUND);
if(err != 0) return err;
stats.unmarshal(context.data());
if(stats.sumSize_ == 0 && stats.numberOfNodes_ != 0) {
// Fill in an estimate for the missing size values that you get with DLS
StructuralStats allStats;
err = getStats(context, id1, allStats);
if(err != 0) return err;
stats.sumSize_ = allStats.sumSize_;
stats.sumChildSize_ = allStats.sumChildSize_;
stats.sumDescendantSize_ = allStats.sumDescendantSize_;
}
return 0;
}
int SyntaxDatabase::updateStatistics(OperationContext &context, DbtIn &key, const KeyStatistics &statistics)
{
Cursor cursor(*statistics_.get(), context.txn(), CURSOR_WRITE);
int err = cursor.get(key, context.data(), DB_SET | (context.txn() ? DB_RMW : 0)); // could throw on error
if (err == 0) {
KeyStatistics existing;
existing.setThisFromDbt(context.data());
existing.add(statistics);
existing.setDbtFromThis(context.data());
err = cursor.put(key, context.data(), DB_CURRENT); // could throw on error
} else if (err == DB_NOTFOUND) {
statistics.setDbtFromThis(context.data());
err = cursor.put(key, context.data(), DB_KEYFIRST); // could throw on error
}
return err;
}
// mutex is locked if present and necessary
int DictionaryDatabase::defineName(OperationContext &context,
const Name &name, NameID &id)
{
// Primary { id -> name\0uri\0 }
// Secondary { name:uri -> id }
id.reset();
name.setDbtFromThis_PrimaryValue(context.data());
int err = primary_->appendPrimary(context, id, context.data(), /*no flags*/0);
if (err == 0) {
/*Cache after insert. This prevents the cache from returning bad answers
* on a transaction abort, and speeds up access to newly inserted element
* and attribute names that are needed for indexing after an insert.
*/
cache_.insert(id.raw(), context.data());
id.setDbtFromThis(context.key());
name.setDbtFromThis_SecondaryKey(context.data());
Transaction *txn = getTxn(context);
err = secondary_->put(txn, &context.data(), &context.key(), /*no flags*/0);
if (err == 0) {
/* Add to the transaction's string cache, that way if an abort occurs
* the values can be removed from the cache.
*/
if (txn) {
DictionaryStringCache *dsc = txn->getStringCache(this, true);
dsc->insert(&context.data(), id.raw());
}
if(Log::isLogEnabled(Log::C_DICTIONARY, Log::L_INFO)) {
ostringstream oss;
oss << "Define new name " << id << " -> " << name;
Log::log(environment_, Log::C_DICTIONARY, Log::L_INFO,
name_.c_str(),
oss.str().c_str());
}
}
}
return err;
}
double IndexDatabase::percentage(OperationContext &context, Operation operation, Operation gto, Operation lto, const Key &key1, const Key &key2) const
{
DbtOut &dbt1 = context.key();
DbtOut &dbt2 = context.data();
DB_KEY_RANGE krMin;
getMinKeyDbt(key1, dbt1);
key_range(context.txn(), &dbt1, &krMin, 0);
DB_KEY_RANGE krMax;
getMaxKeyDbt(key1, dbt1);
key_range(context.txn(), &dbt1, &krMax, 0);
// range is the % of the database keys that the keys for this index occupy.
double range = krMax.less - krMin.less;
double extent = 0.0;
if (range > 0.0) {
// extent is the % of the database keys that the keys for this index match this operation.
DB_KEY_RANGE kr1;
DB_KEY_RANGE kr2;
switch(operation) {
case DbWrapper::PREFIX: {
key1.setDbtFromThis(dbt1);
key_range(context.txn(), &dbt1, &kr1, 0);
getNextKeyDbt(key1, dbt2);
key_range(context.txn(), &dbt2, &kr2, 0);
extent = kr2.less - kr1.less;
break;
}
case DbWrapper::LTX:
case DbWrapper::LTE: {
key1.setDbtFromThis(dbt2);
key_range(context.txn(), &dbt2, &kr2, 0);
extent = kr2.less - krMin.less + (operation == DbWrapper::LTE ? kr2.equal : 0);
break;
}
case DbWrapper::GTX:
case DbWrapper::GTE: {
key1.setDbtFromThis(dbt1);
key_range(context.txn(), &dbt1, &kr1, 0);
extent = krMax.less + krMax.equal - kr1.less + (operation == DbWrapper::GTX ? kr1.equal : 0);
break;
}
case DbWrapper::RANGE: {
key1.setDbtFromThis(dbt1);
key_range(context.txn(), &dbt1, &kr1, 0);
key2.setDbtFromThis(dbt2);
key_range(context.txn(), &dbt2, &kr2, 0);
extent = kr2.less - kr1.less + (lto == DbWrapper::LTE ? kr2.equal : 0) + (gto == DbWrapper::GTX ? kr1.equal : 0);
break;
}
case DbWrapper::EQUALITY: {
key1.setDbtFromThis(dbt2);
key_range(context.txn(), &dbt2, &kr2, 0);
extent = kr2.equal;
break;
}
case DbWrapper::ALL: {
extent = range;
break;
}
default: {
break;
}
}
}
// Return a small percentage in the case of a zero range or extent -
// it's unlikely that zero is really the right answer
if(range == 0 || extent == 0) return 0.001;
// extent/range is the % of keys within this index that match this operation.
return extent / range;
}