ibis::query Class Reference

A data structure for representing user queries. More...

#include <query.h>

Inheritance diagram for ibis::query:

Classes
class	readLock
	A read lock on a query object. More...
class	result
	The class ibis::query::result allows user to retrieve query result one row at a time. More...
class	weight
	A class to be used for reordering the terms in the where clauses. More...
class	writeLock
	A write lock on a query object. More...

Public Types
enum	QUERY_STATE {   UNINITIALIZED, SET_COMPONENTS, SET_RIDS, SET_PREDICATE,   SPECIFIED, QUICK_ESTIMATE, FULL_EVALUATE, BUNDLES_TRUNCATED,   HITS_TRUNCATED }

Public Member Functions
int	addConditions (const ibis::qExpr *qexp)
	Add a set of conditions to the existing where clause. More...
int	addConditions (const char *)
	Add a set of conditions to the existing where clause. More...
void	clear ()
	Releases the resources held by the query object. More...
void	clearErrorMessage () const
	Reset the last error message to blank.
const selectClause &	components () const
	Return a list of names specified in the select clause.
void	contractQuery ()
	Contracts where clause to preferred bounds. More...
long	countHits () const
	Count the number of hits. More...
const char *	dir () const
	Return the directory for any persistent data. More...
int	estimate ()
	Function to perform estimation. More...
int	evaluate (const bool evalSelect=false)
	Computes the exact hits. More...
void	expandQuery ()
	Expands where clause to preferred bounds. More...
long	getCandidateRows (std::vector< uint32_t > &) const
	Extract the positions of candidates. More...
const ibis::bitvector *	getCandidateVector () const
	Return a pointer to the bit vector representing the candidates.
long	getExpandedHits (ibis::bitvector &) const
	Get a bitvector containing all rows satisfying the query condition. More...
long	getHitRows (std::vector< uint32_t > &rids) const
	Extract the positions of the bits that are 1s in the solution. More...
const ibis::bitvector *	getHitVector () const
	Return the pointer to the internal hit vector. More...
const char *	getLastError () const
	Return the last error message recorded internally.
long	getMaxNumHits () const
	Return the number of records in the upper bound.
long	getMinNumHits () const
	Return the number of records in the lower bound.
long	getNumHits () const
	Compute the number of records in the exact solution. More...
virtual const char *	getSelectClause () const
	Return the select clause string.
QUERY_STATE	getState () const
	Return the current state of query.
const RIDSet *	getUserRIDs () const
	Return a const pointer to the copy of the user supplied RID set.
virtual const char *	getWhereClause () const
	Return the where clause string.
const char *	id () const
	Return an identifier of the query.
long	limit (const char *names, uint32_t keep, bool updateHits=true)
	Truncate the results to provide the top-K rows. More...
void	logMessage (const char *event, const char *fmt,...) const
	Used to print information about the progress or state of query processing. More...
int	orderby (const char *names) const
	Re-order the results according to the new "ORDER BY" specification. More...
const part *	partition () const
	Return the pointer to the data partition used to process the query.
void	printSelected (std::ostream &out) const
	Print the values of the selected columns to the specified output stream. More...
void	printSelectedWithRID (std::ostream &out) const
	Print the values of the columns in the select clause without functions. More...
	query (const char *dir, const ibis::partList &tl)
	Constructor. More...
	query (const char *uid=0, const part *et=0, const char *pref=0)
	Constructor. More...
RIDSet *	readRIDs () const
	Read RIDs from the file named "-rids". More...
std::string	removeComplexConditions ()
	Separate out the sub-expressions that are not simple. More...
long	sequentialScan (ibis::bitvector &bv) const
	Perform a simple sequential scan. More...
int	setPartition (const ibis::part *tbl)
	Resets the data partition associated with the query. More...
int	setRIDs (const RIDSet &set)
	Specify a list of Row IDs for the query object. More...
virtual int	setSelectClause (const char *str)
	Specifies the select clause for the query. More...
int	setTable (const ibis::part *tbl)
	This is deprecated, will be removed soon.
int	setWhereClause (const char *str)
	Specify the where clause in the string form. More...
int	setWhereClause (const std::vector< const char * > &names, const std::vector< double > &lbounds, const std::vector< double > &rbounds)
	Specify the where clause as a set of conjunctive ranges. More...
int	setWhereClause (const ibis::qExpr *qexp)
	Specify the where clause through a qExpr object. More...
time_t	timestamp () const
	The time stamp on the data used to process the query.
const char *	userName () const
	User started the query.
void	writeRIDs (const RIDSet *rids) const
	Write the list of RIDs to a file named "-rids".
virtual	~query ()
	Desctructor.
array_t< signed char > *	getQualifiedBytes (const char *column_name)
	The functions getQualifiedTTT return the values of selected columns in the records that satisfies the specified conditions. More...
array_t< unsigned char > *	getQualifiedUBytes (const char *column_name)
	Retrieve the values of column_name as 8-bit unsigned integers. More...
array_t< int16_t > *	getQualifiedShorts (const char *column_name)
	Retrieve the values of column_name as 16-bit integers. More...
array_t< uint16_t > *	getQualifiedUShorts (const char *column_name)
	Retrieve the values of column_name as 16-bit unsigned integers. More...
array_t< int32_t > *	getQualifiedInts (const char *column_name)
	Retrieve integer values from records satisfying the query conditions. More...
array_t< uint32_t > *	getQualifiedUInts (const char *column_name)
	Retrieve unsigned integer values from records satisfying the query conditions. More...
array_t< int64_t > *	getQualifiedLongs (const char *column_name)
	Retrieve values of column_name as 64-bit integers. More...
array_t< uint64_t > *	getQualifiedULongs (const char *column_name)
	Retrieve values of column_name as 64-bit unsigned integers. More...
array_t< float > *	getQualifiedFloats (const char *column_name)
	Retrieve floating-point values from records satisfying the query conditions. More...
array_t< double > *	getQualifiedDoubles (const char *column_name)
	Retrieve double precision floating-point values from records satisfying the query conditions. More...
std::vector< std::string > *	getQualifiedStrings (const char *column_name)
	Retrieve string values from records satisfying the query conditions. More...
RIDSet *	getRIDs () const
	Return the list of row IDs of the hits. More...
RIDSet *	getRIDs (const ibis::bitvector &mask) const
	Return a list of row IDs that match the mask. More...
const RIDSet *	getRIDsInBundle (const uint32_t bid) const
	Return the list of row IDs of the hits within the specified bundle.
template<>
int64_t	countEqualPairs (const array_t< int32_t > &val1, const array_t< uint32_t > &val2) const
template<>
int64_t	countEqualPairs (const array_t< uint32_t > &val1, const array_t< int32_t > &val2) const
template<>
int64_t	countDeltaPairs (const array_t< uint32_t > &val1, const array_t< int32_t > &val2, const uint32_t &delta) const
template<>
int64_t	countDeltaPairs (const array_t< int32_t > &val1, const array_t< uint32_t > &val2, const int32_t &delta) const
template<>
int64_t	recordEqualPairs (const array_t< uint32_t > &val1, const array_t< int32_t > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const char *filename) const
template<>
int64_t	recordEqualPairs (const array_t< int32_t > &val1, const array_t< uint32_t > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const char *filename) const
template<>
int64_t	recordDeltaPairs (const array_t< uint32_t > &val1, const array_t< int32_t > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const uint32_t &delta, const char *filename) const
template<>
int64_t	recordDeltaPairs (const array_t< int32_t > &val1, const array_t< uint32_t > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const int32_t &delta, const char *filename) const

Static Public Member Functions
static bool	isValidToken (const char *tok)
	Is the given string a valid query token. More...
static void	keepQueryRecords ()
	Tell the destructor to leave stored information on disk.
static void	removeQueryRecords ()
	Tell the destructor to remove all stored information about queries.
static unsigned	tokenLength ()
	Length of the query token.

Protected Member Functions
int	computeHits ()
	Generate the hit vector. More...
template<typename T1 , typename T2 >
int64_t	countDeltaPairs (const array_t< T1 > &val1, const array_t< T2 > &val2, const T1 &delta) const
	Assume the two input arrays are sorted in ascending order, count the number of elements that are with delta of each other. More...
template<typename T1 , typename T2 >
int64_t	countEqualPairs (const array_t< T1 > &val1, const array_t< T2 > &val2) const
	Assume the two input arrays are sorted in ascending order, count the number of elements that match. More...
uint32_t	countPages (unsigned wordsize) const
	Count the number of pages accessed to retrieve every value in the hit vector. More...
int	doContract (ibis::qExpr *exp0) const
	Contract range conditions to remove the need of candidate check.
void	doEstimate (const qExpr *term, ibis::bitvector &low, ibis::bitvector &high) const
	Use index only to come up with a upper bound and a lower bound. More...
int	doEvaluate (const qExpr *term, ibis::bitvector &hits) const
	Evaluate the query expression. More...
int	doEvaluate (const qExpr *term, const ibis::bitvector &mask, ibis::bitvector &hits) const
	Evaluate the query expression with mask. More...
int	doExpand (ibis::qExpr *exp0) const
	Expand range conditions to remove the need of candidate check.
int	doScan (const qExpr *term, const ibis::bitvector &mask, ibis::bitvector &hits) const
	Masked sequential scan. More...
int	doScan (const qExpr *term, ibis::bitvector &hits) const
	Perform a sequential scan. More...
void	gainReadAccess (const char *mesg) const
void	gainWriteAccess (const char *mesg) const
void	getBounds ()
	Compute the upper and lower bounds for range queries.
bool	hasBundles () const
void	logError (const char *event, const char *fmt,...) const
	!< The warning/error message
void	logWarning (const char *event, const char *fmt,...) const
int64_t	mergePairs (const char *pairfile) const
void	orderPairs (const char *pairfile) const
	Sort the content of the file as ibis::rid_t. More...
void	printRIDs (const RIDSet &ridset) const
	Export the Row IDs of the hits to log file. More...
int64_t	processJoin ()
	Process the join operation and return the number of pairs. More...
void	readHits ()
	Read the results of the query.
void	readQuery (const ibis::partList &tl)
	Read the status information from disk. More...
template<typename T1 , typename T2 >
int64_t	recordDeltaPairs (const array_t< T1 > &val1, const array_t< T2 > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const T1 &delta, const char *pairfile) const
template<typename T1 , typename T2 >
int64_t	recordEqualPairs (const array_t< T1 > &val1, const array_t< T2 > &val2, const array_t< uint32_t > &ind1, const array_t< uint32_t > &ind2, const char *pairfile) const
void	releaseAccess (const char *mesg) const
void	removeFiles ()
	Remove the files written by this object.
void	reorderExpr ()
int64_t	sortEquiJoin (const ibis::deprecatedJoin &cmp, const ibis::bitvector &mask) const
	Performing an equi-join by sorting the selected values first. More...
int64_t	sortEquiJoin (const ibis::deprecatedJoin &cmp, const ibis::bitvector &mask, const char *pairfile) const
	Perform equi-join by sorting the selected values. More...
int64_t	sortJoin (const std::vector< const ibis::deprecatedJoin * > &terms, const ibis::bitvector &mask) const
int64_t	sortJoin (const ibis::deprecatedJoin &cmp, const ibis::bitvector &mask) const
int64_t	sortRangeJoin (const ibis::deprecatedJoin &cmp, const ibis::bitvector &mask) const
	Performing a range join by sorting the selected values. More...
int64_t	sortRangeJoin (const ibis::deprecatedJoin &cmp, const ibis::bitvector &mask, const char *pairfile) const
	Performing range join by sorting the selected values.
void	storeErrorMesg (const char *) const
	Store the message into member variable lastError for later use. More...
void	writeHits () const
	Write the results of the query.
virtual void	writeQuery ()
	Write the basic information about the query to disk. More...

Protected Attributes
selectClause	comps
	!< Query conditions
whereClause	conds
	!< Name of the user who specified the query
ibis::part::readLock *	dslock
	!< Estimated upper bound
ibis::bitvector *	hits
	!< Status of the query
char	lastError [MAX_LINE+PATH_MAX]
	!< A read lock on the mypart
QUERY_STATE	state
	!< Select clause
ibis::bitvector *	sup
	!< Solution in bitvector form (or lower bound)
char *	user

Friends
class	readLock
class	writeLock

Detailed Description

A data structure for representing user queries.

This is the primary entry for user to take advantage of bitmap indexing facilities. A query is a very limited version of the SQL SELECT statement. It is only defined on one data partition and it takes a where clause and a select clause. The where clause is mandatory!

It contains a list of range conditions joined together with logical operators, such as "temperature > 700 and 100 <= presessure < 350". Records whose values satisfy the conditions defined in the where clause is considered hits. A query may retrieve values of variables/columns specified in the select clause. A select clause is optional. If specified, it contains a list of column names. These columns must not be NULL in order for a record to be a hit. If any additional functions are needed in the select clause, use the function ibis::table::select instead of using this class.

The hits can be computed in two ways by using functions estimate or evaluate. The function estimate can take advantage of the indices to give two approximate solutions, one as an upper bound and the other as a lower bound. The bitmap indices will be automatically built according to the specification if they are not present. The accuracy of the bounds depend on the nature of the indices available. If no index can be constructed, the lower bound would be empty and the upper bound would include every record. When the function evaluate is called, the exact solution is computed no matter whether the function estimate has been called or not. The solution produced is recorded as a bit vector. The user may use ibis::bitvector::indexSet to extract the record numbers of the hits or use one of the functions getQualifiedInts, getQualifiedFloats, and getQualifiedDoubles to retrieve the values of the selected columns. Additionally, one may call either printSelected or printSelectedWithRID to print the selected values to the specified I/O stream.

Member Enumeration Documentation

enum ibis::query::QUERY_STATE

Enumerator
UNINITIALIZED	The query object is currently empty.
SET_COMPONENTS	The query object has a select clause.
SET_RIDS	The query object contains a list of RIDs.
SET_PREDICATE	The query object has a where clause.
SPECIFIED	SET_COMPONENTS & (SET_RIDS | SET_PREDICATE).
QUICK_ESTIMATE	A upper and a lower bound are computed.
FULL_EVALUATE	The exact hits are computed.
BUNDLES_TRUNCATED	Only top-K results are stored.
HITS_TRUNCATED	The hit vector has been updated to match bundles.

Constructor & Destructor Documentation

ibis::query::query	(	const char *	dir,
		const ibis::partList &	tl
	)

Constructor.

Reconstructs query from stored information in the named directory dir. This is only used for recovering from program crashes, not intended for user to manually construct a query in a directory.

Note

that to enable recovery, the query objects must be constructed with the recovery feature, which is enabled through a configuration parameter prefix.enableRecovery = true.

References comps, conds, dir(), ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), hits, QUICK_ESTIMATE, readHits(), readQuery(), SET_COMPONENTS, SET_PREDICATE, SET_RIDS, SPECIFIED, state, ibis::util::strnewdup(), sup, and UNINITIALIZED.

ibis::query::query	(	const char *	uid = 0,
		const part *	et = 0,
		const char *	pref = 0
	)

Constructor.

Generates a new query on the given data partition.

• uid the user name to be associated with the query object. If this arguemnt is a nil pointer, this function will call ibis::util::userName to determine the current user name.

• et the data partition to be used for queried. The data partition to be queried could be altered with the function setPartition.

• pref a special prefix to be used for this query. This prefix is used primarily to identify the query object and to retrieve configuration parameters that are intended for a special class of queries. For example, if recovery is desired, the user can define pref.enableRecovery = true. When this feature is enable, it is possible to also define pref.purgeTempFiles = true to tell the destructor of this class to remove the log file about a query. The default value if pref is a nil pointer, which disables the logging feature.

References ibis::gParameters(), lastError, and ibis::util::strnewdup().

Member Function Documentation

int ibis::query::addConditions

(

const ibis::qExpr *

qx

)

Add a set of conditions to the existing where clause.

The new query expression is joined with the existing conditions with the AND operator.

Note

This object will have a copy of the incoming query expression.

References ibis::whereClause::addExpr(), BUNDLES_TRUNCATED, comps, conds, ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, ibis::part::name(), ibis::qExpr::nItems(), QUICK_ESTIMATE, removeFiles(), ibis::whereClause::resetString(), SET_COMPONENTS, SET_PREDICATE, SPECIFIED, state, sup, UNINITIALIZED, ibis::whereClause::verify(), and writeQuery().

int ibis::query::addConditions

(

const char *

qx

)

Add a set of conditions to the existing where clause.

The new query expression is joined with the existing conditions with the AND operator.

References ibis::whereClause::addConditions(), BUNDLES_TRUNCATED, comps, conds, ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, ibis::part::name(), ibis::qExpr::nItems(), QUICK_ESTIMATE, removeFiles(), ibis::whereClause::resetString(), SET_COMPONENTS, SET_PREDICATE, SPECIFIED, state, sup, UNINITIALIZED, ibis::whereClause::verify(), and writeQuery().

void ibis::query::clear

(

)

Releases the resources held by the query object.

It re-initializes the select clause and the where clause to blank.

References ibis::fileManager::flushDir(), ibis::gParameters(), ibis::fileManager::instance(), and ibis::util::removeDir().

int ibis::query::computeHits ( )

protected

Generate the hit vector.

Make sure mypart is set before calling this function.

References ibis::bitvector::compress(), ibis::bitvector::indexSet::indices(), ibis::bitvector::indexSet::isRange(), ibis::bitvector::indexSet::nIndices(), and ibis::bitvector::size().

Referenced by evaluate().

void ibis::query::contractQuery

(

)

Contracts where clause to preferred bounds.

Similar to function exandQuery, but makes the bounds of the range conditions narrower rather than wider.

References BUNDLES_TRUNCATED, comps, conds, doContract(), dslock, ibis::selectClause::empty(), ibis::whereClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), hits, HITS_TRUNCATED, QUICK_ESTIMATE, removeFiles(), ibis::whereClause::resetString(), SPECIFIED, state, sup, and writeQuery().

template<typename T1 , typename T2 >

int64_t ibis::query::countDeltaPairs ( const array_t< T1 > &  val1, const array_t< T2 > &  val2, const T1 &  delta  ) const

protected

Assume the two input arrays are sorted in ascending order, count the number of elements that are with delta of each other.

Note that both template arguments should be elemental types or they must support operators -, +, == and < with mixed types.

References ibis::array_t< T >::size().

int64_t ibis::query::countDeltaPairs	(	const array_t< uint32_t > &	val1,
		const array_t< int32_t > &	val2,
		const uint32_t &	delta
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), and ibis::array_t< T >::size().

int64_t ibis::query::countDeltaPairs	(	const array_t< int32_t > &	val1,
		const array_t< uint32_t > &	val2,
		const int32_t &	delta
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), and ibis::array_t< T >::size().

template<typename T1 , typename T2 >

int64_t ibis::query::countEqualPairs ( const array_t< T1 > &  val1, const array_t< T2 > &  val2  ) const

protected

Assume the two input arrays are sorted in ascending order, count the number of elements that match.

Note that both template arguments should be elemental types or they must support operators == and < with mixed types.

References ibis::array_t< T >::size().

int64_t ibis::query::countEqualPairs	(	const array_t< int32_t > &	val1,
		const array_t< uint32_t > &	val2
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), and ibis::array_t< T >::size().

int64_t ibis::query::countEqualPairs	(	const array_t< uint32_t > &	val1,
		const array_t< int32_t > &	val2
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), and ibis::array_t< T >::size().

long ibis::query::countHits

(

)

const

Count the number of hits.

Don't generate the hit vector if not already there. It only work for queries containing a single range condition. Furthermore, this function does not obtain a read lock on the query or the partition. Therefore it is possible for another thread to modify the query object while the evaluation is in progress.

References ibis::bitvector::cnt(), conds, ibis::part::countHits(), ibis::whereClause::empty(), ibis::whereClause::getExpr(), hits, ibis::part::nRows(), and sup.

uint32_t ibis::query::countPages ( unsigned  wordsize ) const

protected

Count the number of pages accessed to retrieve every value in the hit vector.

References ibis::bitvector::indexSet::indices(), ibis::bitvector::indexSet::isRange(), ibis::bitvector::indexSet::nIndices(), and ibis::fileManager::pageSize().

Referenced by evaluate().

const char* ibis::query::dir ( ) const

inline

Return the directory for any persistent data.

This is not nil only if the recovery feature is enabled. By default, the recovery feature is disabled.

Referenced by ibis::bundle1::bundle1(), ibis::bundles::bundles(), query(), ibis::bundle1::write(), and ibis::bundles::write().

void ibis::query::doEstimate ( const qExpr *  term, ibis::bitvector &  low, ibis::bitvector &  high  ) const

protected

Use index only to come up with a upper bound and a lower bound.

Treats nil term as matching every row to allow empty where clauses to be interpreted as matching everything (to conform to SQL standard).

References ibis::bitvector::bytes(), ibis::bitvector::clear(), ibis::bitvector::cnt(), ibis::qExists::colName(), ibis::bitvector::copy(), ibis::bitvector::flip(), ibis::compRange::inRange(), ibis::compRange::isConstant(), ibis::math::term::isTrue(), ibis::util::logMessage(), ibis::bitvector::set(), ibis::bitvector::size(), and ibis::bitvector::sloppyCount().

int ibis::query::doEvaluate ( const qExpr *  term, ibis::bitvector &  ht  ) const

protected

Evaluate the query expression.

Combines the operations on index and the sequential scan in one function.

It returns a non-negative value to indicate success and a negative value to indicate error. Note that a return value of zero indicates that no hits is found. However, a positive return value does not necessarily mean the number of hits is greater than zero, it simply means that it will take some work to figure out the actually number of hits.

References ibis::bitvector::bytes(), ibis::bitvector::clear(), ibis::bitvector::cnt(), ibis::qExists::colName(), ibis::bitvector::flip(), ibis::compRange::inRange(), ibis::qExpr::isConstant(), ibis::compRange::isConstant(), ibis::math::term::isTrue(), ibis::bitvector::set(), ibis::bitvector::size(), and ibis::bitvector::sloppyCount().

int ibis::query::doEvaluate ( const qExpr *  term, const ibis::bitvector &  mask, ibis::bitvector &  ht  ) const

protected

Evaluate the query expression with mask.

Combines the operations on index and the sequential scan in one function.

References ibis::bitvector::bytes(), ibis::bitvector::cnt(), ibis::qExists::colName(), ibis::bitvector::copy(), ibis::bitvector::flip(), ibis::compRange::inRange(), ibis::qExpr::isConstant(), ibis::compRange::isConstant(), ibis::math::term::isTrue(), ibis::bitvector::set(), ibis::bitvector::size(), and ibis::bitvector::sloppyCount().

int ibis::query::doScan ( const qExpr *  term, const ibis::bitvector &  mask, ibis::bitvector &  ht  ) const

protected

Masked sequential scan.

Reads the data partition to resolve the query conditions.

References ibis::bitvector::bytes(), ibis::bitvector::cnt(), ibis::qExists::colName(), ibis::bitvector::copy(), ibis::qExpr::getLeft(), ibis::qExpr::getRight(), ibis::qExpr::getType(), ibis::compRange::inRange(), ibis::qExpr::isConstant(), ibis::compRange::isConstant(), ibis::math::term::isTrue(), ibis::util::logMessage(), ibis::horometer::realTime(), ibis::bitvector::set(), ibis::bitvector::size(), ibis::bitvector::sloppyCount(), ibis::horometer::start(), and ibis::horometer::stop().

int ibis::query::doScan ( const qExpr *  term, ibis::bitvector &  ht  ) const

protected

Perform a sequential scan.

Reads the data partition to resolve the query expression.

References ibis::bitvector::cnt(), ibis::qExists::colName(), ibis::bitvector::flip(), ibis::compRange::inRange(), ibis::qExpr::isConstant(), ibis::compRange::isConstant(), ibis::math::term::isTrue(), ibis::bitvector::set(), ibis::bitvector::size(), and ibis::bitvector::sloppyCount().

int ibis::query::estimate

(

)

Function to perform estimation.

It computes a lower bound and an upper bound of hits. It uses the indexes only. If necessary it will build new indexes. The lower bound contains records that are definitely hits and the upper bound contains all hits but may also contain some records that are not hits. We also call the records in the upper bound candidates.

Returns 0 for success, a negative value for error.

References ibis::bitvector::adjustSize(), ibis::bitvector::bytes(), ibis::bitvector::cnt(), comps, conds, ibis::horometer::CPUTime(), ibis::qExpr::directEval(), dslock, ibis::selectClause::empty(), ibis::whereClause::empty(), ibis::util::emptyCache(), getBounds(), ibis::whereClause::getExpr(), ibis::part::getNullMask(), ibis::whereClause::getString(), hits, ibis::fileManager::instance(), logError(), ibis::part::nColumns(), ibis::part::nRows(), ibis::fileManager::pageCount(), QUICK_ESTIMATE, ibis::horometer::realTime(), removeFiles(), ibis::bitvector::set(), ibis::bitvector::size(), SPECIFIED, ibis::horometer::start(), state, ibis::horometer::stop(), sup, and ibis::part::timestamp().

Referenced by getNumHits(), and removeComplexConditions().

int ibis::query::evaluate

(

const bool

evalSelect = false

)

Computes the exact hits.

The same answer shall be computed whether there is any index or not. The argument evalSelect indicates whether the select clause should be evaluated at the same time. If its value is true, the columns specified in the select clause will be retrieved from disk and stored in the temporary location for this query. If not, the qualified values need to be retrieved by calling one of getRIDs, getQualifiedInts, getQualifiedFloats, getQualifiedDoubles and similar functions. These functions work with one column at a time. Note that if the data is dynamically varying, the values retrived later could be different from the values extracted while processing this function.

Returns 0 or a positive integer for success, a negative value for error. Note that when it returns 0, it indicates that the number of hits is 0. However, when it returns a positive value, the return value may not be the number of hits. This semantics is actually implemented in the support functions such as computeHits, doEvaluate, and doScan.

See also

getQualifiedInts

References ibis::bitvector::bytes(), ibis::bitvector::clusteringFactor(), ibis::bitvector::cnt(), comps, computeHits(), conds, countPages(), ibis::horometer::CPUTime(), ibis::bundle::create(), dslock, ibis::selectClause::empty(), ibis::whereClause::empty(), ibis::util::emptyCache(), ibis::util::envLock, FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, ibis::fileManager::iBeat(), ibis::fileManager::instance(), logError(), logMessage(), ibis::part::name(), ibis::part::nColumns(), ibis::part::nRows(), ibis::fileManager::pageCount(), ibis::fileManager::pageSize(), ibis::bitvector::randomSize(), ibis::horometer::realTime(), removeFiles(), ibis::array_t< T >::size(), ibis::bitvector::size(), ibis::bitvector::sloppyCount(), SPECIFIED, ibis::horometer::start(), state, ibis::horometer::stop(), sup, ibis::part::timestamp(), ibis::bundle::write(), writeHits(), and writeQuery().

Referenced by ibis::bord::computeHits(), getQualifiedBytes(), getQualifiedDoubles(), getQualifiedFloats(), getQualifiedInts(), getQualifiedLongs(), getQualifiedShorts(), getQualifiedStrings(), getQualifiedUBytes(), getQualifiedUInts(), getQualifiedULongs(), getQualifiedUShorts(), limit(), ibis::part::quickTest(), and removeComplexConditions().

void ibis::query::expandQuery

(

)

Expands where clause to preferred bounds.

This is to make sure the function estimate will give exact answer. It does nothing if there is no preferred bounds in the indices.

References BUNDLES_TRUNCATED, comps, conds, doExpand(), dslock, ibis::selectClause::empty(), ibis::whereClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), hits, HITS_TRUNCATED, QUICK_ESTIMATE, removeFiles(), ibis::whereClause::resetString(), SPECIFIED, state, sup, and writeQuery().

long ibis::query::getCandidateRows

(

std::vector< uint32_t > &

rids

)

const

Extract the positions of candidates.

This is meant to be used after calling the function estimate. It will return the position the hits if they are known already. If no estimate is known or the query is formed yet, it will return a negative number to indicate error. Upon a successful completion of this function, the return value should be the rids.size().

References ibis::bitvector::cnt(), hits, ibis::bitvector::indexSet::nIndices(), and sup.

long ibis::query::getExpandedHits

(

ibis::bitvector &

res

)

const

Get a bitvector containing all rows satisfying the query condition.

The resulting bitvector inculdes both active rows and inactive rows.

References ibis::bitvector::clear().

Referenced by ibis::part::stringToBitvector().

long ibis::query::getHitRows

(

std::vector< uint32_t > &

rids

)

const

Extract the positions of the bits that are 1s in the solution.

This is only valid after the query has been evaluated. If it has not been evaluated, it will return a negative number to indicate error. Upon a successful completion of this function, the return value should be the rids.size().

References ibis::bitvector::cnt(), hits, ibis::bitvector::indexSet::nIndices(), and sup.

const ibis::bitvector* ibis::query::getHitVector ( ) const

inline

Return the pointer to the internal hit vector.

The user should NOT attempt to free the returned pointer. If this function is called before calling ibis::query::evaluate, it might return a nil pointer.

References hits.

Referenced by ibis::bundle1::bundle1(), ibis::bundles::bundles(), and ibis::part::quickTest().

long int ibis::query::getNumHits

(

)

const

Compute the number of records in the exact solution.

This function will return the number of hits based on the internally stored information or other inexpensive options. It will not perform a full evaluation to compute the numbers of hits. It is intended to be called after calling ibis::query::evaluate. The return value will be -1 if it is not able to determine the number of hits.

References ibis::bitvector::cnt(), conds, ibis::part::countHits(), ibis::whereClause::empty(), estimate(), ibis::whereClause::getExpr(), hits, ibis::part::nRows(), QUICK_ESTIMATE, state, and sup.

Referenced by ibis::bundle0::bundle0(), ibis::bundle1::bundle1(), ibis::bundles::bundles(), ibis::bord::computeHits(), and ibis::part::quickTest().

ibis::array_t< signed char > * ibis::query::getQualifiedBytes

(

const char *

colname

)

The functions getQualifiedTTT return the values of selected columns in the records that satisfies the specified conditions.

An implicit casting will be performed if possible.

The caller must call the operator delete to free the pointers returned.

Any column in the data partition may be used with getQualifiedTTT, not just those given in the select clause. The content returned is read from disk when these functions are called.

Retrieve the values of column_name as 8-bit integers.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectBytes(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< double > * ibis::query::getQualifiedDoubles

(

const char *

colname

)

Retrieve double precision floating-point values from records satisfying the query conditions.

An implicit casting will be performed if the specified column is not of type double.

Note that casting from 64-bit integers to double may cause loss of precision; casting of 32-bit floating-point values to 64-bit version may lead to spurious precision.

References dslock, evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectDoubles(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< float > * ibis::query::getQualifiedFloats

(

const char *

colname

)

Retrieve floating-point values from records satisfying the query conditions.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit floating-point values.

References dslock, evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectFloats(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< int32_t > * ibis::query::getQualifiedInts

(

const char *

colname

)

Retrieve integer values from records satisfying the query conditions.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectInts(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< int64_t > * ibis::query::getQualifiedLongs

(

const char *

colname

)

Retrieve values of column_name as 64-bit integers.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectLongs(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< int16_t > * ibis::query::getQualifiedShorts

(

const char *

colname

)

Retrieve the values of column_name as 16-bit integers.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectShorts(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

std::vector< std::string > * ibis::query::getQualifiedStrings

(

const char *

colname

)

Retrieve string values from records satisfying the query conditions.

The argument colname must be the name of a string-valued column, otherwise a null pointer will be returned.

Note

FastBit does not track the memory usage of neither std::vector nor std::string.

References dslock, evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectStrings(), state, and ibis::part::timestamp().

ibis::array_t< unsigned char > * ibis::query::getQualifiedUBytes

(

const char *

colname

)

Retrieve the values of column_name as 8-bit unsigned integers.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit unsigned integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectUBytes(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< uint32_t > * ibis::query::getQualifiedUInts

(

const char *

colname

)

Retrieve unsigned integer values from records satisfying the query conditions.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit unsigned integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectUInts(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< uint64_t > * ibis::query::getQualifiedULongs

(

const char *

colname

)

Retrieve values of column_name as 64-bit unsigned integers.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit unsigned integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectULongs(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::array_t< uint16_t > * ibis::query::getQualifiedUShorts

(

const char *

colname

)

Retrieve the values of column_name as 16-bit unsigned integers.

An implicit casting will be performed if possible.

A null pointer will be returned if the underlying values can not be safely cast into 32-bit unsigned integers.

References evaluate(), FULL_EVALUATE, hits, logMessage(), ibis::part::selectUShorts(), ibis::array_t< T >::size(), state, and ibis::part::timestamp().

ibis::RIDSet * ibis::query::getRIDs

(

)

const

Return the list of row IDs of the hits.

FastBit has a built-in type called ibis::rid_t.

User may use it to provide a global row identifier for each row. We call such row identifiers (RIDs) the extenal RIDs. In many cases, there is no external RIDs provided by the user, then there is still a set of implicit RIDs numbered from 0 to nRows()-1. This function will retrieve the extenal RIDs if they are present, otherwise, it will return the implicit RIDs.

Note

the returned pointer will be null if the query does not have an exact answer yet, has no hit, or is not associated with any data partition.

References ibis::bitvector::cnt(), ibis::array_t< T >::deepCopy(), ibis::part::explicitRIDs(), FULL_EVALUATE, hits, logMessage(), ibis::bitvector::indexSet::nIndices(), ibis::part::nRows(), ibis::array_t< T >::push_back(), readRIDs(), ibis::array_t< T >::reserve(), ibis::array_t< T >::size(), state, ibis::part::timestamp(), and writeRIDs().

Referenced by ibis::part::quickTest().

ibis::RIDSet * ibis::query::getRIDs

(

const ibis::bitvector &

mask

)

const

Return a list of row IDs that match the mask.

The data type for row identifiers is ibis::rid_t, which can be treated as unsigned 64-bit integer.

These identifiers (RIDs) can be either provided by the user (external RIDs) or internally generated from row positions (implicit RIDs). If the user has not provided external RIDs, then this function simply decodes the positions of the bits that are marked 1 and places the positions in the output array.

The return value can be null if this query object is not associated with a data partition or the mask contains no bit marked 1.

References ibis::bitvector::cnt(), logMessage(), ibis::part::name(), ibis::part::nRows(), and ibis::array_t< T >::size().

bool ibis::query::isValidToken ( const char *  tok )

static

Is the given string a valid query token.

Return true if it has the expected token format, otherwise false.

References ibis::util::charIndex, ibis::util::logMessage(), and tokenLength().

long ibis::query::limit	(	const char *	names,
		uint32_t	keep,
		bool	updateHits = true
	)

Truncate the results to provide the top-K rows.

It returns the number of results kept, which is the smaller of the current number of rows and the input argument keep. A negative value is returned in case of error, e.g., query has not been fully specified. If the 4th argument is true, the internal hit vector is updated to match the truncated solution. Otherwise, the internal hit vector is left unchanged. Since the functions getNumHits and getQualifiedTTT uses this internal hit vector, it is generally a good idea to update the hit vector. On the other hand, one may wish to avoid this update if the hit vector is to be kept for some purpose.

References BUNDLES_TRUNCATED, ibis::horometer::CPUTime(), ibis::bundle::create(), evaluate(), ibis::part::evaluateRIDSet(), FULL_EVALUATE, ibis::bundle::getRIDs(), hits, HITS_TRUNCATED, logMessage(), QUICK_ESTIMATE, ibis::horometer::realTime(), SET_COMPONENTS, SET_PREDICATE, SET_RIDS, ibis::bundle::size(), SPECIFIED, ibis::horometer::start(), state, ibis::horometer::stop(), ibis::bundle::truncate(), UNINITIALIZED, and ibis::bundle::write().

void ibis::query::logMessage	(	const char *	event,
		const char *	fmt,
			...
	)		const

Used to print information about the progress or state of query processing.

It prefixes each message with a query token.

References ibis::util::getLocalTime(), and ibis::util::getLogFile().

Referenced by ibis::bundle::create(), evaluate(), getQualifiedBytes(), getQualifiedDoubles(), getQualifiedFloats(), getQualifiedInts(), getQualifiedLongs(), getQualifiedShorts(), getQualifiedStrings(), getQualifiedUBytes(), getQualifiedUInts(), getQualifiedULongs(), getQualifiedUShorts(), getRIDs(), getRIDsInBundle(), getState(), limit(), orderby(), removeComplexConditions(), setPartition(), setSelectClause(), and setWhereClause().

int ibis::query::orderby

(

const char *

names

)

const

Re-order the results according to the new "ORDER BY" specification.

It returns 0 if it completes successfully. It returns a negative number to indicate error. If direction >= 0, sort the values in ascending order, otherwise, sort them in descending order.

Note

The results stored in ibis::bundle and ibis::query::result are already ordered according to the columns specified in the select clause. One only needs to call this function to re-order the results differently.

References BUNDLES_TRUNCATED, ibis::horometer::CPUTime(), ibis::bundle::create(), FULL_EVALUATE, HITS_TRUNCATED, logMessage(), ibis::horometer::realTime(), ibis::bundle::reorder(), ibis::horometer::start(), state, ibis::horometer::stop(), and ibis::bundle::write().

void ibis::query::orderPairs ( const char *  pfile ) const

protected

Sort the content of the file as ibis::rid_t.

It reads the content of the file one block at a time during the initial sorting of the blocks. It then merges the sorted blocks to produce a overall sorted file. Note that ibis::rid_t is simply a pair of integers. Sinc the pairs are recorded as pairs of integers too, this should work.

References ibis::fileManager::bytesFree(), ibis::util::getFileSize(), ibis::fileManager::instance(), ibis::util::logMessage(), PREFERRED_BLOCK_SIZE, and UnixOpen.

void ibis::query::printRIDs ( const RIDSet &  ridset ) const

protected

Export the Row IDs of the hits to log file.

ibis::query::setMyDir

This function prints a list of RIDs to the log file.

References ibis::array_t< T >::size().

void ibis::query::printSelected

(

std::ostream &

out

)

const

Print the values of the selected columns to the specified output stream.

The printed values are grouped by the columns without functions. For each group, the functions are evaluated on the columns named in the function. This is equivalent to have implicit "GROUP BY" and "ORDER BY" keywords on all columns appears without a function in the select clause.

References ibis::bundle::create(), ibis::bundle::print(), and ibis::bundle::write().

void ibis::query::printSelectedWithRID

(

std::ostream &

out

)

const

Print the values of the columns in the select clause without functions.

One the groups of unique values are printed. For each group, the row ID (RID) of the rows are also printed.

References ibis::bundle::create(), ibis::bundle::printAll(), and ibis::bundle::write().

int64_t ibis::query::processJoin ( )

protected

Process the join operation and return the number of pairs.

This function only counts the number of hits; it does produce the actual tuples for the results of join. Additionally, it performs only self-join, i.e., join a partition with itself. This is only meant to test some algorithms for evaluating joins.

References ibis::index::BINNING, ibis::bitvector::bytes(), ibis::fileManager::bytesFree(), ibis::fileManager::bytesInUse(), ibis::bitvector64::clear(), ibis::bitvector::clusteringFactor(), ibis::bitvector64::cnt(), ibis::bitvector::cnt(), ibis::horometer::CPUTime(), ibis::math::barrel::equivalent(), ibis::qExpr::findRange(), ibis::column::getNullMask(), ibis::util::logMessage(), ibis::bitvector64::markovSize(), ibis::column::name(), ibis::util::outerProduct(), ibis::horometer::realTime(), ibis::math::barrel::recordVariable(), ibis::index::RELIC, ibis::bitvector64::set(), ibis::bitvector::size(), ibis::bitvector64::size(), ibis::horometer::start(), ibis::horometer::stop(), ibis::bitvector64::swap(), and ibis::index::type().

void ibis::query::readQuery ( const ibis::partList &  tl )

protected

Read the status information from disk.

A function to read the query file in a directory – used by the constructor that takes a directory name as the argument the file contains:

• user id
• dataset name
• list of components
• query state
• time stamp on the dataset
• query condition or <NULL>
• list of OIDs

References ibis::util::clear(), and ibis::util::strnewdup().

Referenced by query().

ibis::RIDSet * ibis::query::readRIDs

(

)

const

Read RIDs from the file named "-rids".

Return a pointer to ibis::RIDSet.

References ibis::fileManager::getFile(), ibis::fileManager::instance(), and ibis::array_t< T >::size().

Referenced by ibis::bundle1::bundle1(), ibis::bundles::bundles(), and getRIDs().

int64_t ibis::query::recordDeltaPairs	(	const array_t< uint32_t > &	val1,
		const array_t< int32_t > &	val2,
		const array_t< uint32_t > &	ind1,
		const array_t< uint32_t > &	ind2,
		const uint32_t &	delta,
		const char *	filename
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), ibis::array_t< T >::size(), and UnixOpen.

int64_t ibis::query::recordDeltaPairs	(	const array_t< int32_t > &	val1,
		const array_t< uint32_t > &	val2,
		const array_t< uint32_t > &	ind1,
		const array_t< uint32_t > &	ind2,
		const int32_t &	delta,
		const char *	filename
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), ibis::array_t< T >::size(), and UnixOpen.

int64_t ibis::query::recordEqualPairs	(	const array_t< uint32_t > &	val1,
		const array_t< int32_t > &	val2,
		const array_t< uint32_t > &	ind1,
		const array_t< uint32_t > &	ind2,
		const char *	filename
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), ibis::array_t< T >::size(), and UnixOpen.

int64_t ibis::query::recordEqualPairs	(	const array_t< int32_t > &	val1,
		const array_t< uint32_t > &	val2,
		const array_t< uint32_t > &	ind1,
		const array_t< uint32_t > &	ind2,
		const char *	filename
	)		const

This is an explicit specialization of a protected member of ibis::query class.

Note

The C++ language rules require explicit specialization of template member function be declared in the namespace containing the function, not inside the class! This apparently causes them to be listed as public functions in Doxygen document.

References ibis::array_t< T >::find(), ibis::array_t< T >::size(), and UnixOpen.

std::string ibis::query::removeComplexConditions

(

)

Separate out the sub-expressions that are not simple.

This is intended to allow the overall where clause to be evaluated in separated steps, where the simple conditions are left for this software to handle and the more complex ones are to be handled by another software. The set of conditions remain with this query object and the conditions returned by this function are assumed to be connected with the operator AND. If the top-most operator in the WHERE clause is not an AND operator, the whole clause will be returned if it contains any conditions that is not simple, otherwise, an empty string will be returned.

References BUNDLES_TRUNCATED, ibis::whereClause::clear(), conds, ibis::whereClause::empty(), estimate(), evaluate(), FULL_EVALUATE, getBounds(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, logMessage(), ibis::part::nRows(), ibis::qExpr::print(), QUICK_ESTIMATE, ibis::qExpr::separateSimple(), ibis::bitvector::set(), setWhereClause(), state, and sup.

long ibis::query::sequentialScan

(

ibis::bitvector &

res

)

const

Perform a simple sequential scan.

Return a (new) bitvector that contains the result of directly scan the raw data to determine what records satisfy the user specified conditions. It is mostly used for testing purposes. It can be called any time after the where clause is set, and does not change the state of the current query.

References ibis::bitvector::clear(), ibis::bitvector::cnt(), ibis::bitvector::copy(), ibis::horometer::CPUTime(), ibis::util::emptyCache(), ibis::bitvector::indexSet::indices(), ibis::bitvector::indexSet::isRange(), ibis::bitvector::indexSet::nIndices(), ibis::bitvector::print(), ibis::horometer::realTime(), ibis::horometer::start(), and ibis::horometer::stop().

Referenced by ibis::part::quickTest().

int ibis::query::setPartition

(

const ibis::part *

tbl

)

Resets the data partition associated with the query.

Integer error code: 0: successful completion of the requested operation.

-1: nil pointer to data partition or empty partition. -2: invalid string for select clause. -3: select clause contains invalid column name. -4: invalid string for where clause. -5: where clause can not be parsed correctly. -6: where clause contains invalid column names or unsupported functions. -7: empty rid list for set rid operation. -8: neither rids nor range conditions are set. -9: encountered some exceptional conditions during query evaluations. -10: no private directory to store bundles. -11: Query not fully evaluated.

References BUNDLES_TRUNCATED, comps, conds, dslock, ibis::selectClause::empty(), ibis::whereClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, logMessage(), ibis::part::name(), ibis::part::nColumns(), ibis::part::nRows(), QUICK_ESTIMATE, removeFiles(), SET_COMPONENTS, SET_PREDICATE, SPECIFIED, state, sup, ibis::selectClause::verify(), ibis::whereClause::verify(), and writeQuery().

Referenced by setTable().

int ibis::query::setRIDs

(

const RIDSet &

set

)

Specify a list of Row IDs for the query object.

Select the records with an RID in the list of RIDs.

Note

The incoming RIDs are copied.

The RIDs and the where clause can be used together. When they are both specified, they are used in conjuction. In other word, the hits of the query will contain only the records that satisfy the where clause and have an RID in the list of RIDs.

References BUNDLES_TRUNCATED, comps, ibis::array_t< T >::deepCopy(), ibis::selectClause::empty(), FULL_EVALUATE, hits, HITS_TRUNCATED, QUICK_ESTIMATE, removeFiles(), SET_RIDS, ibis::array_t< T >::size(), SPECIFIED, state, sup, and writeQuery().

Referenced by ibis::part::quickTest().

int ibis::query::setSelectClause ( const char *  str )

virtual

Specifies the select clause for the query.

The select clause is a string of column names separated by spaces, commas (,) or semicolons(;). Repeated calls to this function simply overwrite the previous definition of the select clause. If no select clause is specified, the where clause alone determines whether record is a hit or not. The select clause will be reordered to make the plain column names without functions appear before with functions.

References BUNDLES_TRUNCATED, ibis::part::columnNames(), comps, conds, FULL_EVALUATE, ibis::whereClause::getExpr(), hits, HITS_TRUNCATED, logMessage(), ibis::part::name(), QUICK_ESTIMATE, removeFiles(), SET_COMPONENTS, SPECIFIED, state, sup, ibis::selectClause::swap(), ibis::selectClause::verify(), and writeQuery().

Referenced by ibis::part::quickTest().

int ibis::query::setWhereClause

(

const char *

str

)

Specify the where clause in the string form.

The where clause is a string representing a list of range conditions. By SQL convention, an empty where clause matches all rows. This function may be called multiple times and each invocation will overwrite the previous where clause.

References BUNDLES_TRUNCATED, comps, conds, ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, logMessage(), ibis::part::name(), ibis::qExpr::printFull(), QUICK_ESTIMATE, removeFiles(), SET_PREDICATE, SPECIFIED, state, sup, ibis::whereClause::swap(), ibis::whereClause::verify(), and writeQuery().

Referenced by ibis::bord::computeHits(), ibis::part::quickTest(), removeComplexConditions(), and ibis::part::stringToBitvector().

int ibis::query::setWhereClause	(	const std::vector< const char * > &	names,
		const std::vector< double > &	lbounds,
		const std::vector< double > &	rbounds
	)

Specify the where clause as a set of conjunctive ranges.

This function accepts a set of range conditions expressed by the three vectors. The arrays are expected to be of the same size, and each triplet <names[i], lbounds[i], rbounds[i]> are interpreted as

names[i] between lbounds[i] and rbounds[i]

The range conditions are joined together with the AND operator. If vectors lbounds and rbounds are not the same size, then the missing one is consider to represent an open boundary. For example, if lbounds[4] exists but not rbounds[4], they the range condition is interpreted as

lbounds[4] <= names[4]

References BUNDLES_TRUNCATED, comps, conds, ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), hits, HITS_TRUNCATED, ibis::part::name(), QUICK_ESTIMATE, removeFiles(), SET_COMPONENTS, SET_PREDICATE, ibis::whereClause::setExpr(), ibis::qExpr::setLeft(), ibis::qExpr::setRight(), SPECIFIED, state, sup, UNINITIALIZED, ibis::whereClause::verify(), and writeQuery().

int ibis::query::setWhereClause

(

const ibis::qExpr *

qx

)

Specify the where clause through a qExpr object.

This function accepts a user constructed query expression object. It can be used to bypass the parsing of where clause string.

Note

The query object will hold a copy of the incoming object.

References BUNDLES_TRUNCATED, comps, conds, ibis::selectClause::empty(), FULL_EVALUATE, ibis::whereClause::getExpr(), ibis::whereClause::getString(), hits, HITS_TRUNCATED, ibis::part::name(), ibis::qExpr::nItems(), QUICK_ESTIMATE, removeFiles(), ibis::whereClause::resetString(), SET_COMPONENTS, SET_PREDICATE, ibis::whereClause::setExpr(), SPECIFIED, state, sup, ibis::whereClause::swap(), UNINITIALIZED, ibis::whereClause::verify(), and writeQuery().

int64_t ibis::query::sortEquiJoin ( const ibis::deprecatedJoin &  cmp, const ibis::bitvector &  mask  ) const

protected

Performing an equi-join by sorting the selected values first.

This version reads the values marked to be 1 in the bitvector mask and performs the actual operation of counting the number of pairs with equal values in memory.

References ibis::CATEGORY, ibis::horometer::CPUTime(), ibis::DOUBLE, ibis::FLOAT, ibis::INT, ibis::util::logMessage(), ibis::horometer::realTime(), ibis::column::selectValues(), ibis::horometer::start(), ibis::horometer::stop(), ibis::column::type(), and ibis::UINT.

int64_t ibis::query::sortEquiJoin ( const ibis::deprecatedJoin &  cmp, const ibis::bitvector &  mask, const char *  pairfile  ) const

protected

Perform equi-join by sorting the selected values.

This version reads the values marked to be 1 in the bitvector mask. It writes the the pairs satisfying the join condition to a file name pairfile.

References ibis::CATEGORY, ibis::horometer::CPUTime(), ibis::DOUBLE, ibis::FLOAT, ibis::INT, ibis::util::logMessage(), ibis::horometer::realTime(), ibis::column::selectValues(), ibis::array_t< T >::size(), ibis::array_t< T >::stableSort(), ibis::horometer::start(), ibis::horometer::stop(), ibis::column::type(), and ibis::UINT.

int64_t ibis::query::sortRangeJoin ( const ibis::deprecatedJoin &  cmp, const ibis::bitvector &  mask  ) const

protected

Performing a range join by sorting the selected values.

The sorting is performed through std::sort algorithm.

References ibis::CATEGORY, ibis::horometer::CPUTime(), ibis::DOUBLE, ibis::math::term::eval(), ibis::FLOAT, ibis::INT, ibis::util::logMessage(), ibis::horometer::realTime(), ibis::column::selectValues(), ibis::horometer::start(), ibis::horometer::stop(), ibis::column::type(), and ibis::UINT.

void ibis::query::storeErrorMesg ( const char *  msg ) const

protected

Store the message into member variable lastError for later use.

This function will truncate long messages because lastError is declared with a fixed size of MAX_LINE+PATH_MAX. If the incoming argument is a nil pointer or an empty sttring, lastError will be set to be an empty string as if clearErrorMessage is called.

void ibis::query::writeQuery ( )

protectedvirtual

Write the basic information about the query to disk.

Write the content of the current query into a file.

Referenced by addConditions(), contractQuery(), evaluate(), expandQuery(), setPartition(), setRIDs(), setSelectClause(), and setWhereClause().

---

The documentation for this class was generated from the following files:

• query.h
• query.cpp