Everything necessary to process a query is encapsulated in an object of this type. More...

#include <query_block_max.h>

Inheritance diagram for JASS::query_block_max:

Collaboration diagram for JASS::query_block_max:

Public Member Functions
	query_block_max (compress_integer &codex)
	Constructor.

virtual	~query_block_max ()
	Destructor.

virtual void	init (const std::vector< std::string > &primary_keys, DOCID_TYPE documents=1024, DOCID_TYPE top_k=10, size_t width=7)
	Initialise the object. MUST be called before first use. More...

virtual docid_rsv_pair *	get_first (void)
	Retrun the top result. More...

virtual docid_rsv_pair *	get_next (void)
	After calling get_first(), return the next result. More...

virtual void	rewind (ACCUMULATOR_TYPE smallest_possible_rsv=0, ACCUMULATOR_TYPE top_k_lower_bound=1, ACCUMULATOR_TYPE largest_possible_rsv=0)
	Clear this object after use and ready for re-use.

virtual void	sort (void)
	sort this resuls list before iteration over it. More...

forceinline void	add_rsv (DOCID_TYPE document_id, ACCUMULATOR_TYPE score)
	Add weight to the rsv for document document_id. More...

virtual void	decode_with_writer (size_t integers, const void *compressed, size_t compressed_size)
	Given the integer decoder, the number of integes to decode, and the compressed sequence, decompress (but do not process). More...

Public Member Functions inherited from JASS::query
	query (compress_integer &codex)
	Constructor.

virtual	~query ()
	Destructor.

template<typename STRING_TYPE >
void	parse (const STRING_TYPE &query, parser_query::parser_type which_parser=parser_query::parser_type::query)
	Take the given query and parse it. More...

query_term_list &	terms (void)
	Return a reference to the parsed query. More...

forceinline void	set_impact (ACCUMULATOR_TYPE score)
	Set the impact score to use in a push_back(). More...

forceinline void	decode_and_process (ACCUMULATOR_TYPE impact, size_t integers, const void *compressed, size_t compressed_size)
	Given the integer decoder, the number of integes to decode, and the compressed sequence, decompress (but do not process). More...

virtual void	decode_with_writer (query::printer &writer, size_t integers, const void *compressed, size_t compressed_size)
	Given the integer decoder, the number of integes to decode, and the compressed sequence, decompress (but do not process). More...

Static Public Member Functions
static void	unittest (void)
	Unit test this class.

Private Types
typedef pointer_box< ACCUMULATOR_TYPE >	accumulator_pointer

Private Attributes
accumulator_block_max< ACCUMULATOR_TYPE, MAX_DOCUMENTS >	accumulators
	The accumulators, one per document in the collection.

bool	sorted
	Has the top-k been generates (false after rewind() true after sort())

docid_rsv_pair	next_result
	A single result, used but get_first() and get_next()

DOCID_TYPE	next_result_location
	Used by get_first() and get_next() to determine which result is next.

accumulator_pointer	accumulator_pointers [MAX_TOP_K]
	Array of pointers to the top k accumulators.

heap< accumulator_pointer >	top_results
	Heap containing the top-k results.

DOCID_TYPE	needed_for_top_k
	The number of results we still need in order to fill the top-k.

Additional Inherited Members
Public Types inherited from JASS::query
typedef uint8_t	ACCUMULATOR_TYPE
	the type of an accumulator (probably a uint16_t)

typedef uint32_t	DOCID_TYPE
	the type of a document id (from a compressor)

Public Attributes inherited from JASS::query
DOCID_TYPE	top_k
	The number of results to track.

Static Public Attributes inherited from JASS::query
static constexpr size_t	MAX_DOCUMENTS = 200'000'000
	the maximum number of documents an index can hold

static constexpr size_t	MAX_TOP_K = 1'000
	the maximum top-k value

static constexpr size_t	MAX_RSV = (std::numeric_limits<ACCUMULATOR_TYPE>::max)()

Protected Attributes inherited from JASS::query
ACCUMULATOR_TYPE	impact
	The impact score to be added on a call to add_rsv()

allocator_pool	memory
	All memory allocation happens in this "arena".

std::vector< DOCID_TYPE >	decompress_buffer
	The delta-encoded decopressed integer sequence.

DOCID_TYPE	documents
	The numnber of documents this index contains.

parser_query	parser
	Parser responsible for converting text into a parsed query.

query_term_list *	parsed_query
	The parsed query.

const std::vector< std::string > *	primary_keys
	A vector of strings, each the primary key for the document with an id equal to the vector index.

compress_integer &	codex
	The decompressor to use.

Detailed Description

Everything necessary to process a query is encapsulated in an object of this type.

Member Function Documentation

◆ add_rsv()

forceinline void JASS::query_block_max::add_rsv	(	DOCID_TYPE	document_id,
		ACCUMULATOR_TYPE	score
	)

inline

Add weight to the rsv for document document_id.

Parameters

document_id	[in] which document to increment
score	[in] the amount of weight to add

◆ decode_with_writer()

virtual void JASS::query_block_max::decode_with_writer	(	size_t	integers,
		const void *	compressed,
		size_t	compressed_size
	)

inlinevirtual

Given the integer decoder, the number of integes to decode, and the compressed sequence, decompress (but do not process).

Parameters

integers	[in] The number of integers that are compressed.
compressed	[in] The compressed sequence.
compressed_size	[in] The length of the compressed sequence.

Implements JASS::query.

◆ get_first()

virtual docid_rsv_pair* JASS::query_block_max::get_first ( void )

inlinevirtual

Retrun the top result.

Returns: The first (i.e. top) result in the results list.

Implements JASS::query.

◆ get_next()

virtual docid_rsv_pair* JASS::query_block_max::get_next ( void )

inlinevirtual

After calling get_first(), return the next result.

Returns: The next result in the results list, or NULL if at end of list

Implements JASS::query.

◆ init()

virtual void JASS::query_block_max::init	(	const std::vector< std::string > &	primary_keys,
		DOCID_TYPE	documents = `1024`,
		DOCID_TYPE	top_k = `10`,
		size_t	width = `7`
	)

inlinevirtual

Initialise the object. MUST be called before first use.

Parameters

primary_keys	[in] Vector of the document primary keys used to convert from internal document ids to external primary keys.
documents	[in] The number of documents in the collection.
top_k	[in] The top-k documents to return from the query once executed.
width	[in] The width of the 2-d accumulators (if they are being used).

Reimplemented from JASS::query.

◆ sort()

virtual void JASS::query_block_max::sort ( void )

inlinevirtual

sort this resuls list before iteration over it.

In block_max processing, we scan the block_max scores looking for instances where the score is greater than the bottom of the heap. If we find one then its necessary to scan that block looking for instances where an accumulator is greater than the bottom of the heap. The heap is then built from those accumulators bu pushing pointers to the accumulators onto the heap. Then we sort the heap.

Implements JASS::query.

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

source/query_block_max.h

Public Member Functions

Static Public Member Functions

Private Types

Private Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation

◆ add_rsv()

◆ decode_with_writer()

◆ get_first()

◆ get_next()

◆ init()

◆ sort()