#include <matcher.h>

Collaboration diagram for Matcher:

Public Member Functions
	Matcher (const Xapian::Database &db_, const Xapian::Query &query, Xapian::termcount query_length, const Xapian::RSet *rset, Xapian::Weight::Internal &stats, const Xapian::Weight &wtscheme, bool have_mdecider, Xapian::valueno collapse_key, Xapian::doccount collapse_max, int percent_threshold, double weight_threshold, Xapian::Enquire::docid_order order, Xapian::valueno sort_key, Xapian::Enquire::Internal::sort_setting sort_by, bool sort_val_reverse, double time_limit, const std::vector< opt_ptr_spy > &matchspies)
	Constructor. More...

Xapian::MSet	get_mset (Xapian::doccount first, Xapian::doccount maxitems, Xapian::doccount check_at_least, Xapian::Weight::Internal &stats, const Xapian::Weight &wtscheme, const Xapian::MatchDecider mdecider, const Xapian::KeyMaker sorter, Xapian::valueno collapse_key, Xapian::doccount collapse_max, int percent_threshold, double weight_threshold, Xapian::Enquire::docid_order order, Xapian::valueno sort_key, Xapian::Enquire::Internal::sort_setting sort_by, bool sort_val_reverse, double time_limit, const std::vector< opt_ptr_spy > &matchspies)
	Run the match and produce an MSet object. More...

Private Types
typedef Xapian::Internal::opt_intrusive_ptr< Xapian::MatchSpy >	opt_ptr_spy

Private Member Functions
	Matcher (const Matcher &)=delete

Matcher &	operator= (const Matcher &)=delete

Xapian::MSet	get_local_mset (Xapian::doccount first, Xapian::doccount maxitems, Xapian::doccount check_at_least, const Xapian::Weight &wtscheme, const Xapian::MatchDecider mdecider, const Xapian::KeyMaker sorter, Xapian::valueno collapse_key, Xapian::doccount collapse_max, int percent_threshold, double percent_threshold_factor, double weight_threshold, Xapian::Enquire::docid_order order, Xapian::valueno sort_key, Xapian::Enquire::Internal::sort_setting sort_by, bool sort_val_reverse, double time_limit, const std::vector< opt_ptr_spy > &matchspies)

template<typename Action >
void	for_all_remotes (Action action)
	Perform action on remotes as they become ready using poll() or select(). More...

Private Attributes
Xapian::Database	db

std::vector< std::unique_ptr< LocalSubMatch > >	locals
	LocalSubMatch objects for local databases. More...

std::vector< std::unique_ptr< RemoteSubMatch > >	remotes
	RemoteSubMatch objects for remote databases. More...

std::size_t	first_nonselectable
	Partition point in remotes. More...

Detailed Description

Definition at line 46 of file matcher.h.

Member Typedef Documentation

◆ opt_ptr_spy

typedef Xapian::Internal::opt_intrusive_ptr<Xapian::MatchSpy> Matcher::opt_ptr_spy

private

Definition at line 47 of file matcher.h.

Constructor & Destructor Documentation

◆ Matcher() [1/2]

Matcher::Matcher ( const Matcher & )

privatedelete

◆ Matcher() [2/2]

Matcher::Matcher	(	const Xapian::Database &	db_,
		const Xapian::Query &	query,
		Xapian::termcount	query_length,
		const Xapian::RSet *	rset,
		Xapian::Weight::Internal &	stats,
		const Xapian::Weight &	wtscheme,
		bool	have_mdecider,
		Xapian::valueno	collapse_key,
		Xapian::doccount	collapse_max,
		int	percent_threshold,
		double	weight_threshold,
		Xapian::Enquire::docid_order	order,
		Xapian::valueno	sort_key,
		Xapian::Enquire::Internal::sort_setting	sort_by,
		bool	sort_val_reverse,
		double	time_limit,
		const std::vector< opt_ptr_spy > &	matchspies
	)

Constructor.

Parameters

db_	Database to search
query	Query object
query_length	Query length
rset	Relevance set (NULL for none)
stats	Object to collate stats into
wtscheme	Weight object to use as factory
have_mdecider	MatchDecider specified?
collapse_key	value slot to collapse on (Xapian::BAD_VALUENO which means no collapsing)
collapse_max	Maximum number of documents with the same key to allow
percent_threshold	Lower bound on percentage score
weight_threshold	Lower bound on weight
order	Xapian::docid sort order
sort_key	Value slot to sort on
sort_by	What to sort results on
sort_val_reverse	Reverse direction keys sort in?
time_limit	time in seconds after which to disable check_at_least (0.0 means don't).
matchspies	MatchSpy objects to use

Member Function Documentation

◆ for_all_remotes()

template<typename Action >

void Matcher::for_all_remotes ( Action action )

inlineprivate

Perform action on remotes as they become ready using poll() or select().

Definition at line 80 of file matcher.cc.

References socket_errno().

Referenced by get_mset().

◆ get_local_mset()

Xapian::MSet Matcher::get_local_mset	(	Xapian::doccount	first,
		Xapian::doccount	maxitems,
		Xapian::doccount	check_at_least,
		const Xapian::Weight &	wtscheme,
		const Xapian::MatchDecider *	mdecider,
		const Xapian::KeyMaker *	sorter,
		Xapian::valueno	collapse_key,
		Xapian::doccount	collapse_max,
		int	percent_threshold,
		double	percent_threshold_factor,
		double	weight_threshold,
		Xapian::Enquire::docid_order	order,
		Xapian::valueno	sort_key,
		Xapian::Enquire::Internal::sort_setting	sort_by,
		bool	sort_val_reverse,
		double	time_limit,
		const std::vector< opt_ptr_spy > &	matchspies
	)

private

EstimateOp tree to calculate an Estimates object for each local shard.

This allows the estimate to be calculated at the end of the match so that it can incorporate information about things such as how many documents were accepted and rejected by positional checks.

Definition at line 341 of file matcher.cc.

References Xapian::Internal::intrusive_base::_refs, Assert, db, EstimateOp::DECIDER, PostListTree::delete_postlists(), Xapian::Enquire::DESCENDING, DOCID, dummy, ProtoMSet::early_reject(), Estimates::est, Xapian::Internal::PostListAndEstimate::est, ProtoMSet::finalise(), PostListTree::get_docid(), ProtoMSet::get_min_weight(), get_msetcmp_function(), ValueStreamDocument::get_value(), PostListTree::get_weight(), locals, Estimates::max, Estimates::min, PostListTree::next(), Xapian::Internal::PostListAndEstimate::pl, ProtoMSet::process(), Xapian::Vec< T, COW, UNIQUEPTR, typename >::push_back(), PostListTree::recalc_maxweight(), REL, REL_VAL, ValueStreamDocument::set_document(), ProtoMSet::set_new_min_weight(), PostListTree::set_postlists(), Result::set_sort_key(), Result::set_weight(), Xapian::Vec< T, COW, UNIQUEPTR, typename >::size(), VAL, and VAL_REL.

Referenced by get_mset().

◆ get_mset()

Xapian::MSet Matcher::get_mset	(	Xapian::doccount	first,
		Xapian::doccount	maxitems,
		Xapian::doccount	check_at_least,
		Xapian::Weight::Internal &	stats,
		const Xapian::Weight &	wtscheme,
		const Xapian::MatchDecider *	mdecider,
		const Xapian::KeyMaker *	sorter,
		Xapian::valueno	collapse_key,
		Xapian::doccount	collapse_max,
		int	percent_threshold,
		double	weight_threshold,
		Xapian::Enquire::docid_order	order,
		Xapian::valueno	sort_key,
		Xapian::Enquire::Internal::sort_setting	sort_by,
		bool	sort_val_reverse,
		double	time_limit,
		const std::vector< opt_ptr_spy > &	matchspies
	)

Run the match and produce an MSet object.

Parameters

first	Zero-based index of the first result to return (which supports retrieving pages of results).
maxitems	The maximum number of documents to return.
checkatleast	Check at least this many documents. By default Xapian will avoiding considering documents which it can prove can't match, which is faster but can result in a loose bounds on and a poor estimate of the total number of matches - setting checkatleast higher allows trading off speed for tighter bounds and a more accurate estimate.
mset	MSet object to full in
stats	Collated stats
wtscheme	Weight object to use as factory
mdecider	MatchDecider to use (NULL for none)
sorter	KeyMaker for sort keys (NULL for none)
collapse_key	value slot to collapse on (Xapian::BAD_VALUENO which means no collapsing)
collapse_max	Maximum number of documents with the same key to allow
collapse_key	value slot to collapse on (Xapian::BAD_VALUENO which means no collapsing)
collapse_max	Maximum number of documents with the same key to allow
percent_threshold	Lower bound on percentage score
weight_threshold	Lower bound on weight
order	Xapian::docid sort order
sort_key	Value slot to sort on
sort_by	What to sort results on
sort_val_reverse	Reverse direction keys sort in?
time_limit	time in seconds after which to disable check_at_least (0.0 means don't).
matchspies	MatchSpy objects to use

Definition at line 577 of file matcher.cc.

References Assert, AssertRel, db, Xapian::MSet::empty(), for_all_remotes(), Xapian::Internal::intrusive_ptr_nonnull< T >::get(), get_local_mset(), RemoteSubMatch::get_mset(), RemoteSubMatch::get_shard(), if(), Xapian::Database::internal, Xapian::MSet::internal, locals, remotes, and RemoteSubMatch::start_match().

Referenced by Xapian::Enquire::Internal::get_mset(), and RemoteServer::msg_query().

◆ operator=()

Matcher& Matcher::operator= ( const Matcher & )

privatedelete

Member Data Documentation

◆ db

Xapian::Database Matcher::db

private

Definition at line 49 of file matcher.h.

Referenced by get_local_mset(), and get_mset().

◆ first_nonselectable

std::size_t Matcher::first_nonselectable

private

Partition point in remotes.

We call select() on remotes[i]->get_read_fd() for all i < first_nonselectable. See doc comment for remotes for more details.

Definition at line 86 of file matcher.h.

◆ locals

std::vector<std::unique_ptr<LocalSubMatch> > Matcher::locals

private

LocalSubMatch objects for local databases.

The entries are at the same index as the corresponding shard in the Database object, with NULL entries for any remote shards.

Definition at line 56 of file matcher.h.

Referenced by get_local_mset(), and get_mset().

◆ remotes

std::vector<std::unique_ptr<RemoteSubMatch> > Matcher::remotes

private

RemoteSubMatch objects for remote databases.

Unlike locals, this only contains entries for remote shards, and each RemoteSubMatch object knows its shard index.

If poll() isn't available then select() has to be used to wait for fds to become ready to read, which imposes some limitations:

Generally select() only works for fds < FD_SETSIZE
On Microsoft Windows, select() only works for sockets and FD_SETSIZE has a different meaning - it's the maximum number of sockets which we can select() on at once.

To work within these limitations, we order the objects in the vector so the ones we can select() on all come before the ones we can't select() on.

The partition point is recorded in first_nonselectable.

Definition at line 78 of file matcher.h.

Referenced by get_mset().

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

matcher/matcher.h
matcher/matcher.cc

Public Member Functions

Private Types

Private Member Functions

Private Attributes

Detailed Description

Member Typedef Documentation

◆ opt_ptr_spy

Constructor & Destructor Documentation

◆ Matcher() [1/2]

◆ Matcher() [2/2]

Member Function Documentation

◆ for_all_remotes()

◆ get_local_mset()

◆ get_mset()

◆ operator=()

Member Data Documentation

◆ db

◆ first_nonselectable

◆ locals

◆ remotes