#include "univ.i"
#include "data0data.h"
#include "dict0types.h"
#include "row0mysql.h"
#include "fts0fts.h"
#include "fts0types.h"
#include "fts0priv.h"
#include "row0merge.h"

Include dependency graph for row0ftsort.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	fts_doc_item
struct	fts_psort_common_t
struct	fts_psort_t
struct	fts_tokenize_ctx
struct	fts_psort_insert

Macros
#define	FTS_NUM_AUX_INDEX 6
#define	FTS_PLL_MERGE 1
#define	FTS_PARENT_COMPLETE 1
#define	FTS_CHILD_COMPLETE 1
#define	FTS_CHILD_EXITING 2
#define	FTSORT_PRINT
#define	DEBUG_FTS_SORT_PRINT(str)

Typedefs
typedef struct fts_doc_item	fts_doc_item_t
typedef struct fts_tokenize_ctx	fts_tokenize_ctx_t
typedef struct fts_psort_insert	fts_psort_insert_t

Functions
typedef	UT_LIST_BASE_NODE_T (fts_doc_item_t) fts_doc_list_t
UNIV_INTERN dict_index_t *	row_merge_create_fts_sort_index (dict_index_t index, const dict_table_t table, ibool *opt_doc_id_size)
UNIV_INTERN ibool	row_fts_psort_info_init (trx_t trx, row_merge_dup_t dup, const dict_table_t new_table, ibool opt_doc_id_size, fts_psort_t psort, fts_psort_t *merge) __attribute__((nonnull))
UNIV_INTERN void	row_fts_psort_info_destroy (fts_psort_t psort_info, fts_psort_t merge_info)
UNIV_INTERN void	row_fts_free_pll_merge_buf (fts_psort_t *psort_info)
UNIV_INTERN os_thread_ret_t	fts_parallel_tokenization (void *arg)
UNIV_INTERN void	row_fts_start_psort (fts_psort_t *psort_info)
UNIV_INTERN os_thread_ret_t	fts_parallel_merge (void *arg)
UNIV_INTERN void	row_fts_start_parallel_merge (fts_psort_t *merge_info)
UNIV_INTERN void	row_fts_insert_tuple (fts_psort_insert_t ins_ctx, fts_tokenizer_word_t word, ib_vector_t positions, doc_id_t in_doc_id, dtuple_t *dtuple)
UNIV_INTERN int	row_merge_fts_sel_propagate (int propogated, int sel_tree, ulint level, const mrec_t mrec, ulint offsets, dict_index_t index)
UNIV_INTERN dberr_t	row_fts_merge_insert (dict_index_t index, dict_table_t table, fts_psort_t *psort_info, ulint id) __attribute__((nonnull))

Detailed Description

Create Full Text Index with (parallel) merge sort

Created 10/13/2010 Jimmy Yang

Definition in file row0ftsort.h.

Macro Definition Documentation

#define DEBUG_FTS_SORT_PRINT ( str )

Value:

do {                                    \
                ut_print_timestamp(stderr);     \
                fprintf(stderr, str);           \
        } while (0)

Definition at line 134 of file row0ftsort.h.

#define FTS_PARENT_COMPLETE 1

status bit used for communication between parent and child thread

Definition at line 126 of file row0ftsort.h.

#define FTSORT_PRINT

Print some debug information

Definition at line 131 of file row0ftsort.h.

Typedef Documentation

typedef struct fts_doc_item fts_doc_item_t

This structure defineds information the scan thread will fetch

and put to the linked list for parallel tokenization/sort threads to process

Definition at line 41 of file row0ftsort.h.

Function Documentation

UNIV_INTERN os_thread_ret_t fts_parallel_merge ( void * arg )

Function performs the merge and insertion of the sorted records.

Returns: OS_THREAD_DUMMY_RETURN in: parallel merge info

Function performs the merge and insertion of the sorted records.

Returns: OS_THREAD_DUMMY_RETURN

Parameters

arg	in: parallel merge info

Definition at line 881 of file row0ftsort.cc.

Here is the call graph for this function:

Here is the caller graph for this function:

UNIV_INTERN os_thread_ret_t fts_parallel_tokenization ( void * arg )

Function performs parallel tokenization of the incoming doc strings.

Returns: OS_THREAD_DUMMY_RETURN in: psort_info for the thread

Function performs parallel tokenization of the incoming doc strings. It also performs the initial in memory sort of the parsed records.

Returns: OS_THREAD_DUMMY_RETURN

Parameters

arg	in: psort_info for the thread

Definition at line 553 of file row0ftsort.cc.

Here is the call graph for this function:

Here is the caller graph for this function:

UNIV_INTERN void row_fts_free_pll_merge_buf ( fts_psort_t * psort_info )

Free up merge buffers when merge sort is done in: parallel sort info

Free up merge buffers when merge sort is done

Parameters

psort_info in: parallel sort info

Definition at line 332 of file row0ftsort.cc.

Here is the call graph for this function:

UNIV_INTERN void row_fts_insert_tuple	(	fts_psort_insert_t *	ins_ctx,
		fts_tokenizer_word_t *	word,
		ib_vector_t *	positions,
		doc_id_t *	in_doc_id,
		dtuple_t *	dtuple
	)

Read sorted FTS data files and insert data tuples to auxillary tables.

Returns: DB_SUCCESS or error number in: entry to insert

Parameters

ins_ctx	in: insert context
word	in: last processed tokenized word
positions	in: word position
in_doc_id	in: last item doc id

UNIV_INTERN dberr_t row_fts_merge_insert	(	dict_index_t *	index,
		dict_table_t *	table,
		fts_psort_t *	psort_info,
		ulint	id
	)

Read sorted file containing index data tuples and insert these data tuples to the index

Returns: DB_SUCCESS or error number

Parameters

index	in: index
table	in: new table
psort_info	parallel sort info

Here is the caller graph for this function:

UNIV_INTERN void row_fts_psort_info_destroy	(	fts_psort_t *	psort_info,
		fts_psort_t *	merge_info
	)

Clean up and deallocate FTS parallel sort structures, and close temparary merge sort files parallel merge info

Clean up and deallocate FTS parallel sort structures, and close the merge sort files

Parameters

psort_info	parallel sort info
merge_info	parallel merge info

Definition at line 294 of file row0ftsort.cc.

Here is the call graph for this function:

Here is the caller graph for this function:

UNIV_INTERN ibool row_fts_psort_info_init	(	trx_t *	trx,
		row_merge_dup_t *	dup,
		const dict_table_t *	new_table,
		ibool	opt_doc_id_size,
		fts_psort_t **	psort,
		fts_psort_t **	merge
	)

Initialize FTS parallel sort structures.

Returns: TRUE if all successful

Parameters

trx	in: transaction
dup	in,own: descriptor of FTS index being created
new_table	in: table on which indexes are created
opt_doc_id_size	in: whether to use 4 bytes instead of 8 bytes integer to store Doc ID during sort
psort	out: parallel sort info to be instantiated
merge	out: parallel merge info to be instantiated

Definition at line 171 of file row0ftsort.cc.

Here is the call graph for this function:

UNIV_INTERN void row_fts_start_parallel_merge ( fts_psort_t * merge_info )

Kick off the parallel merge and insert thread in: parallel sort info

Kick off the parallel merge and insert thread

Parameters

merge_info in: parallel sort info

Definition at line 913 of file row0ftsort.cc.

Here is the call graph for this function:

UNIV_INTERN void row_fts_start_psort ( fts_psort_t * psort_info )

Start the parallel tokenization and parallel merge sort in: parallel sort info

Start the parallel tokenization and parallel merge sort

Parameters

psort_info parallel sort structure

Definition at line 861 of file row0ftsort.cc.

Here is the call graph for this function:

UNIV_INTERN dict_index_t* row_merge_create_fts_sort_index	(	dict_index_t *	index,
		const dict_table_t *	table,
		ibool *	opt_doc_id_size
	)

Create a temporary "fts sort index" used to merge sort the tokenized doc string. The index has three "fields":

1) Tokenized word, 2) Doc ID 3) Word's position in original 'doc'.

Returns: dict_index_t structure for the fts sort index out: whether to use 4 bytes instead of 8 bytes integer to store Doc ID during sort

Create a temporary "fts sort index" used to merge sort the tokenized doc string. The index has three "fields":

1) Tokenized word, 2) Doc ID (depend on number of records to sort, it can be a 4 bytes or 8 bytes integer value) 3) Word's position in original doc.

Returns: dict_index_t structure for the fts sort index

Parameters

index	in: Original FTS index based on which this sort index is created
table	in: table that FTS index is being created on
opt_doc_id_size	out: whether to use 4 bytes instead of 8 bytes integer to store Doc ID during sort

Definition at line 63 of file row0ftsort.cc.

Here is the call graph for this function:

UNIV_INTERN int row_merge_fts_sel_propagate	(	int	propogated,
		int *	sel_tree,
		ulint	level,
		const mrec_t **	mrec,
		ulint **	offsets,
		dict_index_t *	index
	)

Propagate a newly added record up one level in the selection tree

Returns: parent where this value propagated to

typedef UT_LIST_BASE_NODE_T ( fts_doc_item_t )

This defines the list type that scan thread would feed the parallel

tokenization threads and sort threads.

Classes

Macros

Typedefs

Functions

Detailed Description

Macro Definition Documentation

Typedef Documentation

Function Documentation