manual merge (42af46ca) · Commits · Software / OSDI20 Artifacts / mariadb

Docs/manual.texi

+24 −1

Original line number	Diff line number	Diff line
		@@ -12423,6 +12423,12 @@ mysql> SELECT name, birth FROM pet ORDER BY birth;
		+----------+------------+
		@end example

		On character type columns, sorting@-like all other comparison
		operations@-is normally performed in a case-insensitive fashion.
		This means that the order will be undefined for columns that are identical
		except for their case. You can force a case-sensitive sort by using the
		BINARY cast: @code{ORDER BY BINARY(field)}.

		To sort in reverse order, add the @code{DESC} (descending) keyword to the
		name of the column you are sorting by:

		@@ -29339,6 +29345,15 @@ interpreted as the year @code{'2010-11-12'}. The value @code{'10:45:15'}
		will be converted to @code{'0000-00-00'} because @code{'45'} is not a legal
		month.

		@item
		The MySQL server only performs basic checking on the validity of a date:
		days @code{00-31}, months @code{00-12}, years @code{1000-9999}.
		Any date not within this range will revert to @code{0000-00-00}.
		Please note that this still allows you to store invalid dates such as
		@code{2002-04-31}. It allows web applications to store data from a form
		without further checking. To ensure a date is valid, perform a check in
		your application.

		@item
		Year values specified as two digits are ambiguous, because the century is
		unknown. MySQL interprets 2-digit year values using the following
		@@ -49098,7 +49113,15 @@ Our TODO section contains what we plan to have in 4.0. @xref{TODO MySQL 4.0}.

		@itemize @bullet
		@item
		Query cache disabled in @code{mysqldump}.
		Fixed bug in phrase operator @code{"..."} in boolean fulltext search.
		@item
		Fixed bug that caused duplicated rows when using truncation operator
		@code{*} in boolean fulltext search.
		@item
		Fixed bug in boolean fulltext search, that caused a crash when an identical
		@code{MATCH} expression that did not use an index appeared twice.
		@item
		Query cache disabled in mysqldump.
		@item
		Fixed problem on Windows 98 that made sending of results very slow.
		@item

include/my_getopt.h

+2 −1

Original line number	Diff line number	Diff line
		@@ -16,7 +16,8 @@

		C_MODE_START

		enum get_opt_var_type { GET_NO_ARG, GET_BOOL, GET_LONG, GET_LL, GET_STR };
		enum get_opt_var_type { GET_NO_ARG, GET_BOOL, GET_INT, GET_UINT, GET_LONG,
		GET_ULONG, GET_LL, GET_ULL, GET_STR };
		enum get_opt_arg_type { NO_ARG, OPT_ARG, REQUIRED_ARG };

		struct my_option

myisam/ft_boolean_search.c

+44 −12

Original line number	Diff line number	Diff line
		@@ -77,7 +77,7 @@ typedef struct st_ftb_word {
		my_off_t docid[2]; /* for index search and for scan */
		uint ndepth;
		int len;
		/* ... there can be docid cache added here. SerG */
		/* ... docid cache can be added here. SerG */
		byte word[1];
		} FTB_WORD;

		@@ -90,6 +90,7 @@ typedef struct st_ft_info {
		uint with_scan;
		FTB_EXPR *root;
		QUEUE queue;
		TREE no_dupes;
		FTB_WORD **list;
		MEM_ROOT mem_root;
		} FTB;
		@@ -165,6 +166,7 @@ void _ftb_parse_query(FTB ftb, byte start, byte end,
		if ((ftbe->quot=param.quot)) ftb->with_scan\|=2;
		if (param.yesno > 0) up->ythresh++;
		_ftb_parse_query(ftb, start, end, ftbe, depth+1);
		param.quot=0;
		break;
		case 3: /* right bracket */
		if (up->quot) up->qend=param.quot;
		@@ -174,6 +176,11 @@ void _ftb_parse_query(FTB ftb, byte start, byte end,
		return;
		}

		static int _ftb_no_dupes_cmp(void* not_used, const void a,const void b)
		{
		return CMP_NUM((((my_off_t)a)), (((my_off_t)b)));
		}

		void _ftb_init_index_search(FT_INFO *ftb)
		{
		int i, r;
		@@ -193,8 +200,8 @@ void _ftb_init_index_search(FT_INFO *ftb)
		{
		ftbw=(FTB_WORD *)(ftb->queue.root[i]);

		if (ftbw->flags&FTB_FLAG_TRUNC &&
		(ftbw->up->ythresh > test(ftbw->flags&FTB_FLAG_YES)))
		if (ftbw->flags&FTB_FLAG_TRUNC) /* special treatment :(( */
		if (ftbw->up->ythresh > test(ftbw->flags&FTB_FLAG_YES))
		{
		/* no need to search for this prefix in the index -
		* it cannot ADD new matches, and to REMOVE half-matched
		@@ -203,6 +210,21 @@ void _ftb_init_index_search(FT_INFO *ftb)
		ftbw->up->yweaks++;
		continue;
		}
		else
		{
		/* We have to index-search for this prefix.
		* It may cause duplicates, as in the index (sorted by <word,docid>)
		* <aaaa,row1>
		* <aabb,row2>
		* <aacc,row1>
		* Searching for "aa*" will find row1 twice...
		*/
		if (!is_tree_inited(& ftb->no_dupes))
		{
		init_tree(& ftb->no_dupes,0,0,sizeof(my_off_t),
		_ftb_no_dupes_cmp,0,0,0);
		}
		}

		r=_mi_search(info, keyinfo, (uchar*) ftbw->word, ftbw->len,
		SEARCH_FIND \| SEARCH_BIGGER, keyroot);
		@@ -250,6 +272,7 @@ FT_INFO * ft_init_boolean_search(MI_INFO info, uint keynr, byte query,
		default_charset_info :
		info->s->keyinfo[keynr].seg->charset);
		ftb->with_scan=0;
		bzero(& ftb->no_dupes, sizeof(TREE));

		init_alloc_root(&ftb->mem_root, 1024, 1024);

		@@ -438,6 +461,11 @@ int ft_boolean_read_next(FT_INFO ftb, char record)
		ftbe->yesses>=(ftbe->ythresh-ftbe->yweaks) && !ftbe->nos)
		{
		/* curdoc matched ! */
		if (is_tree_inited(& ftb->no_dupes) &&
		tree_insert(& ftb->no_dupes, &curdoc, 0)->count >1)
		/* but it managed to get past this line once */
		continue;

		info->lastpos=curdoc;
		info->update&= (HA_STATE_CHANGED \| HA_STATE_ROW_CHANGED); /* why is this ? */

		@@ -523,6 +551,10 @@ float ft_boolean_find_relevance(FT_INFO ftb, byte record, uint length)

		void ft_boolean_close_search(FT_INFO *ftb)
		{
		if (is_tree_inited(& ftb->no_dupes))
		{
		delete_tree(& ftb->no_dupes);
		}
		free_root(& ftb->mem_root, MYF(0));
		my_free((gptr)ftb,MYF(0));
		}

myisam/ft_parser.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -134,7 +134,7 @@ byte ft_get_word(byte *start, byte end, FT_WORD word, FTB_PARAM param)
		{
		if (true_word_char(*doc)) break;
		if (*doc == FTB_RQUOT && param->quot) {
		param->quot=doc-1;
		param->quot=doc;
		*start=doc+1;
		return 3; /* FTB_RBR */
		}

mysql-test/r/fulltext.result

+15 −0

Original line number	Diff line number	Diff line
		@@ -70,6 +70,16 @@ Full-text search in MySQL implements vector space model 0
		select * from t1 where MATCH a,b AGAINST ('"Now sUPPort"' IN BOOLEAN MODE);
		a b
		MySQL has now support for full-text search
		select * from t1 where MATCH a,b AGAINST ('"text search" "now support"' IN BOOLEAN MODE);
		a b
		MySQL has now support for full-text search
		Full-text search in MySQL implements vector space model
		select * from t1 where MATCH a,b AGAINST ('"text search" -"now support"' IN BOOLEAN MODE);
		a b
		Full-text search in MySQL implements vector space model
		select * from t1 where MATCH a,b AGAINST ('"text search" +"now support"' IN BOOLEAN MODE);
		a b
		MySQL has now support for full-text search
		select * from t1 where MATCH a AGAINST ("search" IN BOOLEAN MODE);
		a b
		Full-text search in MySQL implements vector space model
		@@ -173,4 +183,9 @@ CREATE TABLE t1 (a int(11), b text, FULLTEXT KEY (b)) TYPE=MyISAM;
		insert into t1 values (1,"I wonder why the fulltext index doesnt work?");
		SELECT * from t1 where MATCH (b) AGAINST ('apples');
		a b
		insert into t1 values (2,"fullaaa fullzzz");
		select * from t1 where match b against ('full*' in boolean mode);
		a b
		2 fullaaa fullzzz
		1 I wonder why the fulltext index doesnt work?
		drop table t1;