tsquery.c

来自「postgresql8.3.4源码,开源数据库」· C语言 代码 · 共 980 行 · 第 1/2 页

/*-------------------------------------------------------------------------
 *
 * tsquery.c
 *	  I/O functions for tsquery
 *
 * Portions Copyright (c) 1996-2008, PostgreSQL Global Development Group
 *
 *
 * IDENTIFICATION
 *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery.c,v 1.15.2.1 2008/04/11 22:52:17 tgl Exp $
 *
 *-------------------------------------------------------------------------
 */

#include "postgres.h"

#include "libpq/pqformat.h"
#include "miscadmin.h"
#include "tsearch/ts_locale.h"
#include "tsearch/ts_type.h"
#include "tsearch/ts_utils.h"
#include "utils/memutils.h"
#include "utils/pg_crc.h"


struct TSQueryParserStateData
{
	/* State for gettoken_query */
	char	   *buffer;			/* entire string we are scanning */
	char	   *buf;			/* current scan point */
	int			state;
	int			count;			/* nesting count, incremented by (,
								 * decremented by ) */

	/* polish (prefix) notation in list, filled in by push* functions */
	List	   *polstr;

	/*
	 * Strings from operands are collected in op. curop is a pointer to the
	 * end of used space of op.
	 */
	char	   *op;
	char	   *curop;
	int			lenop;			/* allocated size of op */
	int			sumlen;			/* used size of op */

	/* state for value's parser */
	TSVectorParseState valstate;
};

/* parser's states */
#define WAITOPERAND 1
#define WAITOPERATOR	2
#define WAITFIRSTOPERAND 3
#define WAITSINGLEOPERAND 4

/*
 * subroutine to parse the weight part, like ':1AB' of a query.
 */
static char *
get_weight(char *buf, int16 *weight)
{
	*weight = 0;

	if (!t_iseq(buf, ':'))
		return buf;

	buf++;
	while (*buf && pg_mblen(buf) == 1)
	{
		switch (*buf)
		{
			case 'a':
			case 'A':
				*weight |= 1 << 3;
				break;
			case 'b':
			case 'B':
				*weight |= 1 << 2;
				break;
			case 'c':
			case 'C':
				*weight |= 1 << 1;
				break;
			case 'd':
			case 'D':
				*weight |= 1;
				break;
			default:
				return buf;
		}
		buf++;
	}

	return buf;
}

/*
 * token types for parsing
 */
typedef enum
{
	PT_END = 0,
	PT_ERR = 1,
	PT_VAL = 2,
	PT_OPR = 3,
	PT_OPEN = 4,
	PT_CLOSE = 5
} ts_tokentype;

/*
 * get token from query string
 *
 * *operator is filled in with OP_* when return values is PT_OPR
 * *strval, *lenval and *weight are filled in when return value is PT_VAL
 */
static ts_tokentype
gettoken_query(TSQueryParserState state,
			   int8 *operator,
			   int *lenval, char **strval, int16 *weight)
{
	while (1)
	{
		switch (state->state)
		{
			case WAITFIRSTOPERAND:
			case WAITOPERAND:
				if (t_iseq(state->buf, '!'))
				{
					(state->buf)++;		/* can safely ++, t_iseq guarantee
										 * that pg_mblen()==1 */
					*operator = OP_NOT;
					state->state = WAITOPERAND;
					return PT_OPR;
				}
				else if (t_iseq(state->buf, '('))
				{
					state->count++;
					(state->buf)++;
					state->state = WAITOPERAND;
					return PT_OPEN;
				}
				else if (t_iseq(state->buf, ':'))
				{
					ereport(ERROR,
							(errcode(ERRCODE_SYNTAX_ERROR),
							 errmsg("syntax error in tsquery: \"%s\"",
									state->buffer)));
				}
				else if (!t_isspace(state->buf))
				{
					/*
					 * We rely on the tsvector parser to parse the value for
					 * us
					 */
					reset_tsvector_parser(state->valstate, state->buf);
					if (gettoken_tsvector(state->valstate, strval, lenval, NULL, NULL, &state->buf))
					{
						state->buf = get_weight(state->buf, weight);
						state->state = WAITOPERATOR;
						return PT_VAL;
					}
					else if (state->state == WAITFIRSTOPERAND)
						return PT_END;
					else
						ereport(ERROR,
								(errcode(ERRCODE_SYNTAX_ERROR),
								 errmsg("no operand in tsquery: \"%s\"",
										state->buffer)));
				}
				break;
			case WAITOPERATOR:
				if (t_iseq(state->buf, '&'))
				{
					state->state = WAITOPERAND;
					*operator = OP_AND;
					(state->buf)++;
					return PT_OPR;
				}
				if (t_iseq(state->buf, '|'))
				{
					state->state = WAITOPERAND;
					*operator = OP_OR;
					(state->buf)++;
					return PT_OPR;
				}
				else if (t_iseq(state->buf, ')'))
				{
					(state->buf)++;
					state->count--;
					return (state->count < 0) ? PT_ERR : PT_CLOSE;
				}
				else if (*(state->buf) == '\0')
					return (state->count) ? PT_ERR : PT_END;
				else if (!t_isspace(state->buf))
					return PT_ERR;
				break;
			case WAITSINGLEOPERAND:
				if (*(state->buf) == '\0')
					return PT_END;
				*strval = state->buf;
				*lenval = strlen(state->buf);
				state->buf += strlen(state->buf);
				state->count++;
				return PT_VAL;
			default:
				return PT_ERR;
				break;
		}
		state->buf += pg_mblen(state->buf);
	}
	return PT_END;
}

/*
 * Push an operator to state->polstr
 */
void
pushOperator(TSQueryParserState state, int8 oper)
{
	QueryOperator *tmp;

	Assert(oper == OP_NOT || oper == OP_AND || oper == OP_OR);

	tmp = (QueryOperator *) palloc0(sizeof(QueryOperator));
	tmp->type = QI_OPR;
	tmp->oper = oper;
	/* left is filled in later with findoprnd */

	state->polstr = lcons(tmp, state->polstr);
}

static void
pushValue_internal(TSQueryParserState state, pg_crc32 valcrc, int distance, int lenval, int weight)
{
	QueryOperand *tmp;

	if (distance >= MAXSTRPOS)
		ereport(ERROR,
				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
				 errmsg("value is too big in tsquery: \"%s\"",
						state->buffer)));
	if (lenval >= MAXSTRLEN)
		ereport(ERROR,
				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
				 errmsg("operand is too long in tsquery: \"%s\"",
						state->buffer)));

	tmp = (QueryOperand *) palloc0(sizeof(QueryOperand));
	tmp->type = QI_VAL;
	tmp->weight = weight;
	tmp->valcrc = (int32) valcrc;
	tmp->length = lenval;
	tmp->distance = distance;

	state->polstr = lcons(tmp, state->polstr);
}

/*
 * Push an operand to state->polstr.
 *
 * strval must point to a string equal to state->curop. lenval is the length
 * of the string.
 */
void
pushValue(TSQueryParserState state, char *strval, int lenval, int2 weight)
{
	pg_crc32	valcrc;

	if (lenval >= MAXSTRLEN)
		ereport(ERROR,
				(errcode(ERRCODE_PROGRAM_LIMIT_EXCEEDED),
				 errmsg("word is too long in tsquery: \"%s\"",
						state->buffer)));

	INIT_CRC32(valcrc);
	COMP_CRC32(valcrc, strval, lenval);
	FIN_CRC32(valcrc);
	pushValue_internal(state, valcrc, state->curop - state->op, lenval, weight);

	/* append the value string to state.op, enlarging buffer if needed first */
	while (state->curop - state->op + lenval + 1 >= state->lenop)
	{
		int			used = state->curop - state->op;

		state->lenop *= 2;
		state->op = (char *) repalloc((void *) state->op, state->lenop);
		state->curop = state->op + used;
	}
	memcpy((void *) state->curop, (void *) strval, lenval);
	state->curop += lenval;
	*(state->curop) = '\0';
	state->curop++;
	state->sumlen += lenval + 1 /* \0 */ ;
}


/*
 * Push a stopword placeholder to state->polstr
 */
void
pushStop(TSQueryParserState state)
{
	QueryOperand *tmp;

	tmp = (QueryOperand *) palloc0(sizeof(QueryOperand));
	tmp->type = QI_VALSTOP;

	state->polstr = lcons(tmp, state->polstr);
}


#define STACKDEPTH	32

/*
 * Make polish (prefix) notation of query.
 *
 * See parse_tsquery for explanation of pushval.
 */
static void
makepol(TSQueryParserState state,
		PushFunction pushval,
		Datum opaque)
{
	int8		operator = 0;
	ts_tokentype type;
	int			lenval = 0;
	char	   *strval = NULL;
	int8		opstack[STACKDEPTH];
	int			lenstack = 0;
	int16		weight = 0;

	/* since this function recurses, it could be driven to stack overflow */
	check_stack_depth();

	while ((type = gettoken_query(state, &operator, &lenval, &strval, &weight)) != PT_END)
	{
		switch (type)
		{
			case PT_VAL:
				pushval(opaque, state, strval, lenval, weight);
				while (lenstack && (opstack[lenstack - 1] == OP_AND ||
									opstack[lenstack - 1] == OP_NOT))
				{
					lenstack--;
					pushOperator(state, opstack[lenstack]);
				}
				break;
			case PT_OPR:
				if (lenstack && operator == OP_OR)
					pushOperator(state, OP_OR);
				else
				{
					if (lenstack == STACKDEPTH) /* internal error */
						elog(ERROR, "tsquery stack too small");
					opstack[lenstack] = operator;
					lenstack++;
				}
				break;
			case PT_OPEN:
				makepol(state, pushval, opaque);

				if (lenstack && (opstack[lenstack - 1] == OP_AND ||
								 opstack[lenstack - 1] == OP_NOT))
				{
					lenstack--;
					pushOperator(state, opstack[lenstack]);
				}
				break;
			case PT_CLOSE:
				while (lenstack)
				{
					lenstack--;
					pushOperator(state, opstack[lenstack]);
				};
				return;
			case PT_ERR:
			default:
				ereport(ERROR,
						(errcode(ERRCODE_SYNTAX_ERROR),
						 errmsg("syntax error in tsquery: \"%s\"",
								state->buffer)));
		}
	}
	while (lenstack)
	{
		lenstack--;
		pushOperator(state, opstack[lenstack]);
	}
}

static void
findoprnd_recurse(QueryItem *ptr, uint32 *pos, int nnodes)
{
	/* since this function recurses, it could be driven to stack overflow. */
	check_stack_depth();

	if (*pos >= nnodes)
		elog(ERROR, "malformed tsquery: operand not found");

	if (ptr[*pos].type == QI_VAL ||
		ptr[*pos].type == QI_VALSTOP)	/* need to handle VALSTOP here, they
										 * haven't been cleaned away yet. */
	{
		(*pos)++;
	}
	else
	{
		Assert(ptr[*pos].type == QI_OPR);

		if (ptr[*pos].operator.oper == OP_NOT)
		{
			ptr[*pos].operator.left = 1;
			(*pos)++;
			findoprnd_recurse(ptr, pos, nnodes);
		}
		else
		{
			QueryOperator *curitem = &ptr[*pos].operator;
			int			tmp = *pos;

			Assert(curitem->oper == OP_AND || curitem->oper == OP_OR);

			(*pos)++;
			findoprnd_recurse(ptr, pos, nnodes);
			curitem->left = *pos - tmp;
			findoprnd_recurse(ptr, pos, nnodes);
		}
	}
}


/*
 * Fills in the left-fields previously left unfilled. The input
 * QueryItems must be in polish (prefix) notation.
 */
static void
findoprnd(QueryItem *ptr, int size)
{
	uint32		pos;

	pos = 0;
	findoprnd_recurse(ptr, &pos, size);

	if (pos != size)
		elog(ERROR, "malformed tsquery: extra nodes");
}


/*
 * Each value (operand) in the query is be passed to pushval. pushval can
 * transform the simple value to an arbitrarily complex expression using
 * pushValue and pushOperator. It must push a single value with pushValue,
 * a complete expression with all operands, or a a stopword placeholder
 * with pushStop, otherwise the prefix notation representation will be broken,
 * having an operator with no operand.
 *
 * opaque is passed on to pushval as is, pushval can use it to store its
 * private state.
 *
 * The returned query might contain QI_STOPVAL nodes. The caller is responsible
 * for cleaning them up (with clean_fakeval)
 */
TSQuery
parse_tsquery(char *buf,
			  PushFunction pushval,
			  Datum opaque,
			  bool isplain)
{
	struct TSQueryParserStateData state;
	int			i;
	TSQuery		query;
	int			commonlen;
	QueryItem  *ptr;
	ListCell   *cell;

	/* init state */
	state.buffer = buf;
	state.buf = buf;
	state.state = (isplain) ? WAITSINGLEOPERAND : WAITFIRSTOPERAND;
	state.count = 0;
	state.polstr = NIL;

	/* init value parser's state */
	state.valstate = init_tsvector_parser(state.buffer, true, true);

	/* init list of operand */
	state.sumlen = 0;
	state.lenop = 64;
	state.curop = state.op = (char *) palloc(state.lenop);