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

	*(state.curop) = '\0';

	/* parse query & make polish notation (postfix, but in reverse order) */
	makepol(&state, pushval, opaque);

	close_tsvector_parser(state.valstate);

	if (list_length(state.polstr) == 0)
	{
		ereport(NOTICE,
				(errmsg("text-search query doesn't contain lexemes: \"%s\"",
						state.buffer)));
		query = (TSQuery) palloc(HDRSIZETQ);
		SET_VARSIZE(query, HDRSIZETQ);
		query->size = 0;
		return query;
	}

	/* Pack the QueryItems in the final TSQuery struct to return to caller */
	commonlen = COMPUTESIZE(list_length(state.polstr), state.sumlen);
	query = (TSQuery) palloc0(commonlen);
	SET_VARSIZE(query, commonlen);
	query->size = list_length(state.polstr);
	ptr = GETQUERY(query);

	/* Copy QueryItems to TSQuery */
	i = 0;
	foreach(cell, state.polstr)
	{
		QueryItem  *item = (QueryItem *) lfirst(cell);

		switch (item->type)
		{
			case QI_VAL:
				memcpy(&ptr[i], item, sizeof(QueryOperand));
				break;
			case QI_VALSTOP:
				ptr[i].type = QI_VALSTOP;
				break;
			case QI_OPR:
				memcpy(&ptr[i], item, sizeof(QueryOperator));
				break;
			default:
				elog(ERROR, "unrecognized QueryItem type: %d", item->type);
		}
		i++;
	}

	/* Copy all the operand strings to TSQuery */
	memcpy((void *) GETOPERAND(query), (void *) state.op, state.sumlen);
	pfree(state.op);

	/* Set left operand pointers for every operator. */
	findoprnd(ptr, query->size);

	return query;
}

static void
pushval_asis(Datum opaque, TSQueryParserState state, char *strval, int lenval,
			 int16 weight)
{
	pushValue(state, strval, lenval, weight);
}

/*
 * in without morphology
 */
Datum
tsqueryin(PG_FUNCTION_ARGS)
{
	char	   *in = PG_GETARG_CSTRING(0);

	pg_verifymbstr(in, strlen(in), false);

	PG_RETURN_TSQUERY(parse_tsquery(in, pushval_asis, PointerGetDatum(NULL), false));
}

/*
 * out function
 */
typedef struct
{
	QueryItem  *curpol;
	char	   *buf;
	char	   *cur;
	char	   *op;
	int			buflen;
} INFIX;

/* Makes sure inf->buf is large enough for adding 'addsize' bytes */
#define RESIZEBUF(inf, addsize) \
while( ( (inf)->cur - (inf)->buf ) + (addsize) + 1 >= (inf)->buflen ) \
{ \
	int len = (inf)->cur - (inf)->buf; \
	(inf)->buflen *= 2; \
	(inf)->buf = (char*) repalloc( (void*)(inf)->buf, (inf)->buflen ); \
	(inf)->cur = (inf)->buf + len; \
}

/*
 * recursive walk on tree and print it in
 * infix (human-readable) view
 */
static void
infix(INFIX *in, bool first)
{
	/* since this function recurses, it could be driven to stack overflow. */
	check_stack_depth();

	if (in->curpol->type == QI_VAL)
	{
		QueryOperand *curpol = &in->curpol->operand;
		char	   *op = in->op + curpol->distance;
		int			clen;

		RESIZEBUF(in, curpol->length * (pg_database_encoding_max_length() + 1) + 2 + 5);
		*(in->cur) = '\'';
		in->cur++;
		while (*op)
		{
			if (t_iseq(op, '\''))
			{
				*(in->cur) = '\'';
				in->cur++;
			}
			else if (t_iseq(op, '\\'))
			{
				*(in->cur) = '\\';
				in->cur++;
			}
			COPYCHAR(in->cur, op);

			clen = pg_mblen(op);
			op += clen;
			in->cur += clen;
		}
		*(in->cur) = '\'';
		in->cur++;
		if (curpol->weight)
		{
			*(in->cur) = ':';
			in->cur++;
			if (curpol->weight & (1 << 3))
			{
				*(in->cur) = 'A';
				in->cur++;
			}
			if (curpol->weight & (1 << 2))
			{
				*(in->cur) = 'B';
				in->cur++;
			}
			if (curpol->weight & (1 << 1))
			{
				*(in->cur) = 'C';
				in->cur++;
			}
			if (curpol->weight & 1)
			{
				*(in->cur) = 'D';
				in->cur++;
			}
		}
		*(in->cur) = '\0';
		in->curpol++;
	}
	else if (in->curpol->operator.oper == OP_NOT)
	{
		bool		isopr = false;

		RESIZEBUF(in, 1);
		*(in->cur) = '!';
		in->cur++;
		*(in->cur) = '\0';
		in->curpol++;

		if (in->curpol->type == QI_OPR)
		{
			isopr = true;
			RESIZEBUF(in, 2);
			sprintf(in->cur, "( ");
			in->cur = strchr(in->cur, '\0');
		}

		infix(in, isopr);
		if (isopr)
		{
			RESIZEBUF(in, 2);
			sprintf(in->cur, " )");
			in->cur = strchr(in->cur, '\0');
		}
	}
	else
	{
		int8		op = in->curpol->operator.oper;
		INFIX		nrm;

		in->curpol++;
		if (op == OP_OR && !first)
		{
			RESIZEBUF(in, 2);
			sprintf(in->cur, "( ");
			in->cur = strchr(in->cur, '\0');
		}

		nrm.curpol = in->curpol;
		nrm.op = in->op;
		nrm.buflen = 16;
		nrm.cur = nrm.buf = (char *) palloc(sizeof(char) * nrm.buflen);

		/* get right operand */
		infix(&nrm, false);

		/* get & print left operand */
		in->curpol = nrm.curpol;
		infix(in, false);

		/* print operator & right operand */
		RESIZEBUF(in, 3 + (nrm.cur - nrm.buf));
		switch (op)
		{
			case OP_OR:
				sprintf(in->cur, " | %s", nrm.buf);
				break;
			case OP_AND:
				sprintf(in->cur, " & %s", nrm.buf);
				break;
			default:
				/* OP_NOT is handled in above if-branch */
				elog(ERROR, "unrecognized operator type: %d", op);
		}
		in->cur = strchr(in->cur, '\0');
		pfree(nrm.buf);

		if (op == OP_OR && !first)
		{
			RESIZEBUF(in, 2);
			sprintf(in->cur, " )");
			in->cur = strchr(in->cur, '\0');
		}
	}
}


Datum
tsqueryout(PG_FUNCTION_ARGS)
{
	TSQuery		query = PG_GETARG_TSQUERY(0);
	INFIX		nrm;

	if (query->size == 0)
	{
		char	   *b = palloc(1);

		*b = '\0';
		PG_RETURN_POINTER(b);
	}
	nrm.curpol = GETQUERY(query);
	nrm.buflen = 32;
	nrm.cur = nrm.buf = (char *) palloc(sizeof(char) * nrm.buflen);
	*(nrm.cur) = '\0';
	nrm.op = GETOPERAND(query);
	infix(&nrm, true);

	PG_FREE_IF_COPY(query, 0);
	PG_RETURN_CSTRING(nrm.buf);
}

/*
 * Binary Input / Output functions. The binary format is as follows:
 *
 * uint32	 number of operators/operands in the query
 *
 * Followed by the operators and operands, in prefix notation. For each
 * operand:
 *
 * uint8	type, QI_VAL
 * uint8	weight
 *			operand text in client encoding, null-terminated
 *
 * For each operator:
 * uint8	type, QI_OPR
 * uint8	operator, one of OP_AND, OP_OR, OP_NOT.
 */
Datum
tsquerysend(PG_FUNCTION_ARGS)
{
	TSQuery		query = PG_GETARG_TSQUERY(0);
	StringInfoData buf;
	int			i;
	QueryItem  *item = GETQUERY(query);

	pq_begintypsend(&buf);

	pq_sendint(&buf, query->size, sizeof(uint32));
	for (i = 0; i < query->size; i++)
	{
		pq_sendint(&buf, item->type, sizeof(item->type));

		switch (item->type)
		{
			case QI_VAL:
				pq_sendint(&buf, item->operand.weight, sizeof(uint8));
				pq_sendstring(&buf, GETOPERAND(query) + item->operand.distance);
				break;
			case QI_OPR:
				pq_sendint(&buf, item->operator.oper, sizeof(item->operator.oper));
				break;
			default:
				elog(ERROR, "unrecognized tsquery node type: %d", item->type);
		}
		item++;
	}

	PG_FREE_IF_COPY(query, 0);

	PG_RETURN_BYTEA_P(pq_endtypsend(&buf));
}

Datum
tsqueryrecv(PG_FUNCTION_ARGS)
{
	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
	TSQuery		query;
	int			i,
				len;
	QueryItem  *item;
	int			datalen;
	char	   *ptr;
	uint32		size;
	const char **operands;

	size = pq_getmsgint(buf, sizeof(uint32));
	if (size > (MaxAllocSize / sizeof(QueryItem)))
		elog(ERROR, "invalid size of tsquery");

	/* Allocate space to temporarily hold operand strings */
	operands = palloc(size * sizeof(char *));

	/* Allocate space for all the QueryItems. */
	len = HDRSIZETQ + sizeof(QueryItem) * size;
	query = (TSQuery) palloc0(len);
	query->size = size;
	item = GETQUERY(query);

	datalen = 0;
	for (i = 0; i < size; i++)
	{
		item->type = (int8) pq_getmsgint(buf, sizeof(int8));

		if (item->type == QI_VAL)
		{
			size_t		val_len;	/* length after recoding to server encoding */
			uint8		weight;
			const char *val;
			pg_crc32	valcrc;

			weight = (uint8) pq_getmsgint(buf, sizeof(uint8));
			val = pq_getmsgstring(buf);
			val_len = strlen(val);

			/* Sanity checks */

			if (weight > 0xF)
				elog(ERROR, "invalid tsquery: invalid weight bitmap");

			if (val_len > MAXSTRLEN)
				elog(ERROR, "invalid tsquery: operand too long");

			if (datalen > MAXSTRPOS)
				elog(ERROR, "invalid tsquery: total operand length exceeded");

			/* Looks valid. */

			INIT_CRC32(valcrc);
			COMP_CRC32(valcrc, val, val_len);
			FIN_CRC32(valcrc);

			item->operand.weight = weight;
			item->operand.valcrc = (int32) valcrc;
			item->operand.length = val_len;
			item->operand.distance = datalen;

			/*
			 * Operand strings are copied to the final struct after this loop;
			 * here we just collect them to an array
			 */
			operands[i] = val;

			datalen += val_len + 1;		/* + 1 for the '\0' terminator */
		}
		else if (item->type == QI_OPR)
		{
			int8		oper;

			oper = (int8) pq_getmsgint(buf, sizeof(int8));
			if (oper != OP_NOT && oper != OP_OR && oper != OP_AND)
				elog(ERROR, "invalid tsquery: unrecognized operator type %d",
					 (int) oper);
			if (i == size - 1)
				elog(ERROR, "invalid pointer to right operand");

			item->operator.oper = oper;
		}
		else
			elog(ERROR, "unrecognized tsquery node type: %d", item->type);

		item++;
	}

	/* Enlarge buffer to make room for the operand values. */
	query = (TSQuery) repalloc(query, len + datalen);
	item = GETQUERY(query);
	ptr = GETOPERAND(query);

	/*
	 * Fill in the left-pointers. Checks that the tree is well-formed as a
	 * side-effect.
	 */
	findoprnd(item, size);

	/* Copy operands to output struct */
	for (i = 0; i < size; i++)
	{
		if (item->type == QI_VAL)
		{
			memcpy(ptr, operands[i], item->operand.length + 1);
			ptr += item->operand.length + 1;
		}
		item++;
	}

	pfree(operands);

	Assert(ptr - GETOPERAND(query) == datalen);

	SET_VARSIZE(query, len + datalen);

	PG_RETURN_TSVECTOR(query);
}

/*
 * debug function, used only for view query
 * which will be executed in non-leaf pages in index
 */
Datum
tsquerytree(PG_FUNCTION_ARGS)
{
	TSQuery		query = PG_GETARG_TSQUERY(0);
	INFIX		nrm;
	text	   *res;
	QueryItem  *q;
	int			len;

	if (query->size == 0)
	{
		res = (text *) palloc(VARHDRSZ);
		SET_VARSIZE(res, VARHDRSZ);
		PG_RETURN_POINTER(res);
	}

	q = clean_NOT(GETQUERY(query), &len);

	if (!q)
	{
		res = (text *) palloc(1 + VARHDRSZ);
		SET_VARSIZE(res, 1 + VARHDRSZ);
		*((char *) VARDATA(res)) = 'T';
	}
	else
	{
		nrm.curpol = q;
		nrm.buflen = 32;
		nrm.cur = nrm.buf = (char *) palloc(sizeof(char) * nrm.buflen);
		*(nrm.cur) = '\0';
		nrm.op = GETOPERAND(query);
		infix(&nrm, true);

		res = (text *) palloc(nrm.cur - nrm.buf + VARHDRSZ);
		SET_VARSIZE(res, nrm.cur - nrm.buf + VARHDRSZ);
		strncpy(VARDATA(res), nrm.buf, nrm.cur - nrm.buf);
		pfree(q);
	}

	PG_FREE_IF_COPY(query, 0);

	PG_RETURN_POINTER(res);
}