Improvements from Heikki Linnakangas <heikki@enterprisedb.com>

- change the alignment requirement of lexemes in TSVector slightly.
Lexeme strings were always padded to 2-byte aligned length to make sure
that if there's position array (uint16[]) it has the right alignment.
The patch changes that so that the padding is not done when there's no
positions. That makes the storage of tsvectors without positions
slightly more compact.

- added some #include "miscadmin.h" lines I missed in the earlier when I
added calls to check_stack_depth().

- Reimplement the send/recv functions, and added a comment
above them describing the on-wire format. The CRC is now recalculated in
tsquery as well per previous discussion.

src/backend/utils/adt/tsginidx.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsginidx.c,v 1.2 2007/09/07 15:09:56 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsginidx.c,v 1.3 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -22,7 +22,7 @@ Datum
 gin_extract_tsvector(PG_FUNCTION_ARGS)
 {
 	TSVector	vector = PG_GETARG_TSVECTOR(0);
-	uint32	   *nentries = (uint32 *) PG_GETARG_POINTER(1);
+	int32	   *nentries = (int32 *) PG_GETARG_POINTER(1);
 	Datum	   *entries = NULL;
 
 	*nentries = 0;
@@ -55,7 +55,7 @@ Datum
 gin_extract_query(PG_FUNCTION_ARGS)
 {
 	TSQuery		query = PG_GETARG_TSQUERY(0);
-	uint32	   *nentries = (uint32 *) PG_GETARG_POINTER(1);
+	int32	   *nentries = (int32 *) PG_GETARG_POINTER(1);
 	StrategyNumber strategy = PG_GETARG_UINT16(2);
 	Datum	   *entries = NULL;
 

src/backend/utils/adt/tsquery_cleanup.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_cleanup.c,v 1.3 2007/09/07 15:35:10 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_cleanup.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -17,6 +17,7 @@
 
 #include "tsearch/ts_type.h"
 #include "tsearch/ts_utils.h"
+#include "miscadmin.h"
 
 typedef struct NODE
 {

src/backend/utils/adt/tsquery_rewrite.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_rewrite.c,v 1.3 2007/09/07 15:35:10 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_rewrite.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -17,6 +17,7 @@
 #include "executor/spi.h"
 #include "tsearch/ts_type.h"
 #include "tsearch/ts_utils.h"
+#include "miscadmin.h"
 
 
 static int

src/backend/utils/adt/tsquery_util.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_util.c,v 1.3 2007/09/07 15:35:10 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsquery_util.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -16,6 +16,7 @@
 
 #include "tsearch/ts_type.h"
 #include "tsearch/ts_utils.h"
+#include "miscadmin.h"
 
 QTNode *
 QT2QTN(QueryItem * in, char *operand)

src/backend/utils/adt/tsrank.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsrank.c,v 1.3 2007/09/07 15:35:10 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsrank.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -18,6 +18,7 @@
 #include "tsearch/ts_type.h"
 #include "tsearch/ts_utils.h"
 #include "utils/array.h"
+#include "miscadmin.h"
 
 
 static float weights[] = {0.1, 0.2, 0.4, 1.0};
@@ -176,8 +177,9 @@ SortAndUniqItems(TSQuery q, int *size)
 	return res;
 }
 
+/* A dummy WordEntryPos array to use when haspos is false */
 static WordEntryPos POSNULL[] = {
-	0,
+	1, /* Number of elements that follow */
 	0
 };
 
@@ -207,7 +209,6 @@ calc_rank_and(float *w, TSVector t, TSQuery q)
 	}
 	pos = (uint16 **) palloc(sizeof(uint16 *) * q->size);
 	memset(pos, 0, sizeof(uint16 *) * q->size);
-	*(uint16 *) POSNULL = lengthof(POSNULL) - 1;
 	WEP_SETPOS(POSNULL[1], MAXENTRYPOS - 1);
 
 	for (i = 0; i < size; i++)
@@ -265,7 +266,6 @@ calc_rank_or(float *w, TSVector t, TSQuery q)
 	QueryOperand **item;
 	int			size = q->size;
 
-	*(uint16 *) POSNULL = lengthof(POSNULL) - 1;
 	item = SortAndUniqItems(q, &size);
 
 	for (i = 0; i < size; i++)
@@ -593,7 +593,6 @@ get_docrep(TSVector txt, TSQuery query, int *doclen)
 	DocRepresentation *doc;
 	char	   *operand;
 
-	*(uint16 *) POSNULL = lengthof(POSNULL) - 1;
 	doc = (DocRepresentation *) palloc(sizeof(DocRepresentation) * len);
 	operand = GETOPERAND(query);
 	reset_istrue_flag(query);

src/backend/utils/adt/tsvector.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector.c,v 1.3 2007/09/07 15:09:56 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -75,18 +75,20 @@ uniquePos(WordEntryPos * a, int l)
 }
 
 static int
-compareentry(const void *a, const void *b, void *arg)
+compareentry(const void *va, const void *vb, void *arg)
 {
 	char	   *BufferStr = (char *) arg;
+	WordEntryIN *a = (WordEntryIN *) va;
+	WordEntryIN *b = (WordEntryIN *) vb;
 
-	if (((WordEntryIN *) a)->entry.len == ((WordEntryIN *) b)->entry.len)
+	if (a->entry.len == b->entry.len)
 	{
-		return strncmp(&BufferStr[((WordEntryIN *) a)->entry.pos],
-					   &BufferStr[((WordEntryIN *) b)->entry.pos],
-					   ((WordEntryIN *) a)->entry.len);
+		return strncmp(&BufferStr[a->entry.pos],
+					   &BufferStr[b->entry.pos],
+					   a->entry.len);
 	}
 
-	return (((WordEntryIN *) a)->entry.len > ((WordEntryIN *) b)->entry.len) ? 1 : -1;
+	return (a->entry.len > b->entry.len) ? 1 : -1;
 }
 
 static int
@@ -104,6 +106,9 @@ uniqueentry(WordEntryIN * a, int l, char *buf, int *outbuflen)
 			a->poslen = uniquePos(a->pos, a->poslen);
 			*outbuflen = SHORTALIGN(a->entry.len) + (a->poslen + 1) * sizeof(WordEntryPos);
 		}
+		else
+			*outbuflen = a->entry.len;
+
 		return l;
 	}
 	res = a;
@@ -118,10 +123,12 @@ uniqueentry(WordEntryIN * a, int l, char *buf, int *outbuflen)
 		{
 			if (res->entry.haspos)
 			{
+				*outbuflen += SHORTALIGN(res->entry.len);
 				res->poslen = uniquePos(res->pos, res->poslen);
 				*outbuflen += res->poslen * sizeof(WordEntryPos);
 			}
-			*outbuflen += SHORTALIGN(res->entry.len);
+			else
+				*outbuflen += res->entry.len;
 			res++;
 			memcpy(res, ptr, sizeof(WordEntryIN));
 		}
@@ -147,12 +154,18 @@ uniqueentry(WordEntryIN * a, int l, char *buf, int *outbuflen)
 		}
 		ptr++;
 	}
+
+	/* add last item */
+
 	if (res->entry.haspos)
 	{
+		*outbuflen += SHORTALIGN(res->entry.len);
+
 		res->poslen = uniquePos(res->pos, res->poslen);
 		*outbuflen += res->poslen * sizeof(WordEntryPos);
 	}
-	*outbuflen += SHORTALIGN(res->entry.len);
+	else
+		*outbuflen += res->entry.len;
 
 	return res + 1 - a;
 }
@@ -367,6 +380,18 @@ tsvectorout(PG_FUNCTION_ARGS)
 	PG_RETURN_CSTRING(outbuf);
 }
 
+/*
+ * Binary Input / Output functions. The binary format is as follows:
+ *
+ * uint32	number of lexemes
+ * 
+ * for each lexeme:
+ *		lexeme text in client encoding, null-terminated
+ * 		uint16	number of positions
+ * 		for each position:
+ *			uint16 WordEntryPos
+ */
+
 Datum
 tsvectorsend(PG_FUNCTION_ARGS)
 {
@@ -381,18 +406,22 @@ tsvectorsend(PG_FUNCTION_ARGS)
 	pq_sendint(&buf, vec->size, sizeof(int32));
 	for (i = 0; i < vec->size; i++)
 	{
-		/*
-		 * We are sure that sizeof(WordEntry) == sizeof(int32)
+		uint16 npos;
+
+		/* the strings in the TSVector array are not null-terminated, so 
+		 * we have to send the null-terminator separately
 		 */
-		pq_sendint(&buf, *(int32 *) weptr, sizeof(int32));
+		pq_sendtext(&buf, STRPTR(vec) + weptr->pos, weptr->len);
+		pq_sendbyte(&buf, '\0');
 
-		pq_sendbytes(&buf, STRPTR(vec) + weptr->pos, weptr->len);
-		if (weptr->haspos)
+		npos = POSDATALEN(vec, weptr);
+		pq_sendint(&buf, npos, sizeof(uint16));
+
+		if(npos > 0)
 		{
 			WordEntryPos *wepptr = POSDATAPTR(vec, weptr);
 
-			pq_sendint(&buf, POSDATALEN(vec, weptr), sizeof(WordEntryPos));
-			for (j = 0; j < POSDATALEN(vec, weptr); j++)
+			for (j = 0; j < npos; j++)
 				pq_sendint(&buf, wepptr[j], sizeof(WordEntryPos));
 		}
 		weptr++;
@@ -407,71 +436,92 @@ tsvectorrecv(PG_FUNCTION_ARGS)
 	StringInfo	buf = (StringInfo) PG_GETARG_POINTER(0);
 	TSVector	vec;
 	int			i;
-	uint32		size;
-	WordEntry  *weptr;
-	int			datalen = 0;
-	Size		len;
-
-	size = pq_getmsgint(buf, sizeof(uint32));
-	if (size < 0 || size > (MaxAllocSize / sizeof(WordEntry)))
+	int32		nentries;
+	int			datalen; /* number of bytes used in the variable size area
+						  * after fixed size TSVector header and WordEntries
+						  */
+	Size		hdrlen;
+	Size		len;  /* allocated size of vec */
+
+	nentries = pq_getmsgint(buf, sizeof(int32));
+	if (nentries < 0 || nentries > (MaxAllocSize / sizeof(WordEntry)))
 		elog(ERROR, "invalid size of tsvector");
 
-	len = DATAHDRSIZE + sizeof(WordEntry) * size;
+	hdrlen = DATAHDRSIZE + sizeof(WordEntry) * nentries;
 
-	len = len * 2; /* times two to make room for lexemes */
+	len = hdrlen * 2; /* times two to make room for lexemes */
 	vec = (TSVector) palloc0(len);
-	vec->size = size;
+	vec->size = nentries;
 
-	weptr = ARRPTR(vec);
-	for (i = 0; i < size; i++)
+	datalen = 0;
+	for (i = 0; i < nentries; i++)
 	{
-		int32 tmp;
+		const char *lexeme;
+		uint16 npos;
+		size_t lex_len;
+
+		lexeme = pq_getmsgstring(buf);
+		npos = (uint16) pq_getmsgint(buf, sizeof(uint16));
+
+		/* sanity checks */
+
+		lex_len = strlen(lexeme);
+		if (lex_len < 0 || lex_len > MAXSTRLEN)
+			elog(ERROR, "invalid tsvector; lexeme too long");
+
+		if (datalen > MAXSTRPOS)
+			elog(ERROR, "invalid tsvector; maximum total lexeme length exceeded"); 
 
-		weptr = ARRPTR(vec) + i;
+		if (npos > MAXNUMPOS)
+			elog(ERROR, "unexpected number of positions");
 
 		/*
-		 * We are sure that sizeof(WordEntry) == sizeof(int32)
+		 * Looks valid. Fill the WordEntry struct, and copy lexeme.
+		 *
+		 * But make sure the buffer is large enough first.
 		 */
-		tmp = pq_getmsgint(buf, sizeof(int32));
-		*weptr = *(WordEntry *) & tmp;
-
-		while (CALCDATASIZE(size, datalen + SHORTALIGN(weptr->len)) >= len)
+		while (hdrlen + SHORTALIGN(datalen + lex_len) +
+			   (npos + 1) * sizeof(WordEntryPos) >= len)
 		{
 			len *= 2;
 			vec = (TSVector) repalloc(vec, len);
-			weptr = ARRPTR(vec) + i;
 		}
 
-		memcpy(STRPTR(vec) + weptr->pos,
-			   pq_getmsgbytes(buf, weptr->len),
-			   weptr->len);
-		datalen += SHORTALIGN(weptr->len);
+		vec->entries[i].haspos = (npos > 0) ? 1 : 0;
+		vec->entries[i].len = lex_len;
+		vec->entries[i].pos = datalen;
+
+		memcpy(STRPTR(vec) + datalen, lexeme, lex_len);
+
+		datalen += lex_len;
 
-		if (i > 0 && WordEntryCMP(weptr, weptr - 1, STRPTR(vec)) <= 0)
+		if (i > 0 && WordEntryCMP(&vec->entries[i], &vec->entries[i - 1], STRPTR(vec)) <= 0)
 			elog(ERROR, "lexemes are unordered");
 
-		if (weptr->haspos)
+		/* Receive positions */
+
+		if (npos > 0)
 		{
-			uint16		j,
-						npos;
+			uint16		j;
 			WordEntryPos *wepptr;
 
-			npos = (uint16) pq_getmsgint(buf, sizeof(uint16));
-			if (npos > MAXNUMPOS)
-				elog(ERROR, "unexpected number of positions");
-
-			while (CALCDATASIZE(size, datalen + (npos + 1) * sizeof(WordEntryPos)) >= len)
+			/*
+			 * Pad to 2-byte alignment if necessary. Though we used palloc0
+			 * for the initial allocation, subsequent repalloc'd memory
+			 * areas are not initialized to zero.
+			 */
+			if (datalen != SHORTALIGN(datalen))
 			{
-				len *= 2;
-				vec = (TSVector) repalloc(vec, len);
-				weptr = ARRPTR(vec) + i;
+				*(STRPTR(vec) + datalen) = '\0';
+				datalen = SHORTALIGN(datalen);
 			}
 
-			memcpy(_POSDATAPTR(vec, weptr), &npos, sizeof(int16));
-			wepptr = POSDATAPTR(vec, weptr);
+			memcpy(STRPTR(vec) + datalen, &npos, sizeof(uint16));
+
+			wepptr = POSDATAPTR(vec, &vec->entries[i]);
 			for (j = 0; j < npos; j++)
 			{
-				wepptr[j] = (WordEntryPos) pq_getmsgint(buf, sizeof(int16));
+				wepptr[j] = (WordEntryPos) pq_getmsgint(buf, sizeof(WordEntryPos));
 				if (j > 0 && WEP_GETPOS(wepptr[j]) <= WEP_GETPOS(wepptr[j - 1]))
 					elog(ERROR, "position information is unordered");
 			}
@@ -480,7 +530,7 @@ tsvectorrecv(PG_FUNCTION_ARGS)
 		}
 	}
 
-	SET_VARSIZE(vec, CALCDATASIZE(vec->size, datalen));
+	SET_VARSIZE(vec, hdrlen + datalen);
 
 	PG_RETURN_TSVECTOR(vec);
 }

src/backend/utils/adt/tsvector_op.c

@@ -7,7 +7,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector_op.c,v 1.3 2007/09/07 15:09:56 teodor Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/tsvector_op.c,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -165,7 +165,7 @@ tsvector_strip(PG_FUNCTION_ARGS)
 	char	   *cur;
 
 	for (i = 0; i < in->size; i++)
-		len += SHORTALIGN(arrin[i].len);
+		len += arrin[i].len;
 
 	len = CALCDATASIZE(in->size, len);
 	out = (TSVector) palloc0(len);
@@ -179,7 +179,7 @@ tsvector_strip(PG_FUNCTION_ARGS)
 		arrout[i].haspos = 0;
 		arrout[i].len = arrin[i].len;
 		arrout[i].pos = cur - STRPTR(out);
-		cur += SHORTALIGN(arrout[i].len);
+		cur += arrout[i].len;
 	}
 
 	PG_FREE_IF_COPY(in, 0);
@@ -351,12 +351,15 @@ tsvector_concat(PG_FUNCTION_ARGS)
 			ptr->len = ptr1->len;
 			memcpy(cur, data1 + ptr1->pos, ptr1->len);
 			ptr->pos = cur - data;
-			cur += SHORTALIGN(ptr1->len);
 			if (ptr->haspos)
 			{
+				cur += SHORTALIGN(ptr1->len);
 				memcpy(cur, _POSDATAPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
 				cur += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
 			}
+			else
+				cur += ptr1->len;
+				
 			ptr++;
 			ptr1++;
 			i1--;
@@ -367,16 +370,20 @@ tsvector_concat(PG_FUNCTION_ARGS)
 			ptr->len = ptr2->len;
 			memcpy(cur, data2 + ptr2->pos, ptr2->len);
 			ptr->pos = cur - data;
-			cur += SHORTALIGN(ptr2->len);
 			if (ptr->haspos)
 			{
 				int			addlen = add_pos(in2, ptr2, out, ptr, maxpos);
 
+				cur += SHORTALIGN(ptr2->len);
+
 				if (addlen == 0)
 					ptr->haspos = 0;
 				else
 					cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
 			}
+			else
+				cur += ptr2->len;
+
 			ptr++;
 			ptr2++;
 			i2--;
@@ -387,9 +394,9 @@ tsvector_concat(PG_FUNCTION_ARGS)
 			ptr->len = ptr1->len;
 			memcpy(cur, data1 + ptr1->pos, ptr1->len);
 			ptr->pos = cur - data;
-			cur += SHORTALIGN(ptr1->len);
 			if (ptr->haspos)
 			{
+				cur += SHORTALIGN(ptr1->len);
 				if (ptr1->haspos)
 				{
 					memcpy(cur, _POSDATAPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
@@ -407,6 +414,9 @@ tsvector_concat(PG_FUNCTION_ARGS)
 						cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
 				}
 			}
+			else
+				cur += ptr1->len;
+
 			ptr++;
 			ptr1++;
 			ptr2++;
@@ -421,12 +431,15 @@ tsvector_concat(PG_FUNCTION_ARGS)
 		ptr->len = ptr1->len;
 		memcpy(cur, data1 + ptr1->pos, ptr1->len);
 		ptr->pos = cur - data;
-		cur += SHORTALIGN(ptr1->len);
 		if (ptr->haspos)
 		{
+			cur += SHORTALIGN(ptr1->len);
 			memcpy(cur, _POSDATAPTR(in1, ptr1), POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16));
 			cur += POSDATALEN(in1, ptr1) * sizeof(WordEntryPos) + sizeof(uint16);
 		}
+		else
+			cur += ptr1->len;
+
 		ptr++;
 		ptr1++;
 		i1--;
@@ -438,16 +451,20 @@ tsvector_concat(PG_FUNCTION_ARGS)
 		ptr->len = ptr2->len;
 		memcpy(cur, data2 + ptr2->pos, ptr2->len);
 		ptr->pos = cur - data;
-		cur += SHORTALIGN(ptr2->len);
 		if (ptr->haspos)
 		{
 			int			addlen = add_pos(in2, ptr2, out, ptr, maxpos);
 
+			cur += SHORTALIGN(ptr2->len);
+
 			if (addlen == 0)
 				ptr->haspos = 0;
 			else
 				cur += addlen * sizeof(WordEntryPos) + sizeof(uint16);
 		}
+		else
+			cur += ptr2->len;
+
 		ptr++;
 		ptr2++;
 		i2--;
@@ -484,8 +501,8 @@ ValCompare(CHKVAL * chkval, WordEntry * ptr, QueryOperand * item)
 static bool
 checkclass_str(CHKVAL * chkval, WordEntry * val, QueryOperand * item)
 {
-	WordEntryPos *ptr = (WordEntryPos *) (chkval->values + val->pos + SHORTALIGN(val->len) + sizeof(uint16));
-	uint16		len = *((uint16 *) (chkval->values + val->pos + SHORTALIGN(val->len)));
+	WordEntryPos *ptr = (WordEntryPos *) (chkval->values + SHORTALIGN(val->pos + val->len) + sizeof(uint16));
+	uint16		len = *((uint16 *) (chkval->values + SHORTALIGN(val->pos + val->len)));
 
 	while (len--)
 	{

src/include/tsearch/ts_type.h

@@ -5,7 +5,7 @@
  *
  * Copyright (c) 1998-2007, PostgreSQL Global Development Group
  *
- * $PostgreSQL: pgsql/src/include/tsearch/ts_type.h,v 1.3 2007/09/07 15:35:11 teodor Exp $
+ * $PostgreSQL: pgsql/src/include/tsearch/ts_type.h,v 1.4 2007/09/07 16:03:40 teodor Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -62,26 +62,33 @@ typedef uint16 WordEntryPos;
  *							bytes from end of WordEntry array to start of
  *							corresponding lexeme.
  * 4) Lexeme's storage:
- *	  SHORTALIGNED(lexeme) and position information if it exists
- *	  Position information: first int2 - is a number of positions and it
- *	  follows array of WordEntryPos
+ *	  lexeme (without null-terminator)
+ *    if haspos is true:
+ *		padding byte if necessary to make the number of positions 2-byte aligned
+ *		uint16		number of positions that follow.
+ *		uint16[]	positions
+ *
+ * The positions must be sorted.
  */
 
 typedef struct
 {
 	int32		vl_len_;		/* varlena header (do not touch directly!) */
-	uint32		size;
-	char		data[1];
+	int32		size;
+	WordEntry	entries[1]; /* var size */
+	/* lexemes follow */
 } TSVectorData;
 
 typedef TSVectorData *TSVector;
 
-#define DATAHDRSIZE (VARHDRSZ + sizeof(int4))
-#define CALCDATASIZE(x, lenstr) ( (x) * sizeof(WordEntry) + DATAHDRSIZE + (lenstr) )
-#define ARRPTR(x)	( (WordEntry*) ( (char*)(x) + DATAHDRSIZE ) )
-#define STRPTR(x)	( (char*)(x) + DATAHDRSIZE + ( sizeof(WordEntry) * ((TSVector)(x))->size ) )
-#define STRSIZE(x)	( ((TSVector)(x))->len - DATAHDRSIZE - ( sizeof(WordEntry) * ((TSVector)(x))->size ) )
-#define _POSDATAPTR(x,e)	(STRPTR(x)+((WordEntry*)(e))->pos+SHORTALIGN(((WordEntry*)(e))->len))
+#define DATAHDRSIZE (offsetof(TSVectorData, entries))
+#define CALCDATASIZE(x, lenstr) (DATAHDRSIZE + (x) * sizeof(WordEntry) + (lenstr) )
+#define ARRPTR(x)	( (x)->entries )
+
+/* returns a pointer to the beginning of lexemes */
+#define STRPTR(x)	( (char *) &(x)->entries[x->size] )
+
+#define _POSDATAPTR(x,e)	(STRPTR(x) + SHORTALIGN((e)->pos + (e)->len))
 #define POSDATALEN(x,e) ( ( ((WordEntry*)(e))->haspos ) ? (*(uint16*)_POSDATAPTR(x,e)) : 0 )
 #define POSDATAPTR(x,e) ( (WordEntryPos*)( _POSDATAPTR(x,e)+sizeof(uint16) ) )
 
@@ -159,7 +166,7 @@ typedef int8 QueryItemType;
 typedef struct
 {
 	QueryItemType		type;	/* operand or kind of operator (ts_tokentype) */
-	int8		weight;			/* weights of operand to search. It's a bitmask of allowed weights.
+	uint8		weight;			/* weights of operand to search. It's a bitmask of allowed weights.
 								 * if it =0 then any weight are allowed.
 								 * Weights and bit map:
 								 * A: 1<<3