Function array_to_tsvector

src/backend/utils/adt/tsvector_op.c:748–807

* Build tsvector from array of lexemes. */

 * Build tsvector from array of lexemes.
 */
Datum
array_to_tsvector(PG_FUNCTION_ARGS)
{
	ArrayType  *v = PG_GETARG_ARRAYTYPE_P(0);
	TSVector	tsout;
	Datum	   *dlexemes;
	WordEntry  *arrout;
	bool	   *nulls;
	int			nitems,
				i,
				tslen,
				datalen = 0;
	char	   *cur;

	deconstruct_array(v, TEXTOID, -1, false, TYPALIGN_INT, &dlexemes, &nulls, &nitems);

	/* Reject nulls (maybe we should just ignore them, instead?) */
	for (i = 0; i < nitems; i++)
	{
		if (nulls[i])
			ereport(ERROR,
					(errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
					 errmsg("lexeme array may not contain nulls")));
	}

	/* Sort and de-dup, because this is required for a valid tsvector. */
	if (nitems > 1)
	{
		qsort(dlexemes, nitems, sizeof(Datum), compare_text_lexemes);
		nitems = qunique(dlexemes, nitems, sizeof(Datum),
						 compare_text_lexemes);
	}

	/* Calculate space needed for surviving lexemes. */
	for (i = 0; i < nitems; i++)
		datalen += VARSIZE(dlexemes[i]) - VARHDRSZ;
	tslen = CALCDATASIZE(nitems, datalen);

	/* Allocate and fill tsvector. */
	tsout = (TSVector) palloc0(tslen);
	SET_VARSIZE(tsout, tslen);
	tsout->size = nitems;

	arrout = ARRPTR(tsout);
	cur = STRPTR(tsout);
	for (i = 0; i < nitems; i++)
	{
		char	   *lex = VARDATA(dlexemes[i]);
		int			lex_len = VARSIZE(dlexemes[i]) - VARHDRSZ;

		memcpy(cur, lex, lex_len);
		arrout[i].haspos = 0;
		arrout[i].len = lex_len;
		arrout[i].pos = cur - STRPTR(tsout);
		cur += lex_len;
	}

	PG_FREE_IF_COPY(v, 0);