Clean up get_pgsm_query_id_hash()

The function was unnecessarily convoluted so clean it up by making the
parser easier to read, asserting against NULL input and wrapping the
comment at the right width. Especially the code for handling whitespace
was much more complicated than necessary.

Also add another test example to check that we merge whitespace even
when split by a comment.

Author: jeltz

regression/expected/pgsm_query_id.out

@@ -35,6 +35,11 @@ SELECT *, add(1, 2) FROM t1;
 ---+-----
 (0 rows)
 
+SELECT * /* x */ FROM t2;
+ b 
+---
+(0 rows)
+
 SELECT * FROM t2;
  b 
 ---
@@ -85,7 +90,7 @@ SELECT datname, pgsm_query_id, query, calls FROM pg_stat_monitor ORDER BY pgsm_q
 --------------------+---------------------+-----------------------------------------------------+-------
  contrib_regression |  689150021118383254 | SELECT pg_stat_monitor_reset()                      |     1
  db2                | 1648576050146230147 | SELECT * FROM t3 WHERE c = 20                       |     1
- db1                | 1897482803466821995 | SELECT * FROM t2                                    |     3
+ db1                | 1897482803466821995 | SELECT * /* x */ FROM t2                            |     4
  db1                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db2                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db1                | 5193804135051352284 | SELECT $1 + $2                                      |     1

regression/expected/pgsm_query_id_1.out

@@ -35,6 +35,11 @@ SELECT *, add(1, 2) FROM t1;
 ---+-----
 (0 rows)
 
+SELECT * /* x */ FROM t2;
+ b 
+---
+(0 rows)
+
 SELECT * FROM t2;
  b 
 ---
@@ -85,7 +90,7 @@ SELECT datname, pgsm_query_id, query, calls FROM pg_stat_monitor ORDER BY pgsm_q
 --------------------+---------------------+-----------------------------------------------------+-------
  contrib_regression |  689150021118383254 | SELECT pg_stat_monitor_reset()                      |     1
  db2                | 1648576050146230147 | SELECT * FROM t3 WHERE c = 20                       |     1
- db1                | 1897482803466821995 | SELECT * FROM t2                                    |     3
+ db1                | 1897482803466821995 | SELECT * /* x */ FROM t2                            |     4
  db1                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db2                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db2                | 6220142855706866455 | set pg_stat_monitor.pgsm_enable_pgsm_query_id = on  |     1

regression/expected/pgsm_query_id_2.out

@@ -35,6 +35,11 @@ SELECT *, add(1, 2) FROM t1;
 ---+-----
 (0 rows)
 
+SELECT * /* x */ FROM t2;
+ b 
+---
+(0 rows)
+
 SELECT * FROM t2;
  b 
 ---
@@ -85,7 +90,7 @@ SELECT datname, pgsm_query_id, query, calls FROM pg_stat_monitor ORDER BY pgsm_q
 --------------------+---------------------+-----------------------------------------------------+-------
  contrib_regression |  689150021118383254 | SELECT pg_stat_monitor_reset()                      |     1
  db2                | 1648576050146230147 | SELECT * FROM t3 WHERE c = 20                       |     1
- db1                | 1897482803466821995 | SELECT * FROM t2                                    |     3
+ db1                | 1897482803466821995 | SELECT * /* x */ FROM t2                            |     4
  db1                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db2                | 1988437669671417938 | SELECT * FROM t1                                    |     1
  db1                | 8828613017640075704 | SELECT *, add(1, 2) FROM t1                         |     1

regression/sql/pgsm_query_id.sql

@@ -28,6 +28,7 @@ SELECT pg_stat_monitor_reset();
 \c db1
 SELECT * FROM t1;
 SELECT *, add(1, 2) FROM t1;
+SELECT * /* x */ FROM t2;
 SELECT * FROM t2;
 -- Check that spaces and comments do not generate a different pgsm_query_id
 SELECT     *     FROM t2 --WHATEVER;

src/pg_stat_monitor.c

@@ -2582,11 +2582,10 @@ get_next_wbucket(pgsmSharedState *pgsm)
 }
 
 /*
- * This function expects a NORMALIZED query as the input.
- * It iterates over the normalized query skipping comments and
- * multiple spaces. All spaces are converted to ' ' so that we
- * the calculation is independent of the space type whether
- * newline, tab, or any other type. Trailing and leading spaces
+ * This function expects a NORMALIZED query as the input. It iterates over the
+ * normalized query skipping comments and multiple spaces. All spaces are
+ * converted to ' ' so that we the calculation is independent of the space
+ * type whether newline, tab, or any other type. Trailing and leading spaces
  * are also removed before calculating the hash.
  */
 static int64
@@ -2595,82 +2594,71 @@ get_pgsm_query_id_hash(const char *norm_query, int norm_len)
 	char	   *query;
 	char	   *q_iter;
 	const char *norm_q_iter = norm_query;
-	int64		pgsm_query_id = 0;
+	bool		space = false;
+	int64		pgsm_query_id;
+
+	Assert(norm_query != NULL);
 
 	if (!pgsm_enable_pgsm_query_id)
 		return 0;
 
 	query = palloc(norm_len + 1);
 	q_iter = query;
 
-	while (norm_q_iter && *norm_q_iter && norm_q_iter < norm_query + norm_len)
+	while (*norm_q_iter && norm_q_iter < norm_query + norm_len)
 	{
-		/*
-		 * Skip multiline comments, + 1 is safe even if we've reach end of
-		 * string
-		 */
 		if (*norm_q_iter == '/' && *(norm_q_iter + 1) == '*')
 		{
-			while (*norm_q_iter && *(norm_q_iter + 1) && (*norm_q_iter != '*' || *(norm_q_iter + 1) != '/'))
-				norm_q_iter++;
-
-			/*
-			 * Skip the end if the current character is valid. norm_q_iter
-			 * points to the *, we have to skip 2 characters
-			 */
-			if (*norm_q_iter)
-				norm_q_iter++;
-			if (*norm_q_iter)
-				norm_q_iter++;
+			/* Skip multiline comments */
+			norm_q_iter++;
+			norm_q_iter++;
 
-			continue;
+			while (*norm_q_iter)
+			{
+				if (*norm_q_iter == '*' && *(norm_q_iter + 1) == '/')
+				{
+					norm_q_iter++;
+					norm_q_iter++;
+					break;
+				}
+				else
+					norm_q_iter++;
+			}
 		}
-
-		/*
-		 * Skip single line comments, + 1 is safe even if we've reach end of
-		 * string
-		 */
-		if (*norm_q_iter == '-' && *(norm_q_iter + 1) == '-')
+		else if (*norm_q_iter == '-' && *(norm_q_iter + 1) == '-')
 		{
+			/* Skip single line comments */
+			norm_q_iter++;
+			norm_q_iter++;
+
 			while (*norm_q_iter && *norm_q_iter != '\n')
 				norm_q_iter++;
 		}
-
-		/* Skip white spaces */
-		if (scanner_isspace(*norm_q_iter))
+		else if (scanner_isspace(*norm_q_iter))
 		{
-			while (scanner_isspace(*++norm_q_iter));
+			/* Convert white spaces into single space */
+			norm_q_iter++;
 
-			/*
-			 * Let's replace it with a simple space. -1 is safe as we are
-			 * making sure we are not at the start of the string.
-			 */
-			if (q_iter != query && !scanner_isspace(*(q_iter - 1)))
-				*q_iter++ = ' ';
+			while (scanner_isspace(*norm_q_iter))
+				norm_q_iter++;
 
-			continue;
+			space = true;
+		}
+		else
+		{
+			if (space && q_iter != query)
+				*q_iter++ = ' ';
+			space = false;
+			*q_iter++ = *norm_q_iter++;
 		}
-
-		*q_iter++ = *norm_q_iter++;
 	}
 
-	/* Ensure we have a terminating zero at the end */
 	*q_iter = '\0';
 
-	/* Get rid of trailing spaces */
-	while (q_iter > query && *q_iter == '\0')
-	{
-		q_iter--;
-
-		/* Continue reducing the string size if space is found. */
-		if (scanner_isspace(*q_iter))
-			*q_iter = '\0';
-	}
-
-	/* Calculate the hash. */
-	pgsm_query_id = pgsm_hash_string(query, strlen(query));
+	pgsm_query_id = pgsm_hash_string(query, q_iter - query);
 
 	pfree(query);
+
 	return pgsm_query_id;
 }