Implement Incremental Sort

Incremental Sort is an optimized variant of multikey sort for cases when
the input is already sorted by a prefix of the requested sort keys. For
example when the relation is already sorted by (key1, key2) and we need
to sort it by (key1, key2, key3) we can simply split the input rows into
groups having equal values in (key1, key2), and only sort/compare the
remaining column key3.

This has a number of benefits:

- Reduced memory consumption, because only a single group (determined by
  values in the sorted prefix) needs to be kept in memory. This may also
  eliminate the need to spill to disk.

- Lower startup cost, because Incremental Sort produce results after each
  prefix group, which is beneficial for plans where startup cost matters
  (like for example queries with LIMIT clause).

We consider both Sort and Incremental Sort, and decide based on costing.

The implemented algorithm operates in two different modes:

- Fetching a minimum number of tuples without check of equality on the
  prefix keys, and sorting on all columns when safe.

- Fetching all tuples for a single prefix group and then sorting by
  comparing only the remaining (non-prefix) keys.

We always start in the first mode, and employ a heuristic to switch into
the second mode if we believe it's beneficial - the goal is to minimize
the number of unnecessary comparions while keeping memory consumption
below work_mem.

This is a very old patch series. The idea was originally proposed by
Alexander Korotkov back in 2013, and then revived in 2017. In 2018 the
patch was taken over by James Coleman, who wrote and rewrote most of the
current code.

There were many reviewers/contributors since 2013 - I've done my best to
pick the most active ones, and listed them in this commit message.

Author: James Coleman, Alexander Korotkov
Reviewed-by: Tomas Vondra, Andreas Karlsson, Marti Raudsepp, Peter Geoghegan, Robert Haas, Thomas Munro, Antonin Houska, Andres Freund, Alexander Kuzmenkov
Discussion: https://postgr.es/m/CAPpHfdscOX5an71nHd8WSUH6GNOCf=V7wgDaTXdDd9=goN-gfA@mail.gmail.com
Discussion: https://postgr.es/m/CAPpHfds1waRZ=NOmueYq0sx1ZSCnt+5QJvizT8ndT2=etZEeAQ@mail.gmail.com

1 files changed, 63 insertions, 9 deletions

diff --git a/src/backend/optimizer/path/pathkeys.c b/src/backend/optimizer/path/pathkeys.c
index 71b9d42c99..21e3f5a987 100644
--- a/src/backend/optimizer/path/pathkeys.c
+++ b/src/backend/optimizer/path/pathkeys.c
@@ -335,6 +335,60 @@ pathkeys_contained_in(List *keys1, List *keys2)
 }
 
 /*
+ * pathkeys_count_contained_in
+ *    Same as pathkeys_contained_in, but also sets length of longest
+ *    common prefix of keys1 and keys2.
+ */
+bool
+pathkeys_count_contained_in(List *keys1, List *keys2, int *n_common)
+{
+	int			n = 0;
+	ListCell   *key1,
+			   *key2;
+
+	/*
+	 * See if we can avoiding looping through both lists. This optimization
+	 * gains us several percent in planning time in a worst-case test.
+	 */
+	if (keys1 == keys2)
+	{
+		*n_common = list_length(keys1);
+		return true;
+	}
+	else if (keys1 == NIL)
+	{
+		*n_common = 0;
+		return true;
+	}
+	else if (keys2 == NIL)
+	{
+		*n_common = 0;
+		return false;
+	}
+
+	/*
+	 * If both lists are non-empty, iterate through both to find out how many
+	 * items are shared.
+	 */
+	forboth(key1, keys1, key2, keys2)
+	{
+		PathKey    *pathkey1 = (PathKey *) lfirst(key1);
+		PathKey    *pathkey2 = (PathKey *) lfirst(key2);
+
+		if (pathkey1 != pathkey2)
+		{
+			*n_common = n;
+			return false;
+		}
+		n++;
+	}
+
+	/* If we ended with a null value, then we've processed the whole list. */
+	*n_common = n;
+	return (key1 == NULL);
+}
+
+/*
  * get_cheapest_path_for_pathkeys
  *	  Find the cheapest path (according to the specified criterion) that
  *	  satisfies the given pathkeys and parameterization.
@@ -1786,26 +1840,26 @@ right_merge_direction(PlannerInfo *root, PathKey *pathkey)
  *		Count the number of pathkeys that are useful for meeting the
  *		query's requested output ordering.
  *
- * Unlike merge pathkeys, this is an all-or-nothing affair: it does us
- * no good to order by just the first key(s) of the requested ordering.
- * So the result is always either 0 or list_length(root->query_pathkeys).
+ * Because we the have the possibility of incremental sort, a prefix list of
+ * keys is potentially useful for improving the performance of the requested
+ * ordering. Thus we return 0, if no valuable keys are found, or the number
+ * of leading keys shared by the list and the requested ordering..
  */
 static int
 pathkeys_useful_for_ordering(PlannerInfo *root, List *pathkeys)
 {
+	int			n_common_pathkeys;
+
 	if (root->query_pathkeys == NIL)
 		return 0;				/* no special ordering requested */
 
 	if (pathkeys == NIL)
 		return 0;				/* unordered path */
 
-	if (pathkeys_contained_in(root->query_pathkeys, pathkeys))
-	{
-		/* It's useful ... or at least the first N keys are */
-		return list_length(root->query_pathkeys);
-	}
+	(void) pathkeys_count_contained_in(root->query_pathkeys, pathkeys,
+										&n_common_pathkeys);
 
-	return 0;					/* path ordering not useful */
+	return n_common_pathkeys;
 }
 
 /*