Avoid O(N^2) behavior when enlarging SPI tuple table in spi_printtup().

tglsfdc · tglsfdc · commit fe939d950e36 · 2015-08-21T20:32:26.000-04:00
For no obvious reason, spi_printtup() was coded to enlarge the tuple
pointer table by just 256 slots at a time, rather than doubling the size at
each reallocation, as is our usual habit.  For very large SPI results, this
makes for O(N^2) time spent in repalloc(), which of course soon comes to
dominate the runtime.  Use the standard doubling approach instead.

This is a longstanding performance bug, so back-patch to all active
branches.

Neil Conway
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
@@ -1769,7 +1769,8 @@ spi_printtup(TupleTableSlot *slot, DestReceiver *self)
 
 	if (tuptable->free == 0)
 	{
-		tuptable->free = 256;
+		/* Double the size of the pointer array */
+		tuptable->free = tuptable->alloced;
 		tuptable->alloced += tuptable->free;
 		tuptable->vals = (HeapTuple *) repalloc(tuptable->vals,
 									  tuptable->alloced * sizeof(HeapTuple));

Original file line number	Diff line number	Diff line change
`@@ -1769,7 +1769,8 @@ spi_printtup(TupleTableSlot slot, DestReceiver self)`
`1769`	`1769`
`1770`	`1770`	`if (tuptable->free == 0)`
`1771`	`1771`	`{`
`1772`		`- tuptable->free = 256;`
	`1772`	`+ /* Double the size of the pointer array */`
	`1773`	`+ tuptable->free = tuptable->alloced;`
`1773`	`1774`	`tuptable->alloced += tuptable->free;`
`1774`	`1775`	`tuptable->vals = (HeapTuple *) repalloc(tuptable->vals,`
`1775`	`1776`	`tuptable->alloced * sizeof(HeapTuple));`