Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Skip to content

Commit c86c2d9

Browse files
committed
Update comment.
mergepreread()/mergeprereadone() don't exist anymore, the function that does roughly the same is now called mergereadnext().
1 parent 61633f7 commit c86c2d9

File tree

1 file changed

+8
-8
lines changed

1 file changed

+8
-8
lines changed

src/backend/utils/sort/tuplesort.c

+8-8
Original file line numberDiff line numberDiff line change
@@ -3083,14 +3083,14 @@ dumpbatch(Tuplesortstate *state, bool alltuples)
30833083
* a call with no subsequent run actually written to destTape), we prefer
30843084
* to write out a 0 tuple run.
30853085
*
3086-
* mergepreread()/mergeprereadone() are prepared for 0 tuple runs, and
3087-
* will reliably mark the tape inactive for the merge when called from
3088-
* beginmerge(). This case is therefore similar to the case where
3089-
* mergeonerun() finds a dummy run for the tape, and so doesn't need to
3090-
* merge a run from the tape (or conceptually "merges" the dummy run, if
3091-
* you prefer). According to Knuth, Algorithm D "isn't strictly optimal"
3092-
* in its method of distribution and dummy run assignment; this edge case
3093-
* seems very unlikely to make that appreciably worse.
3086+
* mergereadnext() is prepared for 0 tuple runs, and will reliably mark
3087+
* the tape inactive for the merge when called from beginmerge(). This
3088+
* case is therefore similar to the case where mergeonerun() finds a dummy
3089+
* run for the tape, and so doesn't need to merge a run from the tape (or
3090+
* conceptually "merges" the dummy run, if you prefer). According to
3091+
* Knuth, Algorithm D "isn't strictly optimal" in its method of
3092+
* distribution and dummy run assignment; this edge case seems very
3093+
* unlikely to make that appreciably worse.
30943094
*/
30953095
Assert(state->status == TSS_BUILDRUNS);
30963096

0 commit comments

Comments
 (0)