Re: [HACKERS] GSoC 2017: weekly progress reports (week 6) - Mailing list pgsql-hackers

On Sat, Sep 30, 2017 at 6:12 PM, Shubham Barai <shubhambaraiss@gmail.com> wrote:

I have made changes according to your suggestions. Please have a look at the updated patch. 

I am also considering your suggestions for my other patches also. But, I will need some time to 

make changes as I am currently busy doing my master's.

I don't understand why sometimes you call PredicateLockPage() only when fast update is off.  For example:

@@ -94,6 +101,9 @@ scanPostingTree(Relation index, GinScanEntry scanEntry,
  break; /* no more pages */
 
  buffer = ginStepRight(buffer, index, GIN_SHARE);
+
+ if (!GinGetUseFastUpdate(index))
+ PredicateLockPage(index, BufferGetBlockNumber(buffer), snapshot);
  }
 
  UnlockReleaseBuffer(buffer);

But sometimes you call PredicateLockPage() unconditionally.

@@ -131,6 +141,8 @@ collectMatchBitmap(GinBtreeData *btree, GinBtreeStack *stack,
  attnum = scanEntry->attnum;
  attr = btree->ginstate->origTupdesc->attrs[attnum - 1];
 
+ PredicateLockPage(btree->index, stack->buffer, snapshot);
+
  for (;;)
  {
  Page page;

As I understand, all page-level predicate locking should happen only when fast update is off.

Also, despite general idea is described in README-SSI, in-code comments would be useful.