|From:||Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com>|
|To:||Alexander Korotkov <a(dot)korotkov(at)postgrespro(dot)ru>|
|Subject:||Re: Page Scan Mode in Hash Index|
|Views:||Raw Message | Whole Thread | Download mbox|
>> I've assigned to review this patch.
>> At first, I'd like to notice that I like idea and general design.
>> Secondly, patch set don't apply cleanly to master. Please, rebase it.
> Thanks for showing your interest towards this patch. I would like to
> inform that this patch has got dependency on patch for 'Write Ahead
> Logging in hash index - ' and 'Microvacuum support in hash index
> '. Hence, until above two patches becomes stable I may have to keep
> on rebasing this patch. However, I will try to share you the updated
> patch asap.
>> On Tue, Feb 14, 2017 at 8:27 AM, Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com> wrote:
>>> 1) 0001-Rewrite-hash-index-scans-to-work-a-page-at-a-time.patch: this
>>> patch rewrites the hash index scan module to work in page-at-a-time
>>> mode. It basically introduces two new functions-- _hash_readpage() and
>>> _hash_saveitem(). The former is used to load all the qualifying tuples
>>> from a target bucket or overflow page into an items array. The latter
>>> one is used by _hash_readpage to save all the qualifying tuples found
>>> in a page into an items array. Apart from that, this patch bascially
>>> cleans _hash_first(), _hash_next and hashgettuple().
>> I see that forward and backward scan cases of function _hash_readpage contain a lot of code duplication
>> Could you please refactor this function to have less code duplication?
> Sure, I will try to avoid the code duplication as much as possible.
I had close look into hash_readpage() function and could see that
there are only few if-else conditions which are similar for both
forward and backward scan cases and can't be optimised further.
However, If you have a cursory look into this function definition it
looks like the code for forward and backward scan are exactly the same
but that's not the case. Attached is the diff report
(hash_readpage.html) for forward and backward scan code used in
hash_readpage(). This shows what all lines in the hash_readpage() are
same or different.
Please note that before applying the patches for page scan mode in
hash index you may need to first apply the following patches on HEAD,
1) v10 patch for WAL in hash index - 
2) v1 patch for wal consistency check for hash index - 
3) v6 patch for microvacuum support in hash index - 
 - https://www.postgresql.org/message-id/CAA4eK1%2Bk5wR4-kAjPqLoKemuHayQd6RkQQT9gheTfpn%2B72o1UA%40mail.gmail.com
 - https://www.postgresql.org/message-id/CAGz5QCKPU2qX75B1bB_LuEC88xWZa5L55J0TLvYMVD8noSH3pA%40mail.gmail.com
 - https://www.postgresql.org/message-id/CAE9k0PkYpAPDJBfgia08o7XhO8nypH9WoO9M8%3DdqLrwwObXKcw%40mail.gmail.com
|Next Message||Heikki Linnakangas||2017-03-14 12:36:51||Authentication tests, and plain 'password' authentication with a SCRAM verifier|
|Previous Message||Alvaro Herrera||2017-03-14 12:07:21||Re: PATCH: pageinspect / add page_checksum and bt_page_items(bytea)|