From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <vdavydov.dev@gmail.com>
Date: Tue, 21 Aug 2018 18:37:54 +0300
From: Vladimir Davydov <vdavydov.dev@gmail.com>
Subject: Re: [PATCH v2 2/7] vinyl: teach write iterator to return overwritten
 tuples
Message-ID: <20180821153754.xy6m5mtzzozhdxrg@esperanza>
References: <cover.1534847663.git.vdavydov.dev@gmail.com>
 <e43cfefb0c13edbd2df9fb26c4198bd967e349c9.1534847663.git.vdavydov.dev@gmail.com>
 <20180821151457.GB28159@chai>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20180821151457.GB28159@chai>
To: Konstantin Osipov <kostja@tarantool.org>
Cc: tarantool-patches@freelists.org
List-ID: <tarantool-patches.dev.tarantool.org>

On Tue, Aug 21, 2018 at 06:14:57PM +0300, Konstantin Osipov wrote:
> * Vladimir Davydov <vdavydov.dev@gmail.com> [18/08/21 15:19]:
> >  	/*
> > +	 * Deferred DELETE statements can only be produced by
> > +	 * primary index compaction.
> > +	 */
> > +	assert(is_primary || handler == NULL);
> > +	/*
> 
> With this assert, do you really need this check:
> 
> > +static int
> > +vy_write_iterator_deferred_delete(struct vy_write_iterator *stream,
> > +				  struct tuple *stmt)
> > +{
> > +	if (!stream->is_primary)
> > +		return 0;
> 
> ?

Yes. When we dump a primary index, we set 'handler' to NULL, just like
for a secondary index, but in contrast to a secondary index, we have to
include the last statement marked as VY_STMT_DEFERRED_DELETE into the
iterator output even if it is not referenced by any read view, because
we will need it to generate a deferred DELETE statement on compaction.