From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <vdavydov.dev@gmail.com>
Date: Wed, 13 Mar 2019 15:45:04 +0300
From: Vladimir Davydov <vdavydov.dev@gmail.com>
Subject: Re: [PATCH v2 11/14] vinyl: do not fill secondary tuples with nulls
 when decoded
Message-ID: <20190313124503.2urnl6vaqmlejrop@esperanza>
References: <cover.1552464666.git.vdavydov.dev@gmail.com>
 <00072f9869878409ebefa45232b5015a8441a655.1552464666.git.vdavydov.dev@gmail.com>
 <20190313122559.GK25066@chai>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20190313122559.GK25066@chai>
To: Konstantin Osipov <kostja@tarantool.org>
Cc: tarantool-patches@freelists.org
List-ID: <tarantool-patches.dev.tarantool.org>

On Wed, Mar 13, 2019 at 03:25:59PM +0300, Konstantin Osipov wrote:
> * Vladimir Davydov <vdavydov.dev@gmail.com> [19/03/13 11:58]:
> > +struct key_def *
> > +key_def_extract(const struct key_def *cmp_def, const struct key_def *pk_def,
> > +		struct region *region)
> 
> The name is not very telling :(
> 
> It also looks like a standalone function, not a key_def method. So
> we could name it build_pk_def_for_cmp_def or
> find_primary_key_parts() or something like it.

Well, key_def_merge is also not a key_def method in this regard.
I think of key_def_ more like as of a namespace.

> 
> >  	if (lsm == NULL) {
> >  		free(index);
> >  		return NULL;
> > @@ -1301,13 +1300,34 @@ vy_get_by_secondary_tuple(struct vy_lsm *lsm, struct vy_tx *tx,
> >  			  const struct vy_read_view **rv,
> >  			  struct tuple *tuple, struct tuple **result)
> >  {
> > +	int rc = 0;
> >  	assert(lsm->index_id > 0);
> >  
> > -	if (vy_point_lookup(lsm->pk, tx, rv, tuple, result) != 0)
> > -		return -1;
> > +	/*
> > +	 * Lookup the full tuple by a secondary statement.
> > +	 * There are two cases: the secondary statement may be
> > +	 * a key, in which case we need to extract the primary
> Please explain: 
> 
> may be a key, if we got this tuple from disk ... or may be a full
> tuple if we got this tuple from the tuple cache or level 0, 

Okay.

> 
> > +	 * key parts from it; or it may be a full tuple, which
> > +	 * we may pass it immediately to the iterator.
> > +	 */
> > +	struct tuple *key;
> > +		/*
> > +		 * To save disk space, we do not store full tuples
> > +		 * in secondary index runs. Instead we only store
> > +		 * extended keys (i.e. keys consisting of secondary
> > +		 * and primary index parts). This is enough to look
> > +		 * up a full tuple in the primary index.
> > +		 */
> > +		lsm->disk_format = lsm_env->key_format;
> 
> Nice.
> 
> > +
> > +		lsm->pk_in_cmp_def = key_def_extract(lsm->cmp_def, pk->key_def,
> 
> Oh. Shall we call the function find_pk_in_cmp_def()? 

The name's fine. Well, at least one can figure out what the function
does, but I don't quite like that there's no key_def_ prefix.

> > +		if (is_primary)
> > +			stmt = vy_stmt_new_surrogate_from_key(request.key,
> 
> Shouldn't this be renamed?

The next patch removes this function altogether.