Re: [PATCH v2 1/2] memtx: do not use space_vtab::commit_alter for freeing tuples

[Vladimir Davydov <vdavydov.dev@gmail.com>]

On Fri, Apr 06, 2018 at 05:08:27PM +0300, Konstantin Osipov wrote:
> * Vladimir Davydov <vdavydov.dev@gmail.com> [18/04/06 16:20]:
> > +
> > +static void
> > +memtx_index_free_tuples(struct index *index)
> > +{
> > -static void
> > -memtx_space_prune(struct space *space)
> 
> I liked the old term (prune) more (memtx_index_prune()).
> Please consider keeping it.

No problem, I will keep the old name.

> 
> > index ec6f6db6..ebb54f05 100644
> > --- a/src/box/memtx_space.c
> > +++ b/src/box/memtx_space.c
> > @@ -803,41 +803,17 @@ memtx_space_build_secondary_key(struct space *old_space,
> >  			break;
> >  		assert(old_tuple == NULL); /* Guaranteed by DUP_INSERT. */
> >  		(void) old_tuple;
> > +		/*
> > +		 * All tuples stored in a memtx space must be
> > +		 * referenced by the primary index.
> > +		 */
> > +		if (new_index->def->iid == 0)
> > +			tuple_ref(tuple);
> 
> I don't understand this change. This method builds a secondary
> key.
>         /**                                                                     
>          * Called with the new empty secondary index.                           
>          * Fill the new index with data from the primary                        
>          * key of the space.                                                    
>          */  
> 
> How is that possible that new_index->def->iid is 0 here?
> 
> If you're re-using this static function, then please rename it.

This method is used for rebuilding primary keys as well as secondary.
Yeah, this is confusing as hell. I have a patch renaming it and I'm
planning to submit it soon (it's a part of my ALTER rework).

> 
> Other than these two comments, the patch is OK to push.