Re: [PATCH v5 8/9] box: introduce offset slot cache in key_part

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

On Tue, Dec 04, 2018 at 12:04:18AM +0300, Vladimir Davydov wrote:
> On Mon, Nov 26, 2018 at 01:49:42PM +0300, Kirill Shcherbatov wrote:
> > diff --git a/src/box/alter.cc b/src/box/alter.cc
> > index 029da02..6291159 100644
> > --- a/src/box/alter.cc
> > +++ b/src/box/alter.cc
> > @@ -856,7 +856,10 @@ alter_space_do(struct txn *txn, struct alter_space *alter)
> >  	 * Create a new (empty) space for the new definition.
> >  	 * Sic: the triggers are not moved over yet.
> >  	 */
> > -	alter->new_space = space_new_xc(alter->space_def, &alter->key_list);
> > +	alter->new_space =
> > +		space_new_xc(alter->space_def, &alter->key_list,
> > +			     alter->old_space->format != NULL ?
> > +			     alter->old_space->format->epoch + 1 : 1);
> 
> Passing the epoch around looks ugly. Let's introduce a global counter
> and bump it right in tuple_format_new(). If you worry about disk_format
> and mem_format having different epochs in vinyl, it's not really a
> problem: make epoch an optional argument to tuple_format_new(); if the
> caller passes an epoch, use it, otherwise generate a new one; in vinyl
> reuse mem_format->epoch for disk_format.
> 
> > diff --git a/src/box/key_def.h b/src/box/key_def.h
> > index 7731e48..3e08eb4 100644
> > --- a/src/box/key_def.h
> > +++ b/src/box/key_def.h
> > @@ -95,6 +95,14 @@ struct key_part {
> >  	char *path;
> >  	/** The length of JSON path. */
> >  	uint32_t path_len;
> > +	/**
> > +	 * Source format for offset_slot_cache hit validations.
> > +	 * Cache is expected to use "the format with the newest
> > +	 * epoch is most relevant" strategy.
> > +	 */
> > +	struct tuple_format *format_cache;
> 
> Why did you decide to store tuple_format in key_part instead of epoch,
> as you did originally? I liked that more.

Discussed verbally, agreed that:
 - Storing a pointer to the format in key_part is potentially dangerous,
   because pointers may coincide after reallocation, resulting in
   invalid behavior.
 - We should make the epoch counter global rather than passing it around
   on space creation. That is, tuple_format_new() will bump a global
   int64_t counter and assign the new value to the new format.
 - In vinyl disk_format and mem_format must have different epochs,
   because they have different field maps so no optional argument
   argument is required for tuple_format_new() - it will always assign a
   new epoch to the new format.
 - 'Epoch' isn't a very good name. We should probably call it 'version'
   or 'format_version' or 'tuple_format_version', depending on the
   context.