[tarantool-patches] Re: [PATCH v1 3/3] sql: get rid of Column structure

["n.pettik" <korablev@tarantool.org>]

Again: why did this patch trapped in patch-set? All three patches seem to be independent.

> On 18 Jul 2018, at 19:52, Kirill Shcherbatov <kshcherbatov@tarantool.org> wrote:
> 
> Get rid of is_primkey in Column structure as it become

Typo: ‘becomes’.

> redundant. Moved the last member coll with collation pointer
> to field_def structure.

Why do you need to move collation ptr to field_def? It already features collation id,
so you can always get pointer to it by simple lookup. It would make sense if it was
utilised everywhere. But I see assignment only in sqlite3SelectAddColumnTypeAndCollation()
and no real usages..Lets remove it at all.

> @@ -1692,22 +1680,9 @@ sqlite3EndTable(Parse * pParse,	/* Parse context */
> 		}
> 	}
> 
> -	/* Set default on_nullable action if required. */
> -	struct field_def *field = p->def->fields;
> -	for (uint32_t i = 0; i < p->def->field_count; ++i, ++field) {
> -		if (field->nullable_action == on_conflict_action_MAX) {
> -			field->nullable_action = ON_CONFLICT_ACTION_NONE;
> -			field->is_nullable = true;
> -		}
> -	}
> -
> -	if (check_on_conflict_replace_entries(p)) {
> -		sqlite3ErrorMsg(pParse,
> -				"only PRIMARY KEY constraint can "
> -				"have ON CONFLICT REPLACE clause "
> -				"- %s", p->def->name);

Here you are simply fixing changes made in first patch, so mb it is better to
move them to first patch?