Re: [Tarantool-patches] [PATCH v4 2/2] box: introduce _vsession_settings sysview

[Konstantin Osipov <kostja.osipov@gmail.com>]

* Mergen Imeev <imeevma@tarantool.org> [19/11/27 13:56]:
> Hi! Thanks for the review. However, I think you are a little
> wrong. We do not save these settings anywhere except for the
> struct session. When a user reads from the space, it creates
> tuples on the fly and passes them to the user. All of this can
> actually be read from the commit-message:
> 
> 
> box: introduce _vsession_settings sysview
> 
> This patch creates the _vsession_settings sysview. This sysview
> contains names and values of the session settings.
> 
> This sysview creates tuples on the fly using its own get() and
> create_iterator() methods. This allows us to get a tuple without
> having to save it anywhere. But this means that every time we get
> a tuple from this system view, it is a new tuple, even if they
> look the same:
> 
> tarantool> v = box.space._vsession_settings
> tarantool> v:get({'sql_default_engine'}) ~= v:get({'sql_default_engine'})
> ---
> - true
> ...
> 
> Currently, only SQL settings can be extracted from this sysview,
> since the only session settings are SQL settings.

Where do you get the value from to build the tuple?

Where is it stored? If it is stored per-session, I mean
struct-session, then my comment is still valid.

I'm not challenging the implementation details (I did not read the
patch indeed), I am challenging the semantics. 

The view is called _vsession_settings, so I assume its state
(semantically) is local to each database connection, to which I
objected.

-- 
Konstantin Osipov, Moscow, Russia