Re: [Tarantool-patches] [PATCH v2 07/16] port: add dump format and request type to port_sql

[Nikita Pettik <korablev@tarantool.org>]

On 04 Dec 14:52, Konstantin Osipov wrote:
> * Nikita Pettik <korablev@tarantool.org> [19/11/21 10:00]:
> > Dump formats of DQL and DML queries are different: the last one contains
> 
> Ergh, this is not dump, but result set/serialization format. Dump
> in SQL is usually used for logical or physical database
> backup.

Okay, let's rename to sql_serialization_format
 
> > number of affected rows and optionally list of autoincremented ids; the
> > first one comprises all meta-information including column names of
> > resulting set and their types. What is more, dump format is going to be
> > different for execute and prepare requests. So let's introduce separate
> > member to struct port_sql responsible for dump format to be used.
> > 
> > What is more, prepared statement finalization is required only for
> > PREPARE-AND-EXECUTE requests. So let's keep request type in port as well.
> 
> > 
> > Note that C standard specifies that enums are integers, but it does not
> > specify the size. Hence, let's use simple uint8 - mentioned enums are
> > small enough to fit into it.
> 
> enum sizeof in C and older C++ is implementation dependent.
> 
> what do you mean here?

To store uint8 numbers instead of enums (due to their implementation
dependent sizes).

> > +	struct port_sql *port_sql = (struct port_sql *) base;
> > +	if (port_sql->request == PREPARE_AND_EXECUTE)
> > +		sql_finalize(((struct port_sql *)base)->stmt);
> 
> Does this work with the statement object only or with the cache as
> well? 

Only with prepared statement objects.
 
> I suggest we introduce a clear naming for API calls:
> 
> sql_stmt_* - for statement object api
> sql_stmt_cache_* - for prepared statement api

> sql_finalize, sql_prepare, sql_execute - for SQL high level API
> which manipulates both statements and the cache.
> 
> What do you think?

Ok, will move renaming sql_finalize -> sql_stmt_finalize etc to
a separate patch. I'm okay with suggested naming and I'm going
to use it in the next patches.
 
> > -port_sql_create(struct port *port, struct sql_stmt *stmt)
> > +port_sql_create(struct port *port, struct sql_stmt *stmt,
> > +		enum sql_dump_format dump_format, enum sql_request_type request)
> >  {
> >  	port_tuple_create(port);
> > -	((struct port_sql *)port)->stmt = stmt;
> >  	port->vtab = &port_sql_vtab;
> > +	struct port_sql *port_sql = (struct port_sql *) port;
> > +	port_sql->stmt = stmt;
> > +	port_sql->dump_format = dump_format;
> 
> Let's use sql_result_set_format? Do you have to introduce this
> enum ? This information can be derived from sql_request_type, no +
> statement type, no? If yes, I suggest to have a function, which
> returns it, rather than store.

To save 1 byte? :) Doesn't seem to be reasonable tho.
Current implementation looks quite suitable in code.
 
> > +/**
> > + * One of possible formats used to dump msgpack/Lua.
> > + * For details see port_sql_dump_msgpack() and port_sql_dump_lua().
> 
> 
> > + */
> > +enum sql_dump_format {
> > +	DQL_EXECUTE = 0,
> > +	DML_EXECUTE = 1,
> > +	DQL_PREPARE = 2,
> > +	DML_PREPARE = 3,
> > +};
> 
> Neither the names of the members nor comments convey what is going
> on here - looks like a simple matrix dml/dql * prepare/execute.

That's it.

> Having a couple of if statments is not such a big deal IMHO.

With this enum code looks way much better.
 
> > +	/**
> > +	 * Dump format depends on type of SQL query: DML or DQL;
> > +	 * and on type of SQL request: execute or prepare.
> > +	 */
> > +	uint8_t dump_format;
> > +	/** enum sql_request_type */
> > +	uint8_t request;
> 
> If I had to add these constants, I would not use type-erasing
> uint8 here, but use max value for enum in C/C++, which is 64 bits
> and the original type. 

Now one is replaced with do_finalize as Vlad suggested.