[Tarantool-patches] [PATCH v3 1/2] vinyl: rework upsert operation

Aleksandr Lyapunov alyapunov at tarantool.org
Tue Oct 13 22:00:27 MSK 2020


Hi, thanks for the patch! Se my 2 comments below:
(the second seems to be important)
Once they resolved, the whole patch is LGTM!

On 10/3/20 4:28 PM, Nikita Pettik wrote:
> into insert the upsert's tuple and the tuple stored in index can be
> different In our particular case, tuple stored on disk has two fields
Dot before `In` is missing.
> +		/*
> +		 * If it turns out that resulting tuple modifies primary
> +		 * key, then simply ignore this upsert.
> +		 */
> +		if (stmt != NULL &&
> +		    vy_apply_result_does_cross_pk(stmt, exec_res,
> +						  exec_res + mp_size, cmp_def,
> +						  column_mask)) {
I guess in case if `stmt_is_void = true` we must compare with `upsert` 
tuple,
not with `stmt` tuple.
> +			if (!suppress_error) {
> +				say_error("upsert operations %s are not applied"\
> +					  " due to primary key modification",
> +					  mp_str(ups_ops));
> +			}
> +			ups_ops = ups_ops_end;
> +			continue;
> +		}
> +		ups_ops = ups_ops_end;
> +		/*
> +		 * Result statement must satisfy space's format. Since upsert's
> +		 * tuple correctness is already checked in vy_upsert(), let's
> +		 * use its format to provide result verification.
> +		 */
> +		struct tuple_format *format = tuple_format(upsert);
> +		if (tuple_validate_raw(format, exec_res) != 0) {
> +			if (! suppress_error)
> +				diag_log();
> +			continue;
> +		}
> +		result_mp = exec_res;
> +		result_mp_end = exec_res + mp_size;
> +	}
> +	struct tuple *new_terminal_stmt = vy_stmt_new_replace(format, result_mp,
> +							      result_mp_end);
> +	region_truncate(region, region_svp);
> +	if (new_terminal_stmt == NULL)
> +		return NULL;
> +	vy_stmt_set_lsn(new_terminal_stmt, vy_stmt_lsn(upsert));
> +	return new_terminal_stmt;
> +}


More information about the Tarantool-patches mailing list