[tarantool-patches] Re: [PATCH v2 3/7] box: create new methods for ephemeral spaces

[Kirill Shcherbatov]

>  /**
> + * Fill up request according to given
> + * operation type. For all cases but
> + * UPSERT argument name match names of
> + * fields of struct request. For UPSERT
> + * arguments key and key_end are fields
> + * ops and ops_end.
> + */
1. Please, write doxygen-style comment with right margin 66.

The other nitpicks are similar: you should be at least consistent with functions
comments: I mean, memtx_space_ephemeral_replace has _invalid_ comment, but at 
the same time memtx_space_execute_upsert doesn't have comment at all.
Please, clean-up this by your own.
 

> +memtx_space_ephemeral_replace(struct space *space, struct txn *txn,
> +			      struct request *request, struct tuple **result)
>  {
> -	struct memtx_space *memtx_space = (struct memtx_space *)space;
> -	struct tuple *new_tuple = memtx_tuple_new(space->format, tuple,
> -						  tuple_end);
> -	if (new_tuple == NULL)
> -		return -1;
> -	tuple_ref(new_tuple);
> +	assert(txn == NULL);
> +	(void)txn;
> +	struct tuple *new_tuple = NULL;
>  	struct tuple *old_tuple = NULL;
> -	if (memtx_space->replace(space, old_tuple, new_tuple,
> -				 DUP_REPLACE_OR_INSERT, &old_tuple) != 0) {
> +	int rc = memtx_space_replace_impl(space, request, &new_tuple,
> +					  &old_tuple, result);
> +	if (rc && new_tuple != NULL)
2. rc != 0

3. Looks like you introduce interesting functional changes. They deserve to be tested, don't
you think? Or there is no way to test this new API yet? I don't insist on C-module tests if this
the only way to test it.