On Sat, May 14, 2011 at 16:26, Hyrum K Wright <hy...@hyrumwright.org> wrote:
> On Sat, May 14, 2011 at 7:29 PM, Greg Stein <gst...@gmail.com> wrote:
>...
>>>...
>>> + struct set_changelist_baton_t scb = { new_changelist, changelist_filter,
>>> + depth };
>>> + int final_queries[] = { STMT_DROP_CHANGELIST_LIST,
>>> STMT_DROP_TARGETS_LIST,
>>> + -1 };
>>
>> It seems that it would be cleaner to just put the 'DROP TABLE
>> targets_list' into the STMT_DROP_CHANGELIST_LIST (maybe rename the
>> latter to something like STMT_FINALIZE_CHANGELIST). Then you could rip
>> out all this multiple statement stuff. I think one STMT would be
>> cleaner than all this code to support multiple.
>
> The goal is to use the target_list stuff for multiple operations, such
> as propset (and potentially things like info, too). As a result, the
> 'DROP TABLE targets_list' will be used in combination with a number of
> other finalize statements. I suppose we could duplicate that single
> drop statement wherever we need to finalize an operation (and maybe
> that's acceptable duplication), but I'd like to leave it separate, at
> least for now.
That was my thought: just put that DROP into the other finalize
statements. We already drop a bunch of triggers and the change_list
table. I see no reason not to throw a drop of the targets_list in
there, too.
It just seems easier to have multiple SQL operations in one STMT,
rather than simulate the same in C code.
Cheers,
-g
