Re: [gnome-db] gda_holder_set_value_static_str

["Vivien Malerba" <vmalerba gmail com>]

2008/8/25 Massimo Cora' <maxcvs email it>:
> Hi Vivien,
>
>
> while trying to optimize the speed of some queries I came on
> g_value_set_static_string (), which is good and faster than its brother
> g_value_set_str ().
> I was wondering if it was possibile to add a
> gda_holder_set_value_static_str ()?

This can be done, even if it would be usefull only for GdaHolder with
a G_TYPE_STRING value.

>
> Another question, always in that fashion:
> gda_holder_take_value () takes as input a *value that is freed when the
> holder is re-set. That method returns TRUE if value has been set.
> I was thinking about this speed improvement:
> create a gda_holder_take_value_do_not_free () [well, the name is
> unlucky...] that when is re-set will return the stored value instead of
> freeing it. In this manner it's possible to reuse that GValue without a
> g_free () and g_value_new_*().

Could you provide a patch for this?

>
> Think for instance to a prepared query that needs a gchararray parameter
> in a function foo_func ():
>
> * create and set a GValue of type G_TYPE_STRING on a
> global_managed_gvalue1 in a static way. [i.e. no string
> allocations/duplications required]
>
> * get the list of holders with gda_statement_get_parameters ().
>
> * set the global_managed_gvalue1 to the holder with
> gda_holder_take_value_do_not_free ().
>
> * the above method will return NULL because no previous values were taken.
>
> * ... do some stuff then return on the foo_func () to re-exec the query
>
> * re-set the holder's value with a global_managed_gvalue2 and take the
> returned old value global_managed_gvalue1, which is now set as
> 'reusable' for a memory objects pool.
>
> total number of allocations: holders for the query and two globals
> values, better if allocated at program start. No string extra dups are
> required at all.
>
> Currently the above method will require something like: holders for the
> query, two allocations for the GValues and two dups, and the respective
> gda_value_free () when a new value replaced within the holder with
> gda_holder_take_value ().
>
>
> The thoughts can be done with gda_holder_set_value_static_str ().
>
> I'm sure I can achieve better performances if those method would be
> available, what do you think?

That seems a good idea!

Cheers,

Vivien