Hi Alik,
I have a new patch ready. See:
http://lists.mysql.com/commits/41506
Note: This is a second patch because I merged the WL#4205 code and can't (or
don't know how) to redo my original patch.
I also address your points below.
> > I would extend open_schema_table() with COND* parameter -
> that would
> > make it much more general as one could open a I_S table
> with arbitrary
> > select condition. The usage would be then:
> >
> > COND *cond;
> >
> > <build condition cond>
> >
> > TABLE *t= open_schema_table(thd, st, cond);
>
> Agreed with open_schema_table() signature.
>
> However, I think the code should be something like:
>
> COND *cond= create_db_select_condition(...) TABLE *t=
> open_schema_table(thd, st, cond); ...
> delete cond;
Well, I cannot change the signature like Rafal requested. To build the COND*
tree one needs the table and open_schema_table returns the table pointer. So
it has to be done in two steps.
> > OPTIONAL: I'm not sure if it really makes sense to extract this
> > functionality into a separate function. The code creating
> appropriate
> > WHERE clause could be used directly inside
> > InformationSchemaIterator::prepare_is_table().
>
> I would keep it in a separate function for the sake of
> keeping functions small.
Agreed.
> > > +COND *create_db_select_condition(THD *thd,
> > > + TABLE *t,
> > > + List<LEX_STRING> db_list)
>
> Chuck, you pass List by value here. I think, you didn't mean that.
> Let's pass it by pointer.
DOH! Sometimes I amaze myself with the obsurdity of my gaffs.
> Another thing: I guess created COND* object should be deleted
> somewhere/somehow?
I'm not sure how to do this. Simply doing "delete cond" doesn't work because
it deletes too much and blows core. However, the method
Query_arena::free_items() does indeed delete the COND* tree so all is well
in this respect.
Chuck
