Re: Are there Sql backends that return an array from big_query()?

On Thu, 30 Oct 2014, Stephen R. van den Berg wrote:

> Stephen R. van den Berg wrote:
>> Martin Nilsson (Opera Mini - AFK!) @ Pike (-) importm?te f?r mailinglistan wrote:
>>> I think the architecture is outdated, so instead it would be more
>>> valuable to look at how to make it properly. Instead of having a proxy
>>> object that calls a database object, there should be inheritance that
>>> allows direct calling when possible.
>
>> That's exactly what I was looking at, which is why I wanted to eliminate
>> big_query()s that return anything else than an object.

Sounds reasonable.

> I'm toying with something like this:
>
> Instantiate the .pike programs in the Sql.pmod directory using:
[...]
> As before, this to make sure that we have a driver.
>
> Then, instead of using the compiled program find_dbm returns, we do
> something along the lines of:
>
> program realclass =
> compile_string("inherit Sql.SQL;inherit Sql."+program_name+";");

I don't see a reason not to have the providers inherit the base 
API directly, now that it is reasonably easy to inherit Pike-code 
from modules written in C (cf Nettle and __builtin.Nettle).

> which then needs to be cached somewhere for repeated requests of the
> same dbtype, which can then be instantiated.

It shouldn't be a problem to use the master()->programs cache for this
(which would be automatic in the direct inherit case).

> Or is there some more direct way to create a new class out of two
> existing programs?

Not at Pike level.

-- 
Henrik Grubbström					[email protected]
Roxen Internet Software AB