[tarantool-patches] Re: [PATCH v2 5/8] sql: introduce a signature_mask for functions

Konstantin Osipov kostja at tarantool.org
Tue Aug 13 11:44:56 MSK 2019


* Kirill Shcherbatov <kshcherbatov at tarantool.org> [19/08/13 11:32]:
> > Good idea, but why make the mask sooo big?-)))
> > I don't think we have more than 3 arguments in any overloaded
> > function?
> 63 arguments is not too much for UDF (this would use for UDF too), I guess.
> Moreover, column_mask API is worked with 64bit bitmasks.

When do you plan to use it for UDFs? I'm not aware of any work in
the pipeline. If it is in 3-6 months or more, then this is clearly
an overkill, as it may never happen
> 
> > 
> >> +	if (!column_mask_fieldno_is_set(p->signature_mask, (uint32_t)nArg))
> >>  		return 0;
> > 
> > Are you sure you want to use column mask api just to test a bit?
> It is convenient, because column_mask is "smart" and it also return true for
> fieldno > 64: this is ok for uniform scalar functions like
> GREATEST(1,.......)

In absence of UDFs you can static-assert that the number of used
bits never goes above 8, and you will never need the smart
features of the column mask.

-- 
Konstantin Osipov, Moscow, Russia




More information about the Tarantool-patches mailing list