Re: WIP: Rework access method interface

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> On Jan 16, 2016, at 11:13 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> There are a couple of things we could do instead:
>>
>> * Put each amvalidate function into its own file (but probably keep it
>> in the same directory as now). This is a reasonable response to
>> points #1 and #2 but isn't very much help for #3.
>>
>> * Collect the amvalidate functions into one file, which then leaves
>> us wondering where to put that; surely not under any one AM's directory.
>> A new file in src/backend/access/index/ is one plausible solution.
>> This file would also be a reasonable place to put the amvalidate()
>> dispatch function itself.
>>
>> I'm somewhat leaning to the second choice, but perhaps someone has
>> a better idea, or an argument against doing that.

> Doesn't seem very modular. How about putting common code there but AM-specific code in each AM's directory? It would be nice if adding a new AM meant mostly adding a new directory, not much touching the common code.

Then we're going to end up with option A; and I suspect that we'll never
bother with factoring out any common code, because it won't be worth the
notational trouble if it involves common code that's in a different file
in a different directory.

As for modularity, nobody's moaned particularly about the amcostestimate
functions all being in selfuncs.c. It all depends on what you think is
"modular".

regards, tom lane