Horton

Thread

Commits

  1. Move syncscan.c to src/backend/access/common.

  1. Move syncscan.c?

    Thomas Munro <thomas.munro@gmail.com> — 2020-06-23T01:30:39Z

    Hi,

It's a bit odd that syncscan.c is used by both heapam.c and tableam.c,
and provides a generic block-synchronization mechanism that other
table AMs might want to use too, but it lives under
src/backend/access/heap.  It doesn't actually do anything heap
specific (beyond being block-oriented), and it's weird that tableam.c
has to include heapam.h.

Perhaps we should move the .c file under src/backend/access/table, as attached.

I suppose it's remotely possible that someone might invent
physical-order index scans, and once you have those you might sync
scans of those too, and then even table would be too specific, but
that may be a bit far fetched.

  2. Re: Move syncscan.c?

    Andres Freund <andres@anarazel.de> — 2020-06-23T18:28:52Z

    Hi,

On 2020-06-23 13:30:39 +1200, Thomas Munro wrote:
> It's a bit odd that syncscan.c is used by both heapam.c and tableam.c,
> and provides a generic block-synchronization mechanism that other
> table AMs might want to use too, but it lives under
> src/backend/access/heap.  It doesn't actually do anything heap
> specific (beyond being block-oriented), and it's weird that tableam.c
> has to include heapam.h.
> 
> Perhaps we should move the .c file under src/backend/access/table, as attached.

Sounds reasonable. I suspect there's a few more files (and definitely
functions) that could be de-heapified.


> I suppose it's remotely possible that someone might invent
> physical-order index scans, and once you have those you might sync
> scans of those too, and then even table would be too specific, but
> that may be a bit far fetched.

Hm. That'd be an argument for moving it to access/common. I don't really
see a reason not to go for that?

Greetings,

Andres Freund

  3. Re: Move syncscan.c?

    Thomas Munro <thomas.munro@gmail.com> — 2020-07-29T05:04:15Z

    On Wed, Jun 24, 2020 at 6:28 AM Andres Freund <andres@anarazel.de> wrote:
> On 2020-06-23 13:30:39 +1200, Thomas Munro wrote:
> > I suppose it's remotely possible that someone might invent
> > physical-order index scans, and once you have those you might sync
> > scans of those too, and then even table would be too specific, but
> > that may be a bit far fetched.
>
> Hm. That'd be an argument for moving it to access/common. I don't really
> see a reason not to go for that?

Ok, done that way.  Thanks.