Re: [HACKERS] ANALYZE command progress checker

On 3/6/17 12:49 AM, Michael Paquier wrote:
> On Sat, Mar 4, 2017 at 5:33 AM, David Steele <david@pgmasters.net> wrote:
>> I think the idea of a general progress view is very valuable and there
>> are a ton of operations it could be used for:  full table scans, index
>> rebuilds, vacuum, copy, etc.
>>
>> However, I feel that this proposal is not flexible enough and comes too
>> late in the release cycle to allow development into something that could
>> be committed.
>
> Well, each command really has its own requirements in terms of data to
> store, so we either finish with a bunch of small tables that anyone
> could query and join as they wish or a somewhat unique table that is
> bloated with all the information, with a set of views on top of it to
> query all the information. For extensibility's sake of each command
> (for example imagine that REINDEX could be extended with a
> CONCURRENTLY option and multiple phases), I would think that having a
> table per command type would not be that bad.

Well, the ideal scenario is that someone uses the raw data to come up 
with a good way to just provide ye olde 0-100% progress bar. At that 
point a single view would do the trick.

Perhaps instead of adding more clutter to \dvS we could just have a SRF 
for now. At over 2800 rows currently, you're not going to notice one 
more addition to \dfS.
-- 
Jim Nasby, Chief Data Architect, OpenSCG
http://OpenSCG.com