|Gregory Stark||Jul 18, 2006 11:35 am|
|Luke Lonergan||Jul 18, 2006 11:44 am|
|Dave Page||Jul 18, 2006 1:08 pm|
|Andreas Pflug||Jul 18, 2006 5:12 pm|
|Neil Conway||Jul 18, 2006 6:52 pm|
|Tom Lane||Jul 18, 2006 8:23 pm|
|Josh Berkus||Jul 18, 2006 9:24 pm|
|Greg Stark||Jul 19, 2006 2:18 am|
|Hannu Krosing||Jul 19, 2006 2:33 am|
|Dave Page||Jul 19, 2006 2:35 am|
|Andreas Pflug||Jul 19, 2006 5:23 am|
|Tom Lane||Jul 19, 2006 7:33 am|
|Darcy Buskermolen||Jul 19, 2006 8:54 am|
|Andrew Hammond||Jul 19, 2006 10:29 am|
|Christopher Kings-Lynne||Jul 19, 2006 6:38 pm|
|Agent M||Jul 19, 2006 7:40 pm|
|Csaba Nagy||Jul 20, 2006 1:51 am|
|Luke Lonergan||Jul 20, 2006 8:36 am|
|Subject:||Progress bar updates|
|From:||Gregory Stark (gsst...@mit.edu)|
|Date:||Jul 18, 2006 11:35:33 am|
Has anyone looked thought about what it would take to get progress bars from clients like pgadmin? (Or dare I even suggest psql:)
My first thought would be a message like CancelQuery which would cause the backend to peek into a static data structure and return a message that the client could parse and display something intelligent. Various commands would then stuff information into this data structure as they worked.
For a first cut this "data structure" could just be a float between 0 and 1. Or perhaps it should be two integers, a "current" and an "estimated final". That would let the client do more intelligent things when the estimates change for the length of the whole job.
Later I could imagine elaborating into more complex structures for representing multi-step processes or even whole query plans. I also see it possibly being interesting to stuff this data structure into shared memory handled just like how Tom handled the "current command". That would let you see the other queries running on the server, how long they've been running, and estimates for how long they'll continue to run.
I would suggest starting with utility functions like index builds or COPY which would have to be specially handled anyways. Handling all optimizable queries in a single generic implementation seems like something to tackle only once the basic infrastructure is there and working for simple cases.
Of course the estimates would be not much better than guesses. But if you want to say it's not worth having since they won't be perfectly accurate be prepared to swear that you've never looked at the "% complete" that modern ftp clients and web browsers display even though they too are, of course, widely inaccurate. They nonetheless provide some feedback the user desperately wants to be reassured that his job is making progress and isn't years away from finishing.
-- Gregory Stark EnterpriseDB http://www.enterprisedb.com