On Mon, Aug 31, 2020 at 12:18:09PM -0400, Alvaro Herrera wrote: >On 2020-Aug-31, Tomas Vondra wrote: > >> I wonder if trying to list info about all stats from the statistics >> object in a single line is necessary. Maybe we should split the info >> into one line per statistics, so for example >> >> CREATE STATISTICS s (mcv, ndistinct, dependencies) ON ... >> >> would result in three lines in the \dX output. The statistics name would >> identify which lines belong together, but other than that the pieces are >> mostly independent. > >Yeah, that's what I'm suggesting. I don't think we need to repeat the >name/definition for each line though. > >It might be useful to know how does pspg show a single entry that's >split in three lines, though. >
Ah, I didn't realize you're proposing that - I assumed it's broken simply to make it readable, or something like that. I think the lines are mostly independent, so I'd suggest to include the name of the object on each line. The question is whether this independence will remain true in the future - for example histograms would be built only on data not represented by the MCV list, so there's a close dependency there.
Not sure about pspg, and I'm not sure it matters too much.
pspg almost ignores multiline rows - the horizontal cursor is one row every time. There is only one use case where pspg detects multiline rows - sorts, and pspg ensures correct content for multiline rows displayed in different (than input) order.
Regards
Pavel
regards
-- Tomas Vondra http://www.2ndQuadrant.com PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services