Thread: controlling psql's use of the pager a bit more
I often get annoyed because psql is a bit too aggressive when it decides whether to put output through the pager, and the only way to avoid this is to turn the pager off (in which case your next query might dump many thousands of lines to the screen). I'd like a way to be able to specify a minumum number of lines of output before psql would invoke the pager, rather than just always using the terminal window size. Thoughts? cheers andrew
On 13 November 2014 15:52, Andrew Dunstan <andrew@dunslane.net> wrote:
--
I often get annoyed because psql is a bit too aggressive when it decides whether to put output through the pager, and the only way to avoid this is to turn the pager off (in which case your next query might dump many thousands of lines to the screen). I'd like a way to be able to specify a minumum number of lines of output before psql would invoke the pager, rather than just always using the terminal window size.
+1
I've found this annoying myself.
Thom
Andrew Dunstan <andrew@dunslane.net> writes:
> I often get annoyed because psql is a bit too aggressive when it decides
> whether to put output through the pager, and the only way to avoid this
> is to turn the pager off (in which case your next query might dump many
> thousands of lines to the screen). I'd like a way to be able to specify
> a minumum number of lines of output before psql would invoke the pager,
> rather than just always using the terminal window size.
Are you saying you'd want to set the threshold to *more* than the window
height? Why?
My impression is that the problem is not with the threshold as such,
it's that psql is not terribly careful about determining how many lines
it will put out (ie, the value being compared to the threshold isn't
100% reliable). If that's correct, then a hand-set threshold isn't
really going to make you happier, and what's needed is some hard work
on improving the quality of the lines-to-be-output estimates.
regards, tom lane
On 2014-11-13 11:09:06 -0500, Tom Lane wrote: > Andrew Dunstan <andrew@dunslane.net> writes: > > I often get annoyed because psql is a bit too aggressive when it decides > > whether to put output through the pager, and the only way to avoid this > > is to turn the pager off (in which case your next query might dump many > > thousands of lines to the screen). I'd like a way to be able to specify > > a minumum number of lines of output before psql would invoke the pager, > > rather than just always using the terminal window size. > > Are you saying you'd want to set the threshold to *more* than the window > height? Why? Not sure what that'd be useful for. What I've noticed more than once is that it'd be useful to force the pager if the result is wider than the screen. > My impression is that the problem is not with the threshold as such, > it's that psql is not terribly careful about determining how many lines > it will put out (ie, the value being compared to the threshold isn't > 100% reliable). If that's correct, then a hand-set threshold isn't > really going to make you happier, and what's needed is some hard work > on improving the quality of the lines-to-be-output estimates. If it's that, I've personally solved that problem for me by adding -F (aka --quit-if-one-screen) to $LESS. Greetings, Andres Freund -- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services
On 11/13/2014 11:09 AM, Tom Lane wrote: > Andrew Dunstan <andrew@dunslane.net> writes: >> I often get annoyed because psql is a bit too aggressive when it decides >> whether to put output through the pager, and the only way to avoid this >> is to turn the pager off (in which case your next query might dump many >> thousands of lines to the screen). I'd like a way to be able to specify >> a minumum number of lines of output before psql would invoke the pager, >> rather than just always using the terminal window size. > Are you saying you'd want to set the threshold to *more* than the window > height? Why? Because I might be quite happy with 100 or 200 lines I can just scroll in my terminal's scroll buffer, but want to use the pager for more than that. This is useful especially if I want to scroll back and see the results from a query or two ago. cheers andrew
On Thu, Nov 13, 2014 at 05:14:47PM +0100, Andres Freund wrote: > On 2014-11-13 11:09:06 -0500, Tom Lane wrote: > > Andrew Dunstan <andrew@dunslane.net> writes: > > > I often get annoyed because psql is a bit too aggressive when it decides > > > whether to put output through the pager, and the only way to avoid this > > > is to turn the pager off (in which case your next query might dump many > > > thousands of lines to the screen). I'd like a way to be able to specify > > > a minumum number of lines of output before psql would invoke the pager, > > > rather than just always using the terminal window size. > > > > Are you saying you'd want to set the threshold to *more* than the window > > height? Why? > > Not sure what that'd be useful for. If the output is of the same order of magnitude as the window height, it is still easy to compare the new output with whatever was above it by conventional terminal scrolling (pgup, screen scrolling, whatever, if present). If a comparison is desired in the first place, that is. If the output goes to the pager, you're stuck with it. Unless of course if the pager could somehow include the previous output up to some threshold (while still being positioned at the beginning of the new output in order to retain current behaviour), so one could scroll up in the pager window. That would rock. Michael -- Michael Banck Tel.: +49 (0) 2161 / 4643-171 credativ GmbH, HRB Mönchengladbach 12080 Hohenzollernstr. 133, 41061 Mönchengladbach Geschäftsführung: Dr. Michael Meskes, Jörg Folz, Sascha Heuer
Andrew Dunstan wrote > On 11/13/2014 11:09 AM, Tom Lane wrote: >> Andrew Dunstan < > andrew@ > > writes: >>> I often get annoyed because psql is a bit too aggressive when it decides >>> whether to put output through the pager, and the only way to avoid this >>> is to turn the pager off (in which case your next query might dump many >>> thousands of lines to the screen). I'd like a way to be able to specify >>> a minumum number of lines of output before psql would invoke the pager, >>> rather than just always using the terminal window size. >> Are you saying you'd want to set the threshold to *more* than the window >> height? Why? > > > Because I might be quite happy with 100 or 200 lines I can just scroll > in my terminal's scroll buffer, but want to use the pager for more than > that. This is useful especially if I want to scroll back and see the > results from a query or two ago. +1 David J. -- View this message in context: http://postgresql.nabble.com/controlling-psql-s-use-of-the-pager-a-bit-more-tp5826819p5826833.html Sent from the PostgreSQL - hackers mailing list archive at Nabble.com.
On Thu, Nov 13, 2014 at 11:54 AM, David G Johnston <david.g.johnston@gmail.com> wrote: >> Because I might be quite happy with 100 or 200 lines I can just scroll >> in my terminal's scroll buffer, but want to use the pager for more than >> that. This is useful especially if I want to scroll back and see the >> results from a query or two ago. > > +1 +1 from me, too. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
On Thu, Nov 13, 2014 at 11:39 AM, Robert Haas <robertmhaas@gmail.com> wrote: > On Thu, Nov 13, 2014 at 11:54 AM, David G Johnston > <david.g.johnston@gmail.com> wrote: >>> Because I might be quite happy with 100 or 200 lines I can just scroll >>> in my terminal's scroll buffer, but want to use the pager for more than >>> that. This is useful especially if I want to scroll back and see the >>> results from a query or two ago. >> >> +1 > > +1 from me, too. me three (as long as the current behaviors are not messed with and the new stuff is 'opt in' somehow -- I also use the force quit option to less). merlin
On Thu, Nov 13, 2014 at 11:39 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Thu, Nov 13, 2014 at 11:54 AM, David G Johnston
> <david.g.johnston@gmail.com> wrote:
>>> Because I might be quite happy with 100 or 200 lines I can just scroll
>>> in my terminal's scroll buffer, but want to use the pager for more than
>>> that. This is useful especially if I want to scroll back and see the
>>> results from a query or two ago.
>>
>> +1
>
> +1 from me, too.
me three (as long as the current behaviors are not messed with and the
new stuff is 'opt in' somehow -- I also use the force quit option to
less).
Being able to use "\pset pager"
to toggle the capability seems useful.
Thus I'd suggest establishing a new "pager_minlines" option that if unset (default) maintains the current behavior but which can have a value (0 to int_max) where 0 ends up basically doing the same thing as "always" mode for "pager". Leaving the value blank will cause the option to be unset again and revert to the current behavior.
David J.
On Thu, Nov 13, 2014 at 7:52 AM, Andrew Dunstan <andrew@dunslane.net> wrote:
I often get annoyed because psql is a bit too aggressive when it decides whether to put output through the pager, and the only way to avoid this is to turn the pager off (in which case your next query might dump many thousands of lines to the screen). I'd like a way to be able to specify a minumum number of lines of output before psql would invoke the pager, rather than just always using the terminal window size.
+1.
If it scrolls 2 or 3 lines off the top, I can just use the scroll-back buffer to look at them, I don't want the pager for that situation.
But I am more interested in width, not length. I don't want it to switch to the pager just because a couple lines would wrap, especially when looking at EXPLAIN output. I have played with changing \pset columns, but then other things get screwy once you set that to a non-actual value. I don't recall of the top of my head which things those are, though.
Cheers,
Jeff
On Thu, Nov 13, 2014 at 8:14 AM, Andres Freund <andres@2ndquadrant.com> wrote:
On 2014-11-13 11:09:06 -0500, Tom Lane wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
> > I often get annoyed because psql is a bit too aggressive when it decides
> > whether to put output through the pager, and the only way to avoid this
> > is to turn the pager off (in which case your next query might dump many
> > thousands of lines to the screen). I'd like a way to be able to specify
> > a minumum number of lines of output before psql would invoke the pager,
> > rather than just always using the terminal window size.
>
> Are you saying you'd want to set the threshold to *more* than the window
> height? Why?
Not sure what that'd be useful for.
What I've noticed more than once is that it'd be useful to force the
pager if the result is wider than the screen.
For which formats do you want that? It is already done that way under \pset format aligned.
Cheers,
Jeff
On Thu, Nov 13, 2014 at 12:03 PM, David Johnston <david.g.johnston@gmail.com> wrote: > On Thu, Nov 13, 2014 at 10:55 AM, Merlin Moncure <mmoncure@gmail.com> wrote: >> >> On Thu, Nov 13, 2014 at 11:39 AM, Robert Haas <robertmhaas@gmail.com> >> wrote: >> > On Thu, Nov 13, 2014 at 11:54 AM, David G Johnston >> > <david.g.johnston@gmail.com> wrote: >> >>> Because I might be quite happy with 100 or 200 lines I can just scroll >> >>> in my terminal's scroll buffer, but want to use the pager for more >> >>> than >> >>> that. This is useful especially if I want to scroll back and see the >> >>> results from a query or two ago. >> >> >> >> +1 >> > >> > +1 from me, too. >> >> me three (as long as the current behaviors are not messed with and the >> new stuff is 'opt in' somehow -- I also use the force quit option to >> less). >> >> > > Being able to use "\pset pager" > > to toggle the capability seems useful. > > Thus I'd suggest establishing a new "pager_minlines" option that if unset > (default) maintains the current behavior but which can have a value (0 to > int_max) where 0 ends up basically doing the same thing as "always" mode for > "pager". Leaving the value blank will cause the option to be unset again > and revert to the current behavior. seems pretty reasonable. merlin
On 11/13/2014 11:41 AM, Andrew Dunstan wrote:
>
> On 11/13/2014 11:09 AM, Tom Lane wrote:
>> Andrew Dunstan <andrew@dunslane.net> writes:
>>> I often get annoyed because psql is a bit too aggressive when it
>>> decides
>>> whether to put output through the pager, and the only way to avoid this
>>> is to turn the pager off (in which case your next query might dump many
>>> thousands of lines to the screen). I'd like a way to be able to specify
>>> a minumum number of lines of output before psql would invoke the pager,
>>> rather than just always using the terminal window size.
>> Are you saying you'd want to set the threshold to *more* than the window
>> height? Why?
>
>
> Because I might be quite happy with 100 or 200 lines I can just scroll
> in my terminal's scroll buffer, but want to use the pager for more
> than that. This is useful especially if I want to scroll back and see
> the results from a query or two ago.
>
>
>
>
This patch shows more or less what I had in mind.
However, there is more work to do. As Tom noted upthread, psql's
calculation of the number of lines is pretty bad. For example, if I do:
\pset pager_min_lines 100
select * from generate_series(1,50);
the pager still gets invoked, which is unfortunate to say the least.
So I'm going to take a peek at that.
cheers
andrew
Attachment
Andrew Dunstan wrote: > However, there is more work to do. As Tom noted upthread, psql's calculation > of the number of lines is pretty bad. For example, if I do: > > \pset pager_min_lines 100 > select * from generate_series(1,50); > > the pager still gets invoked, which is unfortunate to say the least. > > So I'm going to take a peek at that. Any luck there? -- Álvaro Herrera http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
On 12/04/2014 03:53 PM, Alvaro Herrera wrote: > Andrew Dunstan wrote: > >> However, there is more work to do. As Tom noted upthread, psql's calculation >> of the number of lines is pretty bad. For example, if I do: >> >> \pset pager_min_lines 100 >> select * from generate_series(1,50); >> >> the pager still gets invoked, which is unfortunate to say the least. >> >> So I'm going to take a peek at that. > Any luck there? > Some, yes, See <http://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=4077fb4d1d34ad04dfb95ba676c2b43ea1f1da53> cheers andrew
On 11/15/2014 05:56 PM, Andrew Dunstan wrote: > > On 11/13/2014 11:41 AM, Andrew Dunstan wrote: >> >> On 11/13/2014 11:09 AM, Tom Lane wrote: >>> Andrew Dunstan <andrew@dunslane.net> writes: >>>> I often get annoyed because psql is a bit too aggressive when it >>>> decides >>>> whether to put output through the pager, and the only way to avoid >>>> this >>>> is to turn the pager off (in which case your next query might dump >>>> many >>>> thousands of lines to the screen). I'd like a way to be able to >>>> specify >>>> a minumum number of lines of output before psql would invoke the >>>> pager, >>>> rather than just always using the terminal window size. >>> Are you saying you'd want to set the threshold to *more* than the >>> window >>> height? Why? >> >> >> Because I might be quite happy with 100 or 200 lines I can just >> scroll in my terminal's scroll buffer, but want to use the pager for >> more than that. This is useful especially if I want to scroll back >> and see the results from a query or two ago. >> >> >> >> > > > This patch shows more or less what I had in mind. > > However, there is more work to do. As Tom noted upthread, psql's > calculation of the number of lines is pretty bad. For example, if I do: > > \pset pager_min_lines 100 > select * from generate_series(1,50); > > the pager still gets invoked, which is unfortunate to say the least. > > So I'm going to take a peek at that. > The over-eager invocation of the pager due to double counting of lines got fixed recently, so here's a slightly updated patch for a pager_min_lines setting, including docco. cheers andrew
Attachment
On 12/21/2014 02:22 PM, Andrew Dunstan wrote: > > On 11/15/2014 05:56 PM, Andrew Dunstan wrote: >> >> On 11/13/2014 11:41 AM, Andrew Dunstan wrote: >>> >>> On 11/13/2014 11:09 AM, Tom Lane wrote: >>>> Andrew Dunstan <andrew@dunslane.net> writes: >>>>> I often get annoyed because psql is a bit too aggressive when it >>>>> decides >>>>> whether to put output through the pager, and the only way to avoid >>>>> this >>>>> is to turn the pager off (in which case your next query might dump >>>>> many >>>>> thousands of lines to the screen). I'd like a way to be able to >>>>> specify >>>>> a minumum number of lines of output before psql would invoke the >>>>> pager, >>>>> rather than just always using the terminal window size. >>>> Are you saying you'd want to set the threshold to *more* than the >>>> window >>>> height? Why? >>> >>> >>> Because I might be quite happy with 100 or 200 lines I can just >>> scroll in my terminal's scroll buffer, but want to use the pager for >>> more than that. This is useful especially if I want to scroll back >>> and see the results from a query or two ago. >>> >>> >>> >>> >> >> >> This patch shows more or less what I had in mind. >> >> However, there is more work to do. As Tom noted upthread, psql's >> calculation of the number of lines is pretty bad. For example, if I do: >> >> \pset pager_min_lines 100 >> select * from generate_series(1,50); >> >> the pager still gets invoked, which is unfortunate to say the least. >> >> So I'm going to take a peek at that. >> > > > The over-eager invocation of the pager due to double counting of lines > got fixed recently, so here's a slightly updated patch for a > pager_min_lines setting, including docco. > > The assigned reviewer hasn't done a review and hasn't responded to email. If there are no other comments I propose to commit this shortly. cheers andrew
On 2014-12-21 14:22:10 -0500, Andrew Dunstan wrote:
> @@ -301,11 +301,11 @@ slashUsage(unsigned short int pager)
> * show list of available variables (options) from command line
> */
> void
> -helpVariables(unsigned short int pager)
> +helpVariables(unsigned short int pager, int pager_min_lines)
> {
> FILE *output;
Odd space before pager_min_lines.
Without having actually tried it, it looks clean enough to me. If
there's more pager options we might at some point introduce a pager
options struct instead adding more options to PageOutput. But for now it
seems ok enough.
Greetings,
Andres Freund
Andres Freund <andres@2ndquadrant.com> writes:
> Without having actually tried it, it looks clean enough to me. If
> there's more pager options we might at some point introduce a pager
> options struct instead adding more options to PageOutput. But for now it
> seems ok enough.
My reaction is that now would be the time to do that, really. This is
messing up the argument lists of what seems like a whole lot of functions,
and I have little faith that this is the last argument we'll need to add.
regards, tom lane
On 03/26/2015 11:10 AM, Tom Lane wrote: > Andres Freund <andres@2ndquadrant.com> writes: >> Without having actually tried it, it looks clean enough to me. If >> there's more pager options we might at some point introduce a pager >> options struct instead adding more options to PageOutput. But for now it >> seems ok enough. > My reaction is that now would be the time to do that, really. This is > messing up the argument lists of what seems like a whole lot of functions, > and I have little faith that this is the last argument we'll need to add. > > *grumble* OK , I'll take a look. cheers andrew
On 03/26/2015 11:10 AM, Tom Lane wrote: > Andres Freund <andres@2ndquadrant.com> writes: >> Without having actually tried it, it looks clean enough to me. If >> there's more pager options we might at some point introduce a pager >> options struct instead adding more options to PageOutput. But for now it >> seems ok enough. > My reaction is that now would be the time to do that, really. This is > messing up the argument lists of what seems like a whole lot of functions, > and I have little faith that this is the last argument we'll need to add. > > OK, this version only changes the signature of one function: PageOutput(), which instead of taking just the pager flag takes a pointer to the whole printTableOpt structure that contains both the pager and pager_min_lines settings (NULL means don't use the pager). That makes the patch smaller and less invasive, and a bit more future-proof. cheers andrew