5:18 p.m.
Hi all,
the following script yields a plot for which the x-range goes far beyond
2020-02-28 when using daily data. I use latest git version on Ubuntu 19.10.
<hansl>
clear
set verbose off
nulldata 40
setobs 7 2020-01-20
delete index
series a = normal()
series b = normal()
list L = dataset
# x-axis shows all the way long up to 2021
scatters L --with-lines --output=display
</hansl>
Thanks,
Artur
12:59 a.m.
On Fri, 28 Feb 2020, Artur Tarassow wrote:
Hi all,
the following script yields a plot for which the x-range goes far beyond
2020-02-28 when using daily data. I use latest git version on Ubuntu 19.10.
<hansl>
clear
set verbose off
nulldata 40
setobs 7 2020-01-20
delete index
series a = normal()
series b = normal()
list L = dataset
# x-axis shows all the way long up to 2021
scatters L --with-lines --output=display
</hansl>
Yes, that's not good. Handling daily data via "scatters" in time
series mode is not so easy. The problem is "fixed" in current git,
but the fix could perhaps be improved: right now what happens is
that the x-axis just gives the 1-based numbers of the observations.
Maybe we could find a way to show dates, but there's not much space
available in "scatters". At least now we don't over-extend the
x-axis.
Allin
3:02 p.m.
Am 01.03.20 um 01:59 schrieb Allin Cottrell:
On Fri, 28 Feb 2020, Artur Tarassow wrote:
> Hi all,
>
> the following script yields a plot for which the x-range goes far
> beyond 2020-02-28 when using daily data. I use latest git version on
> Ubuntu 19.10.
>
> <hansl>
> clear
> set verbose off
>
> nulldata 40
> setobs 7 2020-01-20
> delete index
> series a = normal()
> series b = normal()
>
> list L = dataset
> # x-axis shows all the way long up to 2021
> scatters L --with-lines --output=display
> </hansl>
Yes, that's not good. Handling daily data via "scatters" in time series
mode is not so easy. The problem is "fixed" in current git, but the fix
could perhaps be improved: right now what happens is that the x-axis
just gives the 1-based numbers of the observations.
Thank you, Allin. Looks much better now.
Maybe we could find a way to show dates, but there's not much
space
available in "scatters". At least now we don't over-extend the x-axis.
Maybe its beneficial to rotate the x-labels to save some space. But it
has no priority I guess.
Artur
9:06 a.m.
On Sun, 1 Mar 2020, Artur Tarassow wrote:
> Maybe we could find a way to show dates, but there's not much
space
> available in "scatters". At least now we don't over-extend the x-axis.
Maybe its beneficial to rotate the x-labels to save some space. But it has no
priority I guess.
Something like this, perhaps?
<hansl>
open IT_companies.gdt --frompkg=MGARCH
scatters 2 3 4 5 --output="(a)dotdir/x.gp"
s = readfile("(a)dotdir/x.gp")
# generate daily xtics
ntics = 6
v = round(seq(1, $nobs, $nobs/ntics))
xtics = "set xtics ("
loop i = 1 .. ntics --quiet
vi = v[i]
xtics ~= sprintf("\"%s\" %d", obslabel(vi), vi)
if i<ntics
xtics ~= ", "
endif
endloop
xtics ~= ") rotate by -45"
# modify gnuplot script and plot
s = regsub(s, "set xtics (.*)", xtics)
outfile "(a)dotdir/y.gp" --quiet
print s
end outfile
gnuplot --input="(a)dotdir/y.gp" --output=display
</hansl>
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
9:10 a.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
Am 02.03.2020 um 10:06 schrieb Riccardo (Jack) Lucchetti:
loop i = 1 .. ntics --quiet
I absolutely don't want to hijack this thread, but we write loop ...
--quiet so often that I wonder whether the equivalent of 'setopt loop
persist --quiet' shouldn't be promoted to a user-settable option in the
GUI preferences.
(I guess to override temporarily then loop would need a --verbose option...)
cheers
sven
1:50 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Mon, 2 Mar 2020, Sven Schreiber wrote:
Am 02.03.2020 um 10:06 schrieb Riccardo (Jack) Lucchetti:
> loop i = 1 .. ntics --quiet
I absolutely don't want to hijack this thread, but we write loop ...
--quiet so often that I wonder whether the equivalent of 'setopt loop
persist --quiet' shouldn't be promoted to a user-settable option in the
GUI preferences.
(I guess to override temporarily then loop would need a --verbose option...)
Agreed. We could maybe even countenance mild backward incompatibility:
make "loop" quiet by default and add a --verbose option. If we did
that we'd continue to accept the --quiet option, of course, but it
would do nothing.
Allin
2:27 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Mon, 2 Mar 2020, Allin Cottrell wrote:
On Mon, 2 Mar 2020, Sven Schreiber wrote:
> Am 02.03.2020 um 10:06 schrieb Riccardo (Jack) Lucchetti:
>
>> loop i = 1 .. ntics --quiet
>
> I absolutely don't want to hijack this thread, but we write loop ...
> --quiet so often that I wonder whether the equivalent of 'setopt loop
> persist --quiet' shouldn't be promoted to a user-settable option in the
> GUI preferences.
> (I guess to override temporarily then loop would need a --verbose option...)
Agreed. We could maybe even countenance mild backward incompatibility:
make "loop" quiet by default and add a --verbose option. If we did
that we'd continue to accept the --quiet option, of course, but it
would do nothing.
Hm, I like this
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
8:32 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Mon, 2 Mar 2020, Riccardo (Jack) Lucchetti wrote:
On Mon, 2 Mar 2020, Allin Cottrell wrote:
> On Mon, 2 Mar 2020, Sven Schreiber wrote:
>
>> Am 02.03.2020 um 10:06 schrieb Riccardo (Jack) Lucchetti:
>>
>>> loop i = 1 .. ntics --quiet
>>
>> I absolutely don't want to hijack this thread, but we write loop ...
>> --quiet so often that I wonder whether the equivalent of 'setopt loop
>> persist --quiet' shouldn't be promoted to a user-settable option in the
>> GUI preferences.
>> (I guess to override temporarily then loop would need a --verbose
>> option...)
>
> Agreed. We could maybe even countenance mild backward incompatibility:
> make "loop" quiet by default and add a --verbose option. If we did
> that we'd continue to accept the --quiet option, of course, but it
> would do nothing.
Hm, I like this
Should we just go ahead and do this for the forthcoming release?
"loop" is something that goes way back in gretl history, and the
verbose behavior in the absence of the --quiet flag was pretty much
a debugging thing, perhaps useful when the facility was first
introduced. I'd be very surprised if anyone would actually miss it.
Allin
9 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Mon, 2 Mar 2020, Allin Cottrell wrote:
Should we just go ahead and do this for the forthcoming release?
I'd say let's
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
9:20 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
Am 02.03.2020 um 22:00 schrieb Riccardo (Jack) Lucchetti:
On Mon, 2 Mar 2020, Allin Cottrell wrote:
> Should we just go ahead and do this for the forthcoming release?
I'd say let's
I'm always a bit worried about unintended side effects. For example,
what about special non-standard cases like the progressive variant of
loop, or other such stuff? I don't have any specific suspicion, but I'd
rather let this sink for a little while.
cheers
sven
10:24 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Mon, 2 Mar 2020, Sven Schreiber wrote:
Am 02.03.2020 um 22:00 schrieb Riccardo (Jack) Lucchetti:
> On Mon, 2 Mar 2020, Allin Cottrell wrote:
>
>> Should we just go ahead and do this for the forthcoming release?
>
> I'd say let's [make "loop" quiet by default]
I'm always a bit worried about unintended side effects. For example,
what about special non-standard cases like the progressive variant of
loop, or other such stuff? I don't have any specific suspicion, but I'd
rather let this sink for a little while.
So far as I'm aware, the --progressive and --verbose "loop" options
are orthogonal. But I take your point about unintended consequences.
Maybe we should do this in the git/snapshot stage as we're working
towards gretl 2020b.
Allin
4:47 a.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On 02.03.2020 23:24, Allin Cottrell wrote:
So far as I'm aware, the --progressive and --verbose
"loop" options
are orthogonal. But I take your point about unintended consequences.
Maybe we should do this in the git/snapshot stage as we're working
towards gretl 2020b.
I think so. IMHO it may need a intensive testing for side effects, as
Sven just told.
Marcin
--
Marcin Błażejowski
8:32 a.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
Am 03.03.2020 um 05:47 schrieb Marcin Błażejowski:
On 02.03.2020 23:24, Allin Cottrell wrote:
> So far as I'm aware, the --progressive and --verbose "loop" options
> are orthogonal. But I take your point about unintended consequences.
> Maybe we should do this in the git/snapshot stage as we're working
> towards gretl 2020b.
I think so. IMHO it may need a intensive testing for side effects, as
Sven just told.
I just checked the changelog which doesn't say anything about it. Is
this implemented yet?
thanks
sven
5:41 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Fri, 10 Apr 2020, Sven Schreiber wrote:
Am 03.03.2020 um 05:47 schrieb Marcin Błażejowski:
> On 02.03.2020 23:24, Allin Cottrell wrote:
>> So far as I'm aware, the --progressive and --verbose "loop"
options
>> are orthogonal. But I take your point about unintended consequences.
>> Maybe we should do this in the git/snapshot stage as we're working
>> towards gretl 2020b.
>
> I think so. IMHO it may need a intensive testing for side effects, as
> Sven just told.
I just checked the changelog which doesn't say anything about it. Is
this implemented yet?
No, I haven't done anything about that yet. But the reference to
2020b reminds me -- a new release got kinda pushed off my agenda by,
let's say, various non-gretl goings-on. Should we release soon? Any
particular things needing attention before we do that?
(For my part I'd be happy to release soon, and defer the loop
verbosity issue till 2020c. Reminder: the idea was to make "quiet"
operation of loops the default, and offer a --verbose option to
obtain the mode that's currently the default.)
Allin
9:15 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
Am 10.04.2020 um 19:41 schrieb Allin Cottrell:
On Fri, 10 Apr 2020, Sven Schreiber wrote:
> Am 03.03.2020 um 05:47 schrieb Marcin Błażejowski:
>> I think so. IMHO it may need a intensive testing for side
effects, as
>> Sven just told.
>
> I just checked the changelog which doesn't say anything about it. Is
> this implemented yet?
No, I haven't done anything about that yet. But the reference to 2020b
reminds me -- a new release got kinda pushed off my agenda by, let's
say, various non-gretl goings-on. Should we release soon? Any particular
things needing attention before we do that?
IIRC it is mainly the Mac situation that needs a release? Personally I'm
not affected, but due to the shift to online teaching I'm observing an
increased demand for gretl, and there is a sizeable fraction of students
who use Macbooks. So I guess it would be good.
(For my part I'd be happy to release soon, and defer the loop
verbosity
issue till 2020c. Reminder: the idea was to make "quiet" operation of
loops the default, and offer a --verbose option to
obtain the mode that's currently the default.)
Sounds good.
cheers
sven
9:30 p.m.
New subject: OT: quiet loop (Re: Re: x-range for scatters using daily data)
On Fri, 10 Apr 2020, Sven Schreiber wrote:
IIRC it is mainly the Mac situation that needs a release? Personally
I'm
not affected, but due to the shift to online teaching I'm observing an
increased demand for gretl, and there is a sizeable fraction of students
who use Macbooks. So I guess it would be good.
Yes, the macOS situation is the most pressing reason for a quick
release, and what you report adds some emphasis. So I'll try to get
to it this weekend.
Allin
1:40 p.m.
On Mon, 2 Mar 2020, Riccardo (Jack) Lucchetti wrote:
On Sun, 1 Mar 2020, Artur Tarassow wrote:
> > Maybe we could find a way to show dates, but there's not much space
> > available in "scatters". At least now we don't over-extend the
x-axis.
>
> Maybe its beneficial to rotate the x-labels to save some space. But it has
> no priority I guess.
Something like this, perhaps?
<hansl>
open IT_companies.gdt --frompkg=MGARCH
scatters 2 3 4 5 --output="(a)dotdir/x.gp"
s = readfile("(a)dotdir/x.gp")
# generate daily xtics
ntics = 6
v = round(seq(1, $nobs, $nobs/ntics))
xtics = "set xtics ("
loop i = 1 .. ntics --quiet
vi = v[i]
xtics ~= sprintf("\"%s\" %d", obslabel(vi), vi)
if i<ntics
xtics ~= ", "
endif
endloop
xtics ~= ") rotate by -45"
# modify gnuplot script and plot
s = regsub(s, "set xtics (.*)", xtics)
outfile "(a)dotdir/y.gp" --quiet
print s
end outfile
gnuplot --input="(a)dotdir/y.gp" --output=display
</hansl>
We already do something very much like this for an individual
time-series plot using daily data. Take Artur's example as a basis:
<hansl>
nulldata 40
setobs 7 2020-01-20
series a = normal()
gnuplot a --time --with-lines --output=display
</hansl>
This shows rotated date labels. In the case of scatters, where space
is at a premium, I doubt this will work nicely -- but it might be
worth experimenting.
Allin
9:12 p.m.
Am 02.03.20 um 10:06 schrieb Riccardo (Jack) Lucchetti:
On Sun, 1 Mar 2020, Artur Tarassow wrote:
>> Maybe we could find a way to show dates, but there's not much space
>> available in "scatters". At least now we don't over-extend the
x-axis.
>
> Maybe its beneficial to rotate the x-labels to save some space. But
> it has no priority I guess.
Something like this, perhaps?
<hansl>
open IT_companies.gdt --frompkg=MGARCH
scatters 2 3 4 5 --output="(a)dotdir/x.gp"
s = readfile("(a)dotdir/x.gp")
# generate daily xtics
ntics = 6
v = round(seq(1, $nobs, $nobs/ntics))
xtics = "set xtics ("
loop i = 1 .. ntics --quiet
vi = v[i]
xtics ~= sprintf("\"%s\" %d", obslabel(vi), vi)
if i<ntics
xtics ~= ", "
endif
endloop
xtics ~= ") rotate by -45"
# modify gnuplot script and plot
s = regsub(s, "set xtics (.*)", xtics)
outfile "(a)dotdir/y.gp" --quiet
print s
end outfile
gnuplot --input="(a)dotdir/y.gp" --output=display
</hansl>
I don't get any xtics here :-/ But apart from this: This is a very nice
example of exploiting "scatters" for enhancing multiplots.
Artur
10:30 p.m.
On Mon, 2 Mar 2020, Artur Tarassow wrote:
Am 02.03.20 um 10:06 schrieb Riccardo (Jack) Lucchetti:
> On Sun, 1 Mar 2020, Artur Tarassow wrote:
>
>>> Maybe we could find a way to show dates, but there's not much space
>>> available in "scatters". At least now we don't over-extend the
x-axis.
>>
>> Maybe its beneficial to rotate the x-labels to save some space. But it has
>> no priority I guess.
>
> Something like this, perhaps?
>
> <hansl>
[nice example script]
> </hansl>
I don't get any xtics here :-/ But apart from this: This is a very nice
example of exploiting "scatters" for enhancing multiplots.
Hmm, why aren't you getting any xtics? They're shown here OK.
Anyway, Jack's script might be a good approach to getting dates in
place for "scatters", maybe better than trying to use gnuplot's
"timefmt".
Allin
12:07 p.m.
On Mon, 2 Mar 2020, Allin Cottrell wrote:
On Mon, 2 Mar 2020, Artur Tarassow wrote:
> Am 02.03.20 um 10:06 schrieb Riccardo (Jack) Lucchetti:
>> On Sun, 1 Mar 2020, Artur Tarassow wrote:
>>
>>>> Maybe we could find a way to show dates, but there's not much space
>>>> available in "scatters". At least now we don't over-extend
the x-axis.
>>>
>>> Maybe its beneficial to rotate the x-labels to save some space. But it
>>> has no priority I guess.
>>
>> Something like this, perhaps?
>>
>> <hansl>
[nice example script]
>> </hansl>
>
> I don't get any xtics here :-/ But apart from this: This is a very nice
> example of exploiting "scatters" for enhancing multiplots.
Hmm, why aren't you getting any xtics? They're shown here OK. Anyway, Jack's
script might be a good approach to getting dates in place for "scatters",
maybe better than trying to use gnuplot's "timefmt".
I think I found it. You should get tyhe xtics only using current git. With
2019d, the x scale is in "years", with current git is just an integer
sequence. This version of the script should give you the xtics with 2019d
as well:
<hansl>
open IT_companies.gdt --frompkg=MGARCH
scatters 2 3 4 5 --output="(a)dotdir/x.gp"
s = readfile("(a)dotdir/x.gp")
# generate daily xtics
ntics = 6
v = round(seq(1, $nobs, $nobs/ntics))
compat = $version < 20200
xtics = "set xtics ("
loop i = 1 .. ntics --quiet
vi = v[i]
if compat
# rough!!!
x = $obsmajor[vi] + ($obsminor[vi] - 1)/12 + ($obsmicro[vi] - 1)/365
else
x = vi
endif
xtics ~= sprintf("\"%s\" %8.4f", obslabel(vi), x)
if i<ntics
xtics ~= ", "
endif
endloop
xtics ~= ") rotate by -45"
# modify gnuplot script and plot
s = regsub(s, "set xtics (.*)", xtics)
outfile "(a)dotdir/y.gp" --quiet
print s
end outfile
gnuplot --input="(a)dotdir/y.gp" --output=display
</hansl>
-------------------------------------------------------
Riccardo (Jack) Lucchetti
Dipartimento di Scienze Economiche e Sociali (DiSES)
Università Politecnica delle Marche
(formerly known as Università di Ancona)
r.lucchetti(a)univpm.it
http://www2.econ.univpm.it/servizi/hpp/lucchetti
-------------------------------------------------------
7:41 p.m.
Am 03.03.20 um 13:07 schrieb Riccardo (Jack) Lucchetti:
On Mon, 2 Mar 2020, Allin Cottrell wrote:
> On Mon, 2 Mar 2020, Artur Tarassow wrote:
>
>> Am 02.03.20 um 10:06 schrieb Riccardo (Jack) Lucchetti:
>>> On Sun, 1 Mar 2020, Artur Tarassow wrote:
>>>
>>>>> Maybe we could find a way to show dates, but there's not much
>>>>> space available in "scatters". At least now we don't
over-extend
>>>>> the x-axis.
>>>>
>>>> Maybe its beneficial to rotate the x-labels to save some space.
>>>> But it has no priority I guess.
>>>
>>> Something like this, perhaps?
>>>
>>> <hansl>
> [nice example script]
>>> </hansl>
>>
>> I don't get any xtics here :-/ But apart from this: This is a very
>> nice example of exploiting "scatters" for enhancing multiplots.
>
> Hmm, why aren't you getting any xtics? They're shown here OK. Anyway,
> Jack's script might be a good approach to getting dates in place for
> "scatters", maybe better than trying to use gnuplot's
"timefmt".
I think I found it. You should get tyhe xtics only using current git.
With 2019d, the x scale is in "years", with current git is just an
integer sequence. This version of the script should give you the xtics
with 2019d as well:
Thanks Jack, with a freshly compiled gretl your code works fine!
Artur
1:23 a.m.
On Tue, 3 Mar 2020, Artur Tarassow wrote:
Am 03.03.20 um 13:07 schrieb Riccardo (Jack) Lucchetti:
> On Mon, 2 Mar 2020, Allin Cottrell wrote:
>
>> On Mon, 2 Mar 2020, Artur Tarassow wrote:
>>
>>> Am 02.03.20 um 10:06 schrieb Riccardo (Jack) Lucchetti:
>>>> On Sun, 1 Mar 2020, Artur Tarassow wrote:
>>>>
>>>>>> Maybe we could find a way to show dates, but there's not much
space
>>>>>> available in "scatters". At least now we don't
over-extend the x-axis.
>>>>>
>>>>> Maybe its beneficial to rotate the x-labels to save some space. But
it
>>>>> has no priority I guess.
>>>>
>>>> Something like this, perhaps?
>>>>
>>>> <hansl>
>> [nice example script]
>>>> </hansl>
>>>
>>> I don't get any xtics here :-/ But apart from this: This is a very nice
>>> example of exploiting "scatters" for enhancing multiplots.
>>
>> Hmm, why aren't you getting any xtics? They're shown here OK. Anyway,
>> Jack's script might be a good approach to getting dates in place for
>> "scatters", maybe better than trying to use gnuplot's
"timefmt".
>
> I think I found it. You should get tyhe xtics only using current git. With
> 2019d, the x scale is in "years", with current git is just an integer
> sequence. This version of the script should give you the xtics with 2019d
> as well:
Thanks Jack, with a freshly compiled gretl your code works fine!
Attaching calendar dates to time-series "scatters" with daily data
is now implemented in git. There may be some rough spots!
Allin
1497
days inactive
1539
days old
21 comments
5 participants
participants (5)
-
Allin Cottrell -
Artur Tarassow -
Marcin Błażejowski -
Riccardo (Jack) Lucchetti -
Sven Schreiber