11:06 a.m.
Dear Allin
I'm calculating the Gini index on a very huge dataset.
While in the command line the operation hardly takes a couple of
seconds, through the GUI (Variable -> Gini coeff) it lasts quite long
and gretl crashes with the following message error. Well, maybe that's a
gnuplot issue I guess!
Best,
Artur
4:37 p.m.
On Sat, 8 Oct 2011, artur bala wrote:
I'm calculating the Gini index on a very huge dataset.
While in the command line the operation hardly takes a couple of seconds,
through the GUI (Variable -> Gini coeff) it lasts quite long and gretl
crashes with the following message error.
It doesn't actually crash, does it? Just fails to display the
Lorenz curve graph? That's fixed in CVS: we down-sample if
there are too many points for gnuplot to handle.
Allin
5:35 p.m.
It doesn't actually crash, does it? Just fails to display the
Lorenz
curve graph?
well, it did crash...without displaying the graph!
That's fixed in CVS: we down-sample if
there are too many points for gnuplot to handle.
That's good! With the new CVS
gretl doesn't crash anymore but fails to
show the graph (neither the text window) and produces the same error
message :-[
Best,
artur
5:46 p.m.
On Sat, 8 Oct 2011, artur bala wrote:
> It doesn't actually crash, does it? Just fails to display the
Lorenz
> curve graph?
well, it did crash...without displaying the graph!
OK, it didn't crash on Linux.
> That's fixed in CVS: we down-sample if
> there are too many points for gnuplot to handle.
That's good! With the new CVS gretl doesn't crash anymore but fails to
show the graph (neither the text window) and produces the same error
message :-[
The update was only in the CVS code repository, not in the
snapshot files -- but the snapshots are now updated.
Allin
7:05 p.m.
> That's good! With the new CVS gretl doesn't crash
anymore
The update was only in the CVS code repository, not in the
snapshot files
So why did it stop to crash ?!?! Mmmmmm! :-\
-- but the snapshots are now updated.
Anyway, now it works
perfectly and very fast: it takes some 6 seconds
for 10 million obs!
Thank you Allin!
Best,
Artur
6 a.m.
Am 08.10.2011 18:37, schrieb Allin Cottrell:
On Sat, 8 Oct 2011, artur bala wrote:
> I'm calculating the Gini index on a very huge dataset.
> While in the command line the operation hardly takes a couple of seconds,
> through the GUI (Variable -> Gini coeff) it lasts quite long and gretl
> crashes with the following message error.
It doesn't actually crash, does it? Just fails to display the
Lorenz curve graph? That's fixed in CVS: we down-sample if
there are too many points for gnuplot to handle.
You probably thought about that yourself, but are there other contexts
where gretl could give gnuplot more than it can handle?
cheers,
sven
4 p.m.
On Sun, 9 Oct 2011, Sven Schreiber wrote:
Am 08.10.2011 18:37, schrieb Allin Cottrell:
> On Sat, 8 Oct 2011, artur bala wrote:
>
>> I'm calculating the Gini index on a very huge dataset.
>> While in the command line the operation hardly takes a couple of seconds,
>> through the GUI (Variable -> Gini coeff) it lasts quite long and gretl
>> crashes with the following message error.
>
> It doesn't actually crash, does it? Just fails to display the
> Lorenz curve graph? That's fixed in CVS: we down-sample if
> there are too many points for gnuplot to handle.
>
You probably thought about that yourself, but are there other contexts
where gretl could give gnuplot more than it can handle?
Yes, that occurred to me too, but I haven't yet tried to do anything
about it.
Allin
4:48 p.m.
On Sun, 9 Oct 2011, Allin Cottrell wrote:
> You probably thought about that yourself, but are there other
contexts
> where gretl could give gnuplot more than it can handle?
Yes, that occurred to me too, but I haven't yet tried to do anything
about it.
And I'd be tempted to add: time is on our side. Even if there were any,
we've been using gnuplot for long enough that I'm confident that they
ought to be pretty unlikely in real life, and probably easily fixable.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
8:18 p.m.
Am 09.10.2011 18:48, schrieb Riccardo (Jack) Lucchetti:
On Sun, 9 Oct 2011, Allin Cottrell wrote:
>> You probably thought about that yourself, but are there other contexts
>> where gretl could give gnuplot more than it can handle?
>
> Yes, that occurred to me too, but I haven't yet tried to do anything
> about it.
And I'd be tempted to add: time is on our side. Even if there were any,
we've been using gnuplot for long enough that I'm confident that they
ought to be pretty unlikely in real life, and probably easily fixable.
I'm not on Windows right now, otherwise I would try to provoke the crash
by simply throwing many many data at gnuplot (from within gretl) to
produce a scatterplot. And if it were possible to replicate the crash, I
would say it should be prevented. But of course you're right, this isn't
terribly important.
cheers,
sven
10:08 p.m.
Le 09/10/2011 21:18, Sven Schreiber a écrit :
Am 09.10.2011 18:48, schrieb Riccardo (Jack) Lucchetti:
> On Sun, 9 Oct 2011, Allin Cottrell wrote:
>
>>> You probably thought about that yourself, but are there other contexts
>>> where gretl could give gnuplot more than it can handle?
>> Yes, that occurred to me too, but I haven't yet tried to do anything
>> about it.
> And I'd be tempted to add: time is on our side. Even if there were any,
> we've been using gnuplot for long enough that I'm confident that they
> ought to be pretty unlikely in real life, and probably easily fixable.
>
I'm not on Windows right now, otherwise I would try to provoke the crash
by simply throwing many many data at gnuplot (from within gretl) to
produce a scatterplot.
Well, I did try with the same XXL dataset which did work
fine for the
Lorenz curve...
And if it were possible to replicate the crash, I
would say it should be prevented.
...and gretl didn't crash although it failed
to show the scatterplot
producing the same gnuplot's error message.
But of course you're right, this isn't
terribly important.
you're right, it isn't! I dont' think that handling
a 10-million obs
dataset is of every day use :)
Best,
Artur
6:33 a.m.
On Sun, 9 Oct 2011, Sven Schreiber wrote:
Am 09.10.2011 18:48, schrieb Riccardo (Jack) Lucchetti:
> On Sun, 9 Oct 2011, Allin Cottrell wrote:
>
>>> You probably thought about that yourself, but are there other contexts
>>> where gretl could give gnuplot more than it can handle?
>>
>> Yes, that occurred to me too, but I haven't yet tried to do anything
>> about it.
>
> And I'd be tempted to add: time is on our side. Even if there were any,
> we've been using gnuplot for long enough that I'm confident that they
> ought to be pretty unlikely in real life, and probably easily fixable.
>
I'm not on Windows right now, otherwise I would try to provoke the crash
by simply throwing many many data at gnuplot (from within gretl) to
produce a scatterplot. And if it were possible to replicate the crash, I
would say it should be prevented. But of course you're right, this isn't
terribly important.
Besides, if you had a systematic way to provoke a crash in gnuplot, that
would be a gnuplot bug, not a gretl bug. Of course we could (perhaps
should) do all our best to help them fix it, but division of labour is a
good thing.
Riccardo (Jack) Lucchetti
Dipartimento di Economia
Università Politecnica delle Marche
r.lucchetti(a)univpm.it
http://www.econ.univpm.it/lucchetti
5:42 p.m.
Am 10.10.2011 08:33, schrieb Riccardo (Jack) Lucchetti:
Besides, if you had a systematic way to provoke a crash in gnuplot, that
would be a gnuplot bug, not a gretl bug. Of course we could (perhaps
should) do all our best to help them fix it, but division of labour is a
good thing.
good point -- but doesn't this then apply to the case of the OP?
-sven
4827
days inactive
4829
days old
11 comments
4 participants
participants (4)
-
Allin Cottrell -
artur bala -
Riccardo (Jack) Lucchetti -
Sven Schreiber