Re: geom_points data capacity?

[Frans Slothouber]

Hoi,

The following

require(ggplot2)
zz <- data.frame( xx <- runif(n=750000 ), yy <- runif(n=750000) )
ggplot(zz,aes(x=xx,y=yy))+geom_point()

takes about a minute to plot on my machine ( dual core, 4Gig )

But with so many point you might want to use a different kind of plot,
or just plot a sample of data.

If you save such a plot as pdf it will take a very long time to render too.


Frans.

On 13 September 2012 22:15, matdodds <matthe...@hotmail.com> wrote:
> Hi ggplotters,
>
> I'm new to R and ggplot. I'm working in RStudio and trying to scatterplot
> some data from a microarray background correction using ggplot.
>
> I entered the code below
>
> p <- ggplot(data.mas, aes(x=raw.log, y=bgcorr.log, colour=cel))
> p <- p + geom_point(shape=".")
> p
>
> but my machine simply freezes, obviously doing a lot of work but producing
> nothing. Eventually I had to kill RStudio. Is the scale of the data an issue
> here? (there's ~750k rows of data)
>
> summary(p)
> data: cel, raw, bg.correct, bg.only,
> bg.method, raw.log, bgcorr.log,
> bgonly.log [743895x8]
> mapping: x = raw.log, y = bgcorr.log, colour = cel
> faceting: facet_null()
> -----------------------------------
> geom_point: na.rm = FALSE, shape = .
> stat_identity:
> position_identity: (width = NULL, height = NULL)
>
> suggestions, pointers as to how I can get these data plotted would be very
> welcome. please could these be clear-for-a-beginner! thanks
>
> M
>
> --
> You received this message because you are subscribed to the ggplot2 mailing
> list.
> Please provide a reproducible example:
> https://github.com/hadley/devtools/wiki/Reproducibility
>
> To post: email ggp...@googlegroups.com
> To unsubscribe: email ggplot2+u...@googlegroups.com
> More options: http://groups.google.com/group/ggplot2

[Brian Diggs]

On 9/14/2012 6:20 AM, Frans Slothouber wrote:
> Hoi,
>
> The following
>
> require(ggplot2)
> zz <- data.frame( xx <- runif(n=750000 ), yy <- runif(n=750000) )
> ggplot(zz,aes(x=xx,y=yy))+geom_point()
>
> takes about a minute to plot on my machine ( dual core, 4Gig )

I get similar timings:

> system.time(print(ggplot(zz,aes(x=xx,y=yy))+geom_point()))
user system elapsed
18.78 10.31 29.12
> benchplot(ggplot(zz,aes(x=xx,y=yy))+geom_point())
step user.self sys.self elapsed
1 construct 0.05 0.03 0.08
2 build 7.84 0.75 8.62
3 render 4.58 1.00 5.59
4 draw 3.39 8.60 12.02
5 TOTAL 15.86 10.38 26.31

Quad core, 16GB, 64 bit Windows 7, running just Rterm.

You may be having memory issues (though 750,000 * 4 bytes * 2 dimenions
~ 6MB, so I would doubt that), or it may be related to Rstudio (I
believe that Rstudio has its own graphics device, but I may be recalling
that incorrectly).

> But with so many point you might want to use a different kind of plot,
> or just plot a sample of data.
>
> If you save such a plot as pdf it will take a very long time to render too.

Both of these are very good points.

>
> Frans.

>> To post: email ggplot2-/JYPxA39Uh5...@public.gmane.org
>> To unsubscribe: email ggplot2+unsubscribe-/JYPxA39Uh5...@public.gmane.org
>> More options: http://groups.google.com/group/ggplot2
>


--
Brian S. Diggs, PhD
Senior Research Associate, Department of Surgery
Oregon Health & Science University

[Frans Slothouber]

Forgot to specify; the timings I got were under RStudio.
So I don't think RStudio is the problem.

[Brandon Hurr]

system.time(print(ggplot(zz,aes(x=xx,y=yy))+geom_point()))

   user  system elapsed 

 37.061   2.985  40.004 

MacbookAir 1.7Ghz 4GB

Didn't really seem that long considering how many points it's plotting. Definitely look at summarizing the data and then plotting if possible/appropriate. 

[matdodds]

@everyone: many thanks for your suggestions .... I'm trimmed the data size by sampling and hey presto end of issue.
chs
M