Theoretically possible, but it is going to be easiest to just supply the original data (like you do when you run identify).

You can look at the output of the plot2script function in the TeachingDemos package to see the idea. That function creates a script which will recreate the current plot (base graphics), within the script are the coordinates of the points, but also a bunch of other stuff, so you would need to parse all of that to find just the part that you want, then do what you did before.

What conditions are you in that you have the plot but not easy access to the data points plotted?

Gregory (Greg) L. Snow Ph.D.

> -----Original Message-----

> From: Jannis [mailto:

[hidden email]]

> Sent: Saturday, March 05, 2011 11:54 AM

> To: Greg Snow

> Cc: Dieter Menne;

[hidden email]
> Subject: Re: [R] retrieve x y coordinates of points in current plot

>

> On 03/05/2011 07:25 PM, Greg Snow wrote:

> > It is not completely clear what you are trying to accomplish. Do you

> want to draw a shape in the plot then identify all the points in that

> shape? You could use locator (with type='l') to draw a polygon, then

> there are functions in add on packages (mostly the spatial ones) that

> will detect which points are within a polygon that you could use with

> the raw data and the polygon created.

> >

> > If that is not what you want, then maybe describe your goals in more

> detail (examples are good if you can give one).

> >

> Thats exactly what I want. drawing a polygon in a plot and searching

> for

> the points inside the polygon. I managed to create that polygon and to

> check which points are inside but only by supplying my function with

> the

> coordinates of the points. Now I was wondering whether it is also

> possible to retrieve these coordinates from the plot (similar to

> par()$usr and similar...).

>

>

> ideally it would work as follows:

>

> x<-rnorm(20)

> y=rnorm(20)

> plot(x,y)

> points.in.poly <- identify.poly() #see below

> #now click on the plot to identify the points

>

> Right now it only works like

> points.in.poly <- identify.poly(x,y)

>

> Anyway, supplying the points is not too complicated, it would just be

> easier to do without.

>

> identify.poly <- function(x,y,col.points='red')

> {

> require(sp)

> exit=FALSE

> i=0

> coords.all <- list(x=vector(length=100),y=vector(length=100))

> coords.all$x[1:100]<-NA

> coords.all$y[1:100]<-NA

>

> while (i<100)

> {

> coords.t <- locator(n=1)

> exit=!point.in.polygon(coords.t$x,coords.t$y,

>

> par()$usr[c(1,2,2,1)],par()$usr[c(3,3,4,4)])

> if (exit)

> {

> break

> }

> i=i+1

> points(coords.t,col=col.points,pch='+')

>

> coords.all$x[i] <- coords.t$x

> coords.all$y[i] <- coords.t$y

> if (i>1)

> points(coords.all$x[(i-1):i],coords.all$y[(i-1):i],lty=2,

> col=col.points,type='l')

> }

> points(coords.all$x[c(i,1)],coords.all$y[c(i,1)],lty=2,

> col=col.points,type='l')

> coords.all$x <- na.omit(coords.all$x)

> coords.all$y <- na.omit(coords.all$y)

>

>

> points.inpoly <- point.in.polygon(point.x=x,point.y=y,

>

> pol.x=coords.all$x,pol.y=coords.all$y)

>

> points(x[points.inpoly==1],y[points.inpoly==1],pch=par()$pch,col=col.po

> ints)

> data.return=list(in.poly=!!points.inpoly,x=x,y=y)

> }