link_knn: k-Nearest neighbor linking
In ggobi/cranvas: Interactive statistical graphics based on Qt

Description Usage Arguments Details Value Author(s) Examples

When a number of elements in a data are brushed, their k-nearest neighbors (based on a certain distance measure) are brushed as well.

1	link_knn(mf1, var1 = NULL, mf2 = NULL, var2 = var1, k = 10)

`var1`	the variable names or column indices of the first mutaframe to be used to calculate distances
`var2`	(optional) variable names or column indices of the second mutaframe (by default the same as `var1`)
`k`	the number of nearest neighbors to select
`mf1`	the first mutaframe
`mf2`	(optional) the second mutaframe; default `NULL` means `mf1` will be linked to itself

A center point for the variables based on the selected rows is calculated in the first dataset, then the k nearest rows in the second dataset (if not provided, it will be the same as the first dataset) to this center are selected. Only the Euclidean distance has been implemented at the moment.

Similar to categorical linking (link_cat), this function also links two mutaframes together (or one mutaframe to itself), and id's of listeners are returned.

Yihui Xie <http://yihui.name>

### (1) linking to oneself
data(flea, package = "tourr")
qflea <- qdata(flea, color = species)
qscatter(tars1, tars2, data = qflea)

## brush the nearest 10 points around the center
id <- link_knn(qflea, c("tars1", "tars2"), k = 10)

## remove the linking
remove_link(qflea, id)


### (2) variables on the same scale
mf <- qdata(data.frame(x = rnorm(300), y = rnorm(300)), size = 3)
qscatter(x, y, data = mf)

id <- link_knn(mf, c("x", "y"), k = 20)

remove_link(mf, id)


### (3) link two datasets

mf1 <- qdata(flea, color = species)
mf2 <- qdata(subset(flea, species == "Concinna "))

qscatter(tars1, tars2, data = mf1)
qscatter(tars1, tars2, data = mf2)

id <- link_knn(mf1, c("tars1", "tars2"), mf2, c("tars1", "tars2"))

remove_link(mf1, id[1])
remove_link(mf2, id[2])

cranvas_off()