How to divide a set of overlapping ranges into non-overlapping ranges? but in R - r

Question

Let's say we have two datasets:
assays:
BHID<-c(127,127,127,127,128)
FROM<-c(950,959,960,961,955)
TO<-c(958,960,961,966,969)
Cu<-c(0.3,0.9,2.5,1.2,0.5)
assays<-data.frame(BHID,FROM,TO,Cu)
and litho:
BHID<-c(125,127,127,127)
FROM<-c(940,949,960,962)
TO<-c(949,960,961,969)
ROCK<-c(1,1,2,3)
litho<-data.frame(BHID,FROM,TO,ROCK)
and I want to join the two sets and the results after running the algorithm would be:
BHID FROM TO CU ROCK
125 940 970 - 1
127 949 950 - 1
127 950 958 0.3 1
127 958 959 - 1
127 959 960 0.9 1
127 960 961 2.5 2
127 961 962 1.2 -
127 962 966 1.2 3
127 966 969 - 3
128 955 962 0.5 -

Answer 1

Use merge
merge(assays, litho, all=T)
In essence, all=T is the SQL equivalent for FULL OUTER JOIN. I haven't specified any columns, because in this case merge function will perform the join across the column with same names.

Answer 2

Tough one but the code seems to work. The idea is to first expand each row into many, each representing a one-increment from FROM to TO. After merging, identify contiguous rows and un-expand them... Obviously it is not a very efficient approach so it may or may not work if your real data has very large FROM and TO ranges.
library(plyr)
ASSAYS <- adply(assays, 1, with, {
SEQ <- seq(FROM, TO)
data.frame(BHID,
FROM = head(seq(FROM, TO), -1),
TO = tail(seq(FROM, TO), -1),
Cu)
})
LITHO <- adply(litho, 1, with, {
SEQ <- seq(FROM, TO)
data.frame(BHID,
FROM = head(seq(FROM, TO), -1),
TO = tail(seq(FROM, TO), -1),
ROCK)
})
not.as.previous <- function(x) {
x1 <- head(x, -1)
x2 <- tail(x, -1)
c(TRUE, !is.na(x1) & !is.na(x2) & x1 != x2 |
is.na(x1) & !is.na(x2) |
!is.na(x1) & is.na(x2))
}
MERGED <- merge(ASSAYS, LITHO, all = TRUE)
MERGED <- transform(MERGED,
gp.id = cumsum(not.as.previous(BHID) |
not.as.previous(Cu) |
not.as.previous(ROCK)))
merged <- ddply(MERGED, "gp.id", function(x) {
out <- head(x, 1)
out$TO <- tail(x$TO, 1)
out
})
merged
# BHID FROM TO Cu ROCK gp.id
# 1 125 940 949 NA 1 1
# 2 127 949 950 NA 1 2
# 3 127 950 958 0.3 1 3
# 4 127 958 959 NA 1 4
# 5 127 959 960 0.9 1 5
# 6 127 960 961 2.5 2 6
# 7 127 961 962 1.2 NA 7
# 8 127 962 966 1.2 3 8
# 9 127 966 969 NA 3 9
# 10 128 955 969 0.5 NA 10
Note that the first row is not exactly the same as in your expected output, but I think mine makes more sense.