Subset a dataframe using start and stop points from another dataframe? - r

Question

I have a dataframe df with 3 columns id, first and last
id <- c(27,27,134,134)
first <- c(14,20,9,16)
last <- c(17,24,13,20)
df <- as.data.frame(cbind(id,first,last))
df
Each row corresponds to a chunk of data from another dataframe that I want to keep.
first and last indicate the first and last frames of the relevant chunk
I want to use this to subset the other dataframe dat which is structured as below
dat_id <- c(rep(27, 30), rep(134,30))
dat_frame <- c(seq(1:30), seq(1:30))
dat_data <- c(sample(1:60))
dat <- as.data.frame(cbind(dat_id,dat_frame,dat_data))
dat
The only way I know to extract the relevant portion is with a for loop as below (this produces the expected output), but I expect this is a horribly inefficient way to do it. What's a better way?
#header row
new_df <- data.frame(id = numeric(), frame = numeric(), data = numeric())
#populate
for (i in (seq (1:nrow(df)))){
new_df <- rbind(new_df, subset(dat, dat_id == df[i,"id"])[df[i,"first"]:df[i,"last"],])
}
new_df

Answer 1

This can be done with a complex join in sql. This avoids creating a large intermediate data frame based on joining only on id and then cutting it down.
library(sqldf)
sqldf("
select dat.*
from dat
join df on dat.dat_id = df.id and
dat.dat_frame between df.first and df.last
")
Update
The example in the question changed and the solution has been simplified assuming the new example.

Answer 2

Using dplyr we can do a left_join on dat and df and select only those rows which lie in between first and last of their respective id.
library(dplyr)
left_join(dat, df, by = c("dat_id" = "id")) %>%
filter(between(dat_frame, first, last)) %>%
select(-first, -last)
Or using the same logic in base R
subset(merge(dat, df, by.x = "dat_id", by.y = "id", all.x = TRUE),
dat_frame >= first & dat_frame <= last)

Answer 3

We can use a non-equi join for this. It would be more faster and efficient
library(data.table)
setDT(dat)[, newcol := dat_frame][df, on = .(dat_id = id,
newcol >= first, newcol <=last)][, .(dat_id, dat_frame, dat_data)]
# dat_id dat_frame dat_data
# 1: 27 14 26
# 2: 27 15 56
# 3: 27 16 30
# 4: 27 17 49
# 5: 27 20 23
# 6: 27 21 37
# 7: 27 22 7
# 8: 27 23 40
# 9: 27 24 12
#10: 134 9 57
#11: 134 10 35
#12: 134 11 31
#13: 134 12 53
#14: 134 13 38
#15: 134 16 15
#16: 134 17 14
#17: 134 18 33
#18: 134 19 54
#19: 134 20 43
Or another option is fuzzyjoin
library(fuzzyjoin)
library(dplyr)
dat %>%
mutate(newcol = dat_frame) %>%
fuzzy_left_join(df, by = c("dat_id" = 'id', 'newcol' = 'first',
'newcol' = 'last'), match_fun = list(`==`, `>=`, `<=`)) %>%
na.omit %>%
select(dat_id, dat_frame, dat_data)
# dat_id dat_frame dat_data
#14 27 14 26
#15 27 15 56
#16 27 16 30
#17 27 17 49
#20 27 20 23
#21 27 21 37
#22 27 22 7
#23 27 23 40
#24 27 24 12
#39 134 9 57
#40 134 10 35
#41 134 11 31
#42 134 12 53
#43 134 13 38
#46 134 16 15
#47 134 17 14
#48 134 18 33
#49 134 19 54
#50 134 20 43
Or using base R
out <- do.call(rbind, Map(function(x, y) do.call(rbind,
Map(function(u, v) subset(x, dat_frame >= u & dat_frame <= v),
y$first, y$last)), split(dat, dat$dat_id), split(df, df$id)))
row.names(out) <- NULL
out
# dat_id dat_frame dat_data
#1 27 14 26
#2 27 15 56
#3 27 16 30
#4 27 17 49
#5 27 20 23
#6 27 21 37
#7 27 22 7
#8 27 23 40
#9 27 24 12
#10 134 9 57
#11 134 10 35
#12 134 11 31
#13 134 12 53
#14 134 13 38
#15 134 16 15
#16 134 17 14
#17 134 18 33
#18 134 19 54
#19 134 20 43
NOTE: All the above solutions work
Also, note that the solution in the other post gives Error
left_join(dat, df, by = c("dat_id" = "id")) %>%
filter(between(dat_frame, first, last)) %>%
select(-first, -last)
#Error: Expecting a single value: [extent=120].
NOTE: That the accepted answer is wrong and it is giving error.