Question

I have a data table (dt) that has a datetime column. I also have a separate list of datetimes (L).

How can I return a subset of dt that has +/- N rows for each match of an item in L against the datetime column of dt?

I am interested in doing this generically as possible, so this case is a datetime, but I would also be interested in doing it for factors and integers as well. Ideally this would dedup any overlaps as well.

Basically this would be something like grep 'foo|foo1' -A10 -B10 which translates to “Show lines that match foo or foo1 including 10 lines before and after”.

Answer 1

so as an example

dt <- data.frame (dt = 101 : 200)
L <- c (163, 200, 131)

then the row numbers of your matches are

matches <- which (dt$dt %in% L)

Now say you want from 3 rows before to 1 after:

extrarows <- -3 : 1

Now outer is your friend:

rows <- outer(extrarows, matches, `+`)

is almost what you’re looking for, we just

• need to make sure the entries are all valid row numbers (there could be negative numbers in rows):

  rows <- rows [rows %in% seq_len (nrow (dt))]

• you may also want to get rid of duplicates:

  rows <- unique (rows)

• if you want to preserve the matrix structure of rows, maybe setting invalid indices to NA is better than removing them:

  rows [! rows %in% seq_len (nrow (dt))] <- NA

now you return

dt [rows,]