data.table滚动计算：基于后续Spuer值调整Eingriff字段逻辑

Got it, let's tackle this problem cleanly and scalably—since you need to handle variable lookahead lengths (like 20+ rows), we can avoid manually chaining lead() calls by leveraging data.table's vectorized shift() function. Here's a solution that works for any lookahead size:

Step 1: Setup Your Data #

First, let's confirm your initial data (I'll redefine it here for clarity):

library(data.table)
DT <- data.table(Zeit = c(1, 2, 3, 4, 5, 6, 7, 8, 9), 
                 Spuer = c(45, 45, 32, 25, 30, 44, 34, 42, 44), 
                 Eingriff = c(0, 0, 1, 0, 0, 0, 1, 0, 0))

Step 2: Define Your Lookahead Window #

Set how many subsequent rows you want to check—change this to 20 or any number for your real data:

lookahead_n <- 3  # Adjust this to your actual required lookahead length

Step 3: Update Eingriff with Vectorized Logic #

We can do this in a single concise operation (no messy temporary columns needed):

DT[, Eingriff := fifelse(
  Eingriff == 1,  # Only target rows where Eingriff was originally 1
  # Check if any of the next `lookahead_n` Spuer values are <30 (ignore NAs for end-of-table rows)
  as.integer(sapply(shift(Spuer, n = 1:lookahead_n, type = "lead"), 
                    function(x) any(x < 30, na.rm = TRUE))),
  0  # Leave non-1 Eingriff values as 0
)]

Step 4: Verify the Result #

Running the code above gives you exactly the output you wanted:

DT
#    Zeit Spuer Eingriff
# 1:    1    45        0
# 2:    2    45        0
# 3:    3    32        1
# 4:    4    25        0
# 5:    5    30        0
# 6:    6    44        0
# 7:    7    34        0
# 8:    8    42        0
# 9:    9    44        0

Why This Works #

• Scalability: Just change lookahead_n to 20 (or any number) and it works instantly—no need to add more lead() calls.
• Efficiency: shift() generates all future values in one go, which is much faster than manual lead() chains for large datasets.
• Robustness: na.rm = TRUE ensures rows near the end of the table (where there aren't enough subsequent values) don't break the check—we just ignore missing values.

内容的提问来源于stack exchange，提问作者Bolle