如何用tidyverse优化含if-else的for循环以提升运行速度？

Hey there! Let's get your slow data processing loop sorted out. First, let's clarify what your current code is trying to accomplish: you're iterating over each row in my.query$data, checking if the station exists in the first column of dataPRCP, updating the corresponding value in dataPRCP when there's a match, and likely adding a new row when there isn't (since your code cuts off at else { d...).

Why Your For-Loop Is Slow #

R's base for-loops can be sluggish when modifying data frames row-by-row because each modification forces R to copy the entire data frame in memory. Tidyverse tools like dplyr use vectorized operations and optimized internal code to avoid this overhead, making them way faster for these kinds of tasks.

Solution 1: Use dplyr's Row Operations (Best Practice) #

The cleanest and most efficient way to handle this is with dplyr's rows_update() and rows_append() functions (available in dplyr 1.0.0+). These are purpose-built for updating existing rows and adding new ones without looping:

library(dplyr)

# First, prepare your update data (rename the value column to match the target column in dataPRCP)
update_data <- my.query$data %>%
  select(station, target_column = value)  # Replace "target_column" with the actual name of column i+3 in dataPRCP

# Update existing rows where station matches
dataPRCP <- dataPRCP %>%
  rows_update(update_data, by = "station")

# Add new rows for stations not already in dataPRCP
new_stations <- update_data %>%
  filter(!station %in% dataPRCP$station)

dataPRCP <- dataPRCP %>%
  rows_append(new_stations)

If you don't want to name the target column explicitly (and prefer using column index i+3), you can adjust the code like this:

# Rename the value column to match the column name at index i+3
col_name <- colnames(dataPRCP)[i+3]
update_data <- my.query$data %>%
  select(station, !!col_name := value)

# Proceed with update and append as before
dataPRCP <- dataPRCP %>%
  rows_update(update_data, by = "station") %>%
  rows_append(update_data %>% filter(!station %in% dataPRCP$station))

Solution 2: Vectorized Base R (Alternative) #

If you prefer base R but still want to avoid the loop, you can use match() to find positions vectorially:

# Find matching positions between stations
match_indices <- match(my.query$data$station, dataPRCP[,1])

# Update rows where there's a match
dataPRCP[na.omit(match_indices), i+3] <- my.query$data$value[!is.na(match_indices)]

# Add new rows for non-matching stations
new_rows <- my.query$data[is.na(match_indices), ]
if (nrow(new_rows) > 0) {
  dataPRCP <- rbind(dataPRCP, new_rows)
}

Solution 3: Using purrr (If You Really Want To) #

You mentioned looking for a purrr alternative to if-else. While purrr doesn't have a dedicated if-else function, you can combine walk2() (for side effects) with base R's if-else or dplyr's if_else(). Note that this is less efficient than the dplyr row operations because it still iterates, but it's more idiomatic than a base for-loop:

library(purrr)
library(dplyr)

# Use walk2 to iterate over stations and values
walk2(my.query$data$station, my.query$data$value, function(stat, val) {
  if (stat %in% dataPRCP[,1]) {
    # Update existing row
    row_num <- which(dataPRCP[,1] == stat)
    dataPRCP[row_num, i+3] <<- val  # <<- modifies the global dataPRCP (not ideal, but works here)
  } else {
    # Add new row
    new_row <- tibble(station = stat, !!colnames(dataPRCP)[i+3] := val)
    dataPRCP <<- bind_rows(dataPRCP, new_row)
  }
})

⚠️ Note: Using <<- to modify global variables is generally discouraged in tidyverse workflows because it makes code harder to debug and follow. Stick with the dplyr row operations whenever possible.

Final Notes #

• The dplyr row operations are the fastest and most readable option here—they eliminate the loop entirely and leverage optimized C++ under the hood.
• Always avoid modifying data frames row-by-row in loops in R; vectorized operations or tidyverse's table-based functions will almost always be faster.

内容的提问来源于stack exchange，提问作者antR