如何合并dplyr::filter_at结果，实现多变量批量筛选全同值行？

Great question! Let's break down what's causing your unexpected results and walk through a few clean, scalable solutions that work even for dozens of columns.

First, let's clarify why your initial attempt with all_vars(. == 5 | . == 1) didn't work: that code checks if each individual column is either 1 or 5, but it doesn't enforce that all columns in the row are the same value. That's why you ended up with rows where values mix 1 and 5 (like row 2: A=1, B=5, C=5) — each column passes the "1 or 5" check, but they aren't all identical.

Solution 1: Use dplyr's modern across (recommended, since filter_at is soft-deprecated) #

If you're using dplyr 1.0.0+, across is the preferred way to handle operations across multiple columns. Here's how to filter rows where all values in columns A-C are identical (regardless of what that value is):

library(dplyr)

# First, recreate your data frame (fixed the D column to match your expected output)
A <- c(1, 1, 1, 5, 5)
B <- c(2, 5, 1, 5, 5)
C <- c(3, 5, 1, 5, 5)
D <- c("T", "F", "T", "T", "T")
df <- data.frame(A, B, C, D)

# Filter rows where all values in A-C are the same
df %>%
  rowwise() %>%
  filter(all(c_across(A:C) == first(c_across(A:C)))) %>%
  ungroup()

This will give you exactly your desired output:

# A tibble: 3 × 4
      A     B     C D    
  <dbl> <dbl> <dbl> <chr>
1     1     1     1 T    
2     5     5     5 T    
3     5     5     5 T    

If you only want to keep rows where the identical value is 1 or 5 (in case there are other identical values you don't care about), just add an extra filter:

df %>%
  rowwise() %>%
  filter(all(c_across(A:C) == first(c_across(A:C)))) %>%
  ungroup() %>%
  filter(A %in% c(1, 5)) # Since all columns are identical, checking A is enough

Solution 2: A base R-inspired approach (no rowwise needed) #

If you prefer avoiding rowwise (which can be slower for large datasets), you can use rowSums to check if every column in the row matches the first column:

df %>%
  filter(rowSums(across(A:C) == pull(select(., A))) == ncol(select(., A:C)))

Solution 3: Chain multiple filter_at calls (your original idea) #

If you specifically want to stick with filter_at (though it's not the latest dplyr syntax), you can combine multiple conditions with | to select rows that are all 1 or all 5:

df %>%
  filter(
    filter_at(vars(A:C), all_vars(. == 1)) |
    filter_at(vars(A:C), all_vars(. == 5))
  )

This works, but it's less scalable if you need to check for more values later (you'd have to add more filter_at clauses). The first solution is better for flexibility.

Key Takeaway #

The core issue with your initial code was mixing up "each column is X or Y" vs "all columns are the same value". The rowwise + c_across approach is clean, scalable, and works for any number of columns without manually listing each variable.

内容的提问来源于stack exchange，提问作者numeralpotatochips