为何dplyr的arrange对apply生成的字符型列报错，仅因子型生效？

The Problem Recap #

You built a tibble with tidyverse, generated the Status column using mutate() combined with apply(), and noticed that calling arrange(Status) throws this error:

Error in arrange_impl(.data, dots) : Argument 1 is of unsupported type matrix

But converting Status to a factor fixes the issue, even though you’ve never had problems using arrange() with character columns before. You’re wondering if apply() is the culprit here—and you’re absolutely right.

What’s Actually Happening #

The key issue is that your Status column isn’t actually a character vector—it’s a matrix in disguise. Here’s why:

1. In the tidyverse pipeline, . refers to a tibble (tbl_df). When you use .[, 2:7], you’re extracting columns 2-7 as another tibble, not a base R data frame or matrix.
2. When apply() receives a tibble, it first converts it to a matrix. For row-wise sums (MARGIN = 1), apply() returns a column matrix (not a plain vector) when working with a matrix input.
3. When you combine this matrix with the logical vector from .[, 8] > 0, R’s recycling rules turn the result into another matrix.
4. Finally, ifelse() takes this matrix input and returns a matrix output—so your Status column is a 6x1 matrix, not a character vector.

You can confirm this by running class(df$Status)—it’ll return matrix, not character. That’s why arrange() throws an error: it doesn’t support sorting by matrix columns. When you convert Status to a factor, as.factor() automatically flattens the matrix into a vector, which arrange() can handle.

The Fix: Ditch apply() for Tidyverse-Friendly Row Operations #

Instead of using base R’s apply() (which plays poorly with tibbles), use tidyverse functions designed for row-wise calculations:

Option 1: Use rowSums() with across()

This is the most efficient approach for row sums:

library(tidyverse)

df <- df %>%
  mutate(
    # Calculate row sum of columns 2-7
    period_sum = rowSums(across(2:7)),
    # Generate Status as a proper character vector
    Status = ifelse(period_sum > 0 & `2018-04` > 0, "NOK", "OK")
  ) %>%
  select(-period_sum) # Clean up the helper column

Option 2: Use rowwise() + c_across()

If you need more flexibility (e.g., non-sum row-wise calculations), use rowwise():

df <- df %>%
  rowwise() %>%
  mutate(
    period_sum = sum(c_across(2:7)),
    Status = ifelse(period_sum > 0 & `2018-04` > 0, "NOK", "OK")
  ) %>%
  ungroup() %>% # Important to exit row-wise mode
  select(-period_sum)

Either way, the Status column will now be a plain character vector, and arrange(Status) will work exactly as you expect—no need to convert to a factor.

Why This Works #

Tidyverse functions like across() and c_across() are designed to work with tibbles directly, returning vectors instead of matrices. This keeps your columns in the proper format for downstream dplyr operations like arrange().

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