咨询dplyr::select_处理factor类型列时的异常现象原因

Hey there! Let's unpack that weird behavior you noticed when using dplyr::select_() with factor-type column selections.

First, a quick heads-up: select_() is one of dplyr's older "standard evaluation" functions, and it behaves a bit differently than the modern tidy evaluation version (select()) you might be more familiar with. The core issue here boils down to how R handles factor inputs and what select_() expects from its arguments.

The Root Cause #

When you pass a factor vector to select_() as the list of columns to pick, R doesn't automatically convert that factor to its character-level values (like "a" or "c"). Instead, it coerces the factor to its underlying integer codes—those are the numbers that map each factor value to its position in the factor's level order.

For example, if you created your selection factor like this:

cols_to_pick <- factor(c("a", "c"))

R will set the factor levels to the sorted unique values ("a" then "c"), so the integer codes for "a" is 1 and "c" is 2. When you feed this into select_(), it interprets those integers as column positions, not column names. So instead of picking columns "a" and "c", it picks the 1st and 2nd columns of your data frame—if your 2nd column is "b", that's exactly why you're getting unexpected results!

How to Fix It #

To get the behavior you want (selecting columns by their names stored in a factor), you just need to explicitly convert the factor to a character vector before passing it to select_():

# Convert factor to character first
cols_char <- as.character(cols_to_pick)
# Now use select_() with the character vector
select_(your_data_frame, .dots = cols_char)

Even better, since select_() is deprecated in newer dplyr versions, you should switch to the modern select() function with all_of() for safer, more intuitive column selection:

# Modern approach (works with factors once converted to character)
select(your_data_frame, all_of(as.character(cols_to_pick)))

Why This Doesn't Happen with select() #

The modern select() uses tidy evaluation, which is more user-friendly for handling column selections. It'll throw an error if you pass a factor directly (instead of silently coercing to integers), which helps you catch this type of issue early rather than getting confusing results.

内容的提问来源于stack exchange，提问作者analeigh