如何对数据框指定列单独应用readr::type_convert函数？

Great question! The issue with your initial attempts using map_at and sapply with readr::type_convert is that this function is designed to operate on entire data frames, not individual columns. When you pass a single column (a vector) to it, it throws an error because it expects a data frame input.

Here are a few clean ways to selectively apply type conversion to specific columns while keeping others as character:

1. Use dplyr::across with base::type.convert (simplest for most cases) #

Base R's type.convert works on individual vectors (unlike readr::type_convert), so pairing it with dplyr::across lets you target specific columns directly in your data frame pipeline:

library(tidyverse)

# Original data frame
df <- data.frame(A = letters[1:10], B = as.character(1:10), C = as.character(1:10))

# Convert only column B
df_converted <- df %>%
  mutate(across(B, type.convert))

# Check the result
str(df_converted)
# 'data.frame':	10 obs. of  3 variables:
#  $ A: chr  "a" "b" "c" "d" ...
#  $ B: int  1 2 3 4 5 6 7 8 9 10
#  $ C: chr  "1" "2" "3" "4" ...

For multiple columns, just pass a vector of column names or a selection helper (like starts_with("B")):

# Convert columns B and C
df %>%
  mutate(across(c(B, C), type.convert))

2. Use purrr::map_at with base::type.convert #

If you prefer purrr-style mapping, you can apply type.convert to selected columns and bind the results back into a data frame:

df %>%
  map_at("B", type.convert) %>%
  bind_cols()

3. Use readr::type_convert with explicit column types #

If you want to stick with readr::type_convert, you can define a column type specification that keeps non-target columns as character. This is useful if you need fine-grained control over the exact types:

# Define which columns to convert (others stay character)
col_spec <- cols(
  A = col_character(),
  B = col_guess(),  # Let readr guess the type of B
  C = col_character()
)

df %>% type_convert(col_types = col_spec)

For large data frames with many columns, you can dynamically build the specification to avoid listing every column:

cols_to_convert <- c("B")
col_spec <- cols(.default = col_character())  # Default all to character
col_spec[cols_to_convert] <- list(col_guess())  # Override target columns

df %>% type_convert(col_types = col_spec)

Each of these approaches will let you convert only the columns you want, while preserving the character type of the rest.

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