数据集含非分隔符冒号致tidyverse分隔报错，求解决方案

问题背景 #

原本使用冒号作为分隔符拆分GO注释数据，但数据集中存在非分隔用途的冒号（例如UTP:process2），运行以下tidyverse代码时触发报错：

library(tidyverse)

df <- tribble(
  ~Name, ~Value, ~GOs,
  "Ab1", "1000", "F: f1; F: f2; P: p1",
  "Bb1", "2000", "P: p1; F: f1; F: UTP:process2", # 含额外冒号的测试条目
  "Cb1", "3000", "C: c1; F: f1"
)

df |>
  separate_longer_delim(GOs, delim = ";") |>
  separate_wider_delim(GOs, delim = ":", names_sep = ":") |>
  mutate(across(everything(), str_squish)) |>
  pivot_wider(names_from = `GOs:1`, values_from = `GOs:2`) |>
  unnest(c("F", "P", "C"))

报错信息：

Error in separate_wider_delim():
! Expected 3 pieces in each element of GOs.
! 9 values were too short.

由于数据集规模极大，无法逐个定位异常条目，需要调整代码适配这类情况。

解决方案 #

方法1：利用separate_wider_delim的too_few参数 #

separate_wider_delim提供了too_few参数，可指定拆分片段数不足时的处理逻辑。我们只需要拆分第一个冒号，因此设置too_few = "align_start"，让后续的冒号保留在第二个字段中：

df |>
  separate_longer_delim(GOs, delim = ";") |>
  separate_wider_delim(
    GOs,
    delim = ":",
    names = c("GO_type", "GO_term"),
    too_few = "align_start"  # 不足时从开头对齐，剩余内容归入最后一个字段
  ) |>
  mutate(across(c(GO_type, GO_term), str_squish)) |>
  pivot_wider(names_from = GO_type, values_from = GO_term) |>
  unnest(c(F, P, C), keep_empty = TRUE)

方法2：使用正则表达式拆分第一个冒号 #

如果更习惯用正则，可通过separate_wider_regex匹配第一个冒号，把后续所有内容（包括额外冒号）归入第二个字段：

df |>
  separate_longer_delim(GOs, delim = ";") |>
  separate_wider_regex(
    GOs,
    patterns = c(GO_type = "^[^:]+", ": ", GO_term = ".*")
  ) |>
  pivot_wider(names_from = GO_type, values_from = GO_term) |>
  unnest(c(F, P, C), keep_empty = TRUE)

两种方法都能正确处理带额外冒号的条目，比如F: UTP:process2会被拆分为GO_type = "F"，GO_term = "UTP:process2"，不会因额外冒号导致拆分错误。

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