R语言杂乱dataframe分块标记行提取为新增标识符列的实现方法

现有一个格式杂乱的dataframe，结构为多个观测块循环出现，每个观测块的最后一行是对应块的标记行。该数据来自不可修改/重编程的设备输出，已从存储杂乱字符串的val列中提取出多类字段。
需求为通过匹配标记行中的PD字段识别标记行，将每个块的标记行内容作为标识符存入新增的part列，最终仅保留各块的有效观测行，实现逻辑不受每块观测数量、块总数量限制。

输入示例数据 #

val<-c("Settle0.90s10.0073", "Settle0.80s10.0070", "Settle0.70s10.0067","   
10/22/2110:51:4100PROG 0        PD    Pass        0.0150",
       "Settle0.90s10.0073", "Settle0.80s10.0070", "Settle0.70s10.0067",
       "10/22/2110:52:0300PROG 0        PD    Pass        0.0086")
type<-c("Settle", "Settle", "Settle",NA,"Settle", "Settle", "Settle",NA)
time_sec<-c(0.90,0.80,0.70,"10/22",0.90,0.80,0.70,"10/22")
sec<-c("s","s","s",NA,"s","s","s",NA)
press_psi<-c(10.0073,10.0070, 10.0067,NA,10.0073,10.0070, 10.0067,NA)

df<-data.frame(val,type,time_sec,press_psi)

期望输出效果 #

val<-c("Settle0.90s10.0073", "Settle0.80s10.0070", "Settle0.70s10.0067", 
       "Settle0.90s10.0073", "Settle0.80s10.0070", "Settle0.70s10.0067")
type<-c("Settle", "Settle", "Settle","Settle", "Settle", "Settle")
time_sec<-c(0.90,0.80,0.70,0.90,0.80,0.70)
sec<-c("s","s","s","s","s","s")
press_psi<-c(10.0073,10.0070, 10.0067,10.0073,10.0070, 10.0067)
part<-c("10/22/2110:51:4100PROG 0        PD    Pass        0.0150",
        "10/22/2110:51:4100PROG 0        PD    Pass        0.0150",
        "10/22/2110:51:4100PROG 0        PD    Pass        0.0150",
        "10/22/2110:52:0300PROG 0        PD    Pass        0.0086",
        "10/22/2110:52:0300PROG 0        PD    Pass        0.0086",
        "10/22/2110:52:0300PROG 0        PD    Pass        0.0086")
df<-data.frame(val,type,time_sec,press_psi,part)

现有实现问题 #

已写代码可拆分出各块的有效观测：

lapply(split(df, cumsum(grepl("PD", df$val))), tail, -1)

但需要手动为输出列表中每个dataframe新增part列后再合并，希望有更简便的实现方式。

方案1：tidyverse实现（最简洁） #

通过向上填充的逻辑直接匹配标记值，无需拆分合并：

library(dplyr)
library(tidyr)

result <- df %>%
  # 仅标记行赋值part内容，去除多余空白符
  mutate(part = ifelse(grepl("PD", val), trimws(val), NA_character_)) %>%
  # 从下往上填充NA，每个块的有效观测自动匹配对应块的标记值
  fill(part, .direction = "up") %>%
  # 过滤掉标记行，仅保留有效观测
  filter(!grepl("PD", val))

方案2：base R实现 #

无需依赖第三方包：

# 定位所有标记行，提取对应part值
pd_pos <- grep("PD", df$val)
pd_val <- trimws(df$val[pd_pos])
# 为每行分配所属块编号
df$block_id <- findInterval(seq(nrow(df)), pd_pos, rightmost.closed = TRUE)
# 匹配part值
df$part <- pd_val[df$block_id]
# 移除标记行和辅助列，得到最终结果
result <- df[-pd_pos, setdiff(names(df), "block_id")]

两种方案均完全适配任意块长度、任意块总数的场景，仅靠匹配PD识别标记行，无需手动处理列表拆分合并逻辑。

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