如何使用函数从数据帧的每一行生成多行，然后 join/merge 它们返回？

Question

我从一个数据框开始，其中每一行都有一个长字符串，代表 1D 中的 2D 环境（我们称之为景观）。在实际情况下，这些大约有 6 个值高和 80 个值长，因此在 1D 中，字符串的长度为 480 个字符。我在示例中缩短了这些。每行还有一个唯一名称，是每个景观的 shorthand 标识符。

我有一个函数可以获取每一行，将字符串分成 6 条，并对每条进行分析。在这个例子中，函数的核心动作是压缩条带并获取压缩长度。此函数生成一个 6 行数据框，我需要将其与原始数据框组合，结果是最终数据框，原始数据框的每 1 行有 6 行。

library(dplyr)
library(tibble)

master_df <- tribble(~land_id, ~land_string,
                     "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab",
                     "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb")

compress_it <- function(txt) {
  len.raw <- sum(nchar(txt))
  len.gz <- length(memCompress(txt, "g"))
  return(list("len_raw" = len.raw,
              "len_gz" = len.gz))
}

get_strip_data <- function(land_id, land_string) {
    with_spaces <- gsub("(.{5})", "\1 ", land_string)
    chars_on_lines <- str_replace_all(with_spaces, pattern = " ", "\n")
    prob_matrix <- read.table(text = chars_on_lines, header=FALSE, sep = " ",
                              stringsAsFactors = FALSE)
    prob_matrix <- mutate(prob_matrix, 
                          land_id = land_id,
                          substr_id = 1:nrow(prob_matrix) )
    prob_matrix <- rename(prob_matrix, land_substring = V1)

    mutate(prob_matrix, new = map(land_substring, compress_it)) %>%
    unnest_wider(c(new))
}

get_strip_data(master_df$land_id[[2]], master_df$land_string[[2]]) # to test the above function

这里是我们进入 pseudocode/klugecode 我想做的事情的地方。

首先，我制作了一个空数据框。

subchunks_df <- 
  tribble(~land_id, ~land_string, ~land_substring, ~substr_id, ~len_raw, ~len_gz,
          "", "", "", NA, NA, NA)

尝试 for 循环：

for ( i in 1:nrow(master_df) ) {
  subchunks_df[i, ] <- get_strip_data(master_df$land_id[[i]], master_df$land_string[[i]])
}

改为尝试映射：

subchunks_df <- mapply(get_strip_data, 
                       land_id = master_df$land_id, 
                       land_string = master_df$land_string)

没有。大方一点，我的尝试是 "close but no cigar."

如果我能得到 subchunks_df 正确的形状，我会 right_join:

final_df <- right_join(master_df, subchunks_df, by = "land_id")

考虑到 master_df 遍历函数，这是所需的输出：

final_df <- 
  tribble(~land_id, ~land_string, ~land_substring, ~substr_id, ~len_raw, ~len_gz, 
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "aaaaa", 1, 5, 11,     
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "aaaaa", 2, 5, 11,     
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "baaaa", 3, 5, 11,     
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "abaaa", 4, 5, 13,     
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "babab", 5, 5, 13,     
          "v1-few_bs", "aaaaaaaaaabaaaaabaaabababaabab", "aabab", 6, 5, 13,
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "aaaaa", 1, 5, 11,        
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "aaaaa", 2, 5, 11,        
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "babba", 3, 5, 13,        
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "bbbab", 4, 5, 13,        
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "babab", 5, 5, 13,        
          "v2-more_bs", "aaaaaaaaaababbabbbabbababaabbb", "aabbb", 6, 5, 13)

一如既往，我很感激 dplyr 和 base R 对此的看法。我不相信我的接收条带和获得压缩长度的功能也很好。我找不到更简单的东西。但最后一英里才是真正的麻烦。

Answer 1

地图功能是apply系列的tidyverse版本。 map_dfr 函数正在使用向量 master_df$land_id 索引。把它想象成一个 for 循环。它为您提供了您正在寻找 right_join 调用的数据框。

library(tidyverse)
subchunks_df <- map_dfr(seq_along(master_df$land_id), function(i){
                     get_strip_data(master_df$land_id[[i]], 
                                    master_df$land_string[[i]])})

final_df <- right_join(master_df, subchunks_df, by = "land_id")

Answer 2

您可以使用 Map 然后将数据与 do.call + rbind

绑定在一起

subchunks_df <- do.call(rbind, Map(get_strip_data, master_df$land_id, 
                                    master_df$land_string))

或者如果你更喜欢tidyverse，你可以使用map2_df

subchunks_df <- purrr::map2_df(master_df$land_id, master_df$land_string, 
                               get_strip_data)

# A tibble: 12 x 5
#   land_substring land_id    substr_id len_raw len_gz
#   <chr>          <chr>          <int>   <int>  <int>
# 1 aaaaa          v1-few_bs          1       5     11
# 2 aaaaa          v1-few_bs          2       5     11
# 3 baaaa          v1-few_bs          3       5     11
# 4 abaaa          v1-few_bs          4       5     13
# 5 babab          v1-few_bs          5       5     13
# 6 aabab          v1-few_bs          6       5     13
# 7 aaaaa          v2-more_bs         1       5     11
# 8 aaaaa          v2-more_bs         2       5     11
# 9 babba          v2-more_bs         3       5     13
#10 bbbab          v2-more_bs         4       5     13
#11 babab          v2-more_bs         5       5     13
#12 aabbb          v2-more_bs         6       5     13

然后 right_join :

final_df <- dplyr::right_join(master_df, subchunks_df, by = "land_id")

如何使用函数从数据帧的每一行生成多行，然后 join/merge 它们返回？

How to generate multiple rows from each single row of a dataframe with a function(s) and then join/merge them back in?

r

dataframe

mapply

dplyr