问题描述
我想分析网站上的一些货币交易数据，但这些数据只能通过复制粘贴来访问。我将它复制到我电脑的剪贴板并通过以下方式将它导入到 R 中：
#df <- read.table("clipboard",header = FALSE,sep = "\t",stringsAsFactors = FALSE,na.strings = "",fill = TRUE)
但是当数据框被读入 R 时，它会将单个观察结果放入四行：
df <- structure(list(V1 = c("Buy","Completed","Fee1.00 USD","Total199.00 USD","Buy","Fee0.50 USD","Total100.00 USD","Fee0.64 USD","Total127.00 USD"),V2 = c(NA,"2021-02-11 20:49:19",NA,"2021-02-11 20:48:03","2021-02-11 20:47:22",NA),V3 = c(NA,"0.11057","82.146","30.15",NA)),row.names = c(NA,12L),class = "data.frame")
df
# V1 V2 V3
#1 Buy <NA> <NA>
#2 Completed 2021-02-11 20:49:19 0.11057
#3 Fee1.00 USD <NA> <NA>
#4 Total199.00 USD <NA> <NA>
#5 Buy <NA> <NA>
#6 Completed 2021-02-11 20:48:03 82.146
#7 Fee0.50 USD <NA> <NA>
#8 Total100.00 USD <NA> <NA>
#9 Buy <NA> <NA>
#10 Completed 2021-02-11 20:47:22 30.15
#11 Fee0.64 USD <NA> <NA>
#12 Total127.00 USD <NA> <NA>
因此，我想将每一系列的四行折叠成一个，像这样，覆盖作为数据导入过程的怪癖生成的缺失值：
want <- structure(list(V1 = structure(c(1L,1L,1L),.Label = "Buy",class = "factor"),V2 = structure(c(1L,.Label = "Completed",V3 = structure(3:1,.Label = c("2/11/2021 20:47","2/11/2021 20:48","2/11/2021 20:49"),V4 = c(0.11057,82.146,30.15),V5 = structure(c(3L,2L),.Label = c("Fee0.50 USD","Fee1.00 USD"),V6 = structure(c(3L,.Label = c("Total100.00 USD","Total127.00 USD","Total199.00 USD"),class = "factor")),class = "data.frame",-3L))
want
# V1 V2 V3 V4 V5 V6
#1 Buy Completed 2/11/2021 20:49 0.11057 Fee1.00 USD Total199.00 USD
#2 Buy Completed 2/11/2021 20:48 82.14600 Fee0.50 USD Total100.00 USD
#3 Buy Completed 2/11/2021 20:47 30.15000 Fee0.64 USD Total127.00 USD
显然，事情仍然有点混乱，因为我需要将一些字符串拆分为单独的列（例如 df$V5 = "Fee1.00 USD" 将变为 df$Fee = 1.00），但这是一个不同的问题。
我已经尝试添加一个 id 变量，然后从长到宽重新整形，as discussed here，但是通过获取我需要的值（例如“Fee1.00 USD”中的 1.00）并将它们放入会变得更加混乱作为新的列名：
df$id <- gl((nrow(df)/4),4)
reshape(df,timevar = "V1",idvar = "id",direction = "wide")
我已经尝试将数据框拆分为一系列数据框 as discussed here，但我仍然不确定如何折叠每个框并将其重新拼接在一起：
split(df,f = df$id)
将数据转换为正确格式的最佳方法是什么？
解决方法
这个怎么样：
library(dplyr)
library(tidyr)
df <- df %>% mutate(obs = rep(1:(nrow(.)/4),each=4))
df <- df %>%
pivot_longer(-obs,names_to="var",values_to="vals") %>%
na.omit() %>%
group_by(obs) %>%
mutate(col = seq_along(obs)) %>%
select(obs,col,vals) %>%
pivot_wider(names_from="col",names_prefix="V",values_from="vals")
df
# # A tibble: 3 x 7
# # Groups: obs [3]
# obs V1 V2 V3 V4 V5 V6
# <int> <chr> <chr> <chr> <chr> <chr> <chr>
# 1 1 Buy Completed 2021-02-11 20:49:19 0.11057 Fee1.00 USD Total199.00 USD
# 2 2 Buy Completed 2021-02-11 20:48:03 82.146 Fee0.50 USD Total100.00 USD
# 3 3 Buy Completed 2021-02-11 20:47:22 30.15 Fee0.64 USD Total127.00 USD
Dave 的回答完美且简洁。如果由于某种原因有人不能使用外部包，我使用他的脚本作为指导并尝试使用基本 R 复制它：
df$id <- gl((nrow(df) / 4),4)
df <- reshape(df,idvar = "id",v.names = "val",timevar = "var",times = names(df[1:3]),varying = names(df[1:3]),new.row.names = 1:1000,direction = "long")
df <- na.omit(df)
df <- df[order(df$id),]
df$col <- ave(seq_len(nrow(df)),df$id,FUN = seq_along)
df <- subset(df,select = c("id","col","val"))
df <- reshape(df,timevar = "col",direction = "wide")
colnames(df) <- c("id","V1","V2","V5","V6","V3","V4")
varnames <- c("id","V4","V6")
df <- df[,varnames]
df