dcast 不将变量类型保留为字符，当变量为 NA 时 vapply 出错答案

【问题标题】：dcast not retaining variable type as character, error in vapply when a variable is NAdcast 不将变量类型保留为字符，当变量为 NA 时 vapply 出错
【发布时间】：2019-04-13 04:09:35
【问题描述】：

在尝试使用resphape2::dcast 重塑数据时，我遇到了涉及NA 条目的错误。样本数据在最后。

数据从长到宽重新调整，但有时某些参数包含所有 NA 条目，这似乎是导致问题的原因。或者至少，我认为是。如果我删除任何类似的参数，在本例中为Ammonia，错误就会消失。

在调试dcast 时，似乎固定到了这一行：

ordered <- vaggregate(.value = value, .group = overall, 
  .fun = fun.aggregate, ..., .default = fill, .n = n)

导致错误：

Error in vapply(indices, fun, .default) : 
values must be type 'character',
but FUN(X[[1]]) result is type 'integer'

看到NA变量排在第一位，我想aggregate函数可能默认为整数，即使整列都是character，但是移动这些行并没有解决。我可以找到solve it is by usingna.omit 的唯一方法，它完全删除了该参数。如果可能，我的预期输出将保留所有 NA 的任何参数。造成这种情况的第二个原因是，如果一天/深度没有被采样，它应该被保留并且这些条目应该是ns（未采样）。 有没有一种方法可以解决这个错误，而无需删除所有将被重塑的NA 参数？

可重现的例子（数据在dcast代码下方）：

library(reshape2)
dcast(df, station + date + depth ~ parmcode, value.var = "value_qualif", fill="ns")

dcast(na.omit(df), station + date + depth ~ parmcode,
 value.var = "value_qualif", fill="ns") # solves error, but removes parameter completely

示例数据：

df <- structure(list(station = c("A", "A", "A", "A", "A", "A", "A", 
"A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", 
"A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", 
"A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", 
"A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", 
"A", "A", "A", "A", "A", "A", "A", "A", "A", "A"), date = c("7/2/2018", 
"7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", 
"7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", 
"7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", 
"7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", 
"7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/2/2018", "7/1/2018", 
"7/1/2018", "7/1/2018", "7/1/2018", "7/1/2018", "7/1/2018", "7/1/2018", 
"7/1/2018", "7/1/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", 
"7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", 
"7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", 
"7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", 
"7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", "7/9/2018", 
"7/9/2018", "7/9/2018"), depth = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 
18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 1L, 1L, 1L, 
12L, 12L, 12L, 18L, 18L, 18L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 12L, 18L, 
18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L, 18L), parmcode = c("CDOM", 
"DENSITY", "DO", "ENTERO", "PH", "TOTAL", "XMS", "TEMP", "SAL", 
"FECAL", "TOTAL", "FECAL", "ENTERO", "CDOM", "XMS", "TEMP", "SAL", 
"PH", "DO", "DENSITY", "DO", "DENSITY", "TOTAL", "FECAL", "PH", 
"CDOM", "XMS", "TEMP", "SAL", "ENTERO", "AMMONIA AS N", "AMMONIA AS N", 
"AMMONIA AS N", "AMMONIA AS N", "AMMONIA AS N", "AMMONIA AS N", 
"AMMONIA AS N", "AMMONIA AS N", "AMMONIA AS N", "TOTAL", "XMS", 
"TEMP", "SAL", "PH", "DO", "DENSITY", "CDOM", "FECAL", "ENTERO", 
"CDOM", "FECAL", "ENTERO", "PH", "DO", "TEMP", "XMS", "TOTAL", 
"DENSITY", "SAL", "TOTAL", "FECAL", "ENTERO", "XMS", "TEMP", 
"SAL", "PH", "DO", "DENSITY", "CDOM"), value_qualif = c(NA, NA, 
NA, NA, NA, NA, NA, NA, NA, NA, "<2", "<2", "<2", "1.3", "69.67", 
"16.6", "33.7", "8.1", "7.6", "24.622", "5.5", "25.279", "<2", 
"<2", "7.8", "1.38", "72.96", "13.2", "33.61", "<2", NA, NA, 
NA, NA, NA, NA, NA, NA, NA, "<2", "77.82", "20.8", "33.72", "8.2", 
"8.8", "23.58", "1.01", "<2", "<2", "1.78", "<2", "<2", "8", 
"6.5", "13.5", "67.19", "2e", "25.197", "33.58", "2e", "2e", 
"<2", "75.53", "12.9", "33.61", "7.9", "5.5", "25.34", "1.77"
)), class = "data.frame", row.names = c(NA, -69L))

一些与我的问题无关的问题是POSIXct values become numeric in reshape2 dcast 和Error with custom aggregate function for a cast() call in R reshape2

使用na.omit 我的输出是：

  station     date depth CDOM DENSITY  DO ENTERO FECAL  PH   SAL TEMP TOTAL   XMS
1       A 7/2/2018    12  1.3  24.622 7.6     <2    <2 8.1  33.7 16.6    <2 69.67
2       A 7/2/2018    18 1.38  25.279 5.5     <2    <2 7.8 33.61 13.2    <2 72.96
3       A 7/9/2018     1 1.01   23.58 8.8     <2    <2 8.2 33.72 20.8    <2 77.82
4       A 7/9/2018    12 1.78  25.197 6.5     <2    <2   8 33.58 13.5    2e 67.19
5       A 7/9/2018    18 1.77   25.34 5.5     <2    2e 7.9 33.61 12.9    2e 75.53

不使用na.omit 的预期输出是：

  station     date depth AMMONIA CDOM  DENSITY  DO  ENTERO FECAL  PH   SAL TEMP TOTAL   XMS
1       A 7/2/2018     1    ns    ns      ns    ns     ns    ns   ns    ns   ns    ns    ns
2       A 7/2/2018    12    ns    1.3   24.622  7.6    <2    <2  8.1  33.7 16.6    <2 69.67
3       A 7/2/2018    18    ns    1.38  25.279  5.5    <2    <2  7.8 33.61 13.2    <2 72.96
4       A 7/9/2018     1    ns    1.01  23.58   8.8    <2    <2  8.2 33.72 20.8    <2 77.82
5       A 7/9/2018    12    ns    1.78  25.197  6.5    <2    <2    8 33.58 13.5    2e 67.19
6       A 7/9/2018    18    ns    1.77  25.34   5.5    <2    2e  7.9 33.61 12.9    2e 75.53

【问题讨论】：

标签： r reshape2 dcast

【解决方案1】：

实际问题是所有参数对于（站、日期、深度）的每个三元组只有一个值除了AMMONIA AS N，它有三个NA 条目。

例如，

dcast(df, station + date + depth ~ parmcode, value.var = "value_qualif")
# Aggregation function missing: defaulting to length
#   station     date depth AMMONIA AS N CDOM DENSITY DO ENTERO FECAL PH SAL TEMP TOTAL XMS
# 1       A 7/1/2018     1            3    0       0  0      0     0  0   0    0     0   0
# 2       A 7/1/2018    12            3    0       0  0      0     0  0   0    0     0   0
# 3       A 7/1/2018    18            3    0       0  0      0     0  0   0    0     0   0
# 4       A 7/2/2018     1            0    1       1  1      1     1  1   1    1     1   1
# 5       A 7/2/2018    12            0    1       1  1      1     1  1   1    1     1   1
# 6       A 7/2/2018    18            0    1       1  1      1     1  1   1    1     1   1
# 7       A 7/9/2018     1            0    1       1  1      1     1  1   1    1     1   1
# 8       A 7/9/2018    12            0    1       1  1      1     1  1   1    1     1   1
# 9       A 7/9/2018    18            0    1       1  1      1     1  1   1    1     1   1

一旦我们删除了重复的行，一切都会顺利进行

dcast(df[!duplicated(df), ], station + date + depth ~ parmcode, value.var = "value_qualif")
#   station     date depth AMMONIA AS N CDOM DENSITY   DO ENTERO FECAL   PH   SAL TEMP TOTAL   XMS
# 1       A 7/1/2018     1         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 2       A 7/1/2018    12         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 3       A 7/1/2018    18         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 4       A 7/2/2018     1         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 5       A 7/2/2018    12         <NA>  1.3  24.622  7.6     <2    <2  8.1  33.7 16.6    <2 69.67
# 6       A 7/2/2018    18         <NA> 1.38  25.279  5.5     <2    <2  7.8 33.61 13.2    <2 72.96
# 7       A 7/9/2018     1         <NA> 1.01   23.58  8.8     <2    <2  8.2 33.72 20.8    <2 77.82
# 8       A 7/9/2018    12         <NA> 1.78  25.197  6.5     <2    <2    8 33.58 13.5    2e 67.19
# 9       A 7/9/2018    18         <NA> 1.77   25.34  5.5     <2    2e  7.9 33.61 12.9    2e 75.53

dcast(df[!duplicated(df), ], station + date + depth ~ parmcode, value.var = "value_qualif", fill = "ns")
#   station     date depth AMMONIA AS N CDOM DENSITY  DO ENTERO FECAL  PH   SAL TEMP TOTAL   XMS
# 1       A 7/1/2018     1           ns   ns      ns  ns     ns    ns  ns    ns   ns    ns    ns
# 2       A 7/1/2018    12           ns   ns      ns  ns     ns    ns  ns    ns   ns    ns    ns
# 3       A 7/1/2018    18           ns   ns      ns  ns     ns    ns  ns    ns   ns    ns    ns
# 4       A 7/2/2018     1           ns   ns      ns  ns     ns    ns  ns    ns   ns    ns    ns
# 5       A 7/2/2018    12           ns  1.3  24.622 7.6     <2    <2 8.1  33.7 16.6    <2 69.67
# 6       A 7/2/2018    18           ns 1.38  25.279 5.5     <2    <2 7.8 33.61 13.2    <2 72.96
# 7       A 7/9/2018     1           ns 1.01   23.58 8.8     <2    <2 8.2 33.72 20.8    <2 77.82
# 8       A 7/9/2018    12           ns 1.78  25.197 6.5     <2    <2   8 33.58 13.5    2e 67.19
# 9       A 7/9/2018    18           ns 1.77   25.34 5.5     <2    2e 7.9 33.61 12.9    2e 75.53

或者，你可以运行

dcast(df, station + date + depth ~ parmcode, value.var = "value_qualif", 
      fill = NA_character_, fun.aggregate = head, n = 1)
#   station     date depth AMMONIA AS N CDOM DENSITY   DO ENTERO FECAL   PH   SAL TEMP TOTAL   XMS
# 1       A 7/1/2018     1         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 2       A 7/1/2018    12         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 3       A 7/1/2018    18         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 4       A 7/2/2018     1         <NA> <NA>    <NA> <NA>   <NA>  <NA> <NA>  <NA> <NA>  <NA>  <NA>
# 5       A 7/2/2018    12         <NA>  1.3  24.622  7.6     <2    <2  8.1  33.7 16.6    <2 69.67
# 6       A 7/2/2018    18         <NA> 1.38  25.279  5.5     <2    <2  7.8 33.61 13.2    <2 72.96
# 7       A 7/9/2018     1         <NA> 1.01   23.58  8.8     <2    <2  8.2 33.72 20.8    <2 77.82
# 8       A 7/9/2018    12         <NA> 1.78  25.197  6.5     <2    <2    8 33.58 13.5    2e 67.19
# 9       A 7/9/2018    18         <NA> 1.77   25.34  5.5     <2    2e  7.9 33.61 12.9    2e 75.53

关于NA_character_，请参阅this answer。

【讨论】：

我现在看到了，这些样本可能一式三份。这与我认为的错误相反。我不得不修改 duplicate 调用，因为我的实际数据中的这些重复项具有唯一的样本 ID，但这起到了作用。