将 dplyr::quos() 与列表参数而不是省略号参数一起使用答案

【问题标题】：Using dplyr::quos() with a list argument rather than the ellipsis argument将 dplyr::quos() 与列表参数而不是省略号参数一起使用
【发布时间】：2018-04-14 00:03:10
【问题描述】：

我正在使用 dplyr 并尝试创建一个函数来根据分组参数计算 p.values。我希望能够有一个参数，该参数是要分组的任意长度变量的列表。这是示例数据集：

dataset <- structure(list(Experiment = c(170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170222, 170222, 170222, 170222, 
170222, 170222, 170222, 170222, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824, 170824, 170824, 170824, 170824, 
170824, 170824, 170824, 170824), Sample = c("1: FL_496", "1: FL_496", 
"1: FL_496", "1: FL_496", "1: FL_496", "1: FL_496", "1: FL_496", 
"1: FL_496", "2: FL_505", "2: FL_505", "2: FL_505", "2: FL_505", 
"2: FL_505", "2: FL_505", "2: FL_505", "2: FL_505", "3: FL_509", 
"3: FL_509", "3: FL_509", "3: FL_509", "3: FL_509", "3: FL_509", 
"3: FL_509", "3: FL_509", "4: FL_514", "4: FL_514", "4: FL_514", 
"4: FL_514", "4: FL_514", "4: FL_514", "4: FL_514", "4: FL_514", 
"5: cKO_497", "5: cKO_497", "5: cKO_497", "5: cKO_497", "5: cKO_497", 
"5: cKO_497", "5: cKO_497", "5: cKO_497", "6: cKO_504", "6: cKO_504", 
"6: cKO_504", "6: cKO_504", "6: cKO_504", "6: cKO_504", "6: cKO_504", 
"6: cKO_504", "7: cKO_510", "7: cKO_510", "7: cKO_510", "7: cKO_510", 
"7: cKO_510", "7: cKO_510", "7: cKO_510", "7: cKO_510", "8: cKO_515", 
"8: cKO_515", "8: cKO_515", "8: cKO_515", "8: cKO_515", "8: cKO_515", 
"8: cKO_515", "8: cKO_515", "9: cKO_517", "9: cKO_517", "9: cKO_517", 
"9: cKO_517", "9: cKO_517", "9: cKO_517", "9: cKO_517", "9: cKO_517", 
NA, NA, NA, NA, NA, NA, NA, NA, "1: FL_627", "1: FL_627", "1: FL_627", 
"1: FL_627", "1: FL_627", "1: FL_627", "2: FL_628", "2: FL_628", 
"2: FL_628", "2: FL_628", "2: FL_628", "2: FL_628", "3: FL_633", 
"3: FL_633", "3: FL_633", "3: FL_633", "3: FL_633", "3: FL_633", 
"4: FL_636", "4: FL_636", "4: FL_636", "4: FL_636", "4: FL_636", 
"4: FL_636", "5: cKO_620", "5: cKO_620", "5: cKO_620", "5: cKO_620", 
"5: cKO_620", "5: cKO_620", "6: cKO_625", "6: cKO_625", "6: cKO_625", 
"6: cKO_625", "6: cKO_625", "6: cKO_625", "7: cKO_626", "7: cKO_626", 
"7: cKO_626", "7: cKO_626", "7: cKO_626", "7: cKO_626", "8: cKO_634", 
"8: cKO_634", "8: cKO_634", "8: cKO_634", "8: cKO_634", "8: cKO_634", 
"cKO_620", "cKO_620", "cKO_625", "cKO_625", "cKO_626", "cKO_626", 
"cKO_634", "cKO_634", "FL_627", "FL_627", "FL_628", "FL_628", 
"FL_633", "FL_633", "FL_636", "FL_636"), Genotype = structure(c(1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("miR-15/16 FL", 
"miR-15/16 cKO"), class = "factor"), variable = c("% CD127+", 
"% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", 
"% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", "% CD127+", "% CD127+", 
"% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", 
"% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", 
"% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", "% CD127+", "% CD127+", 
"% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", 
"% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", 
"% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", "% CD127+", "% CD127+", 
"% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% KLRG1+", "% CD127+", 
"% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% CD127+", "% CD127+", "% KLRG1+", "% KLRG1+", 
"% KLRG1+", "% CD127+", "% KLRG1+", "% CD127+", "% KLRG1+", "% CD127+", 
"% KLRG1+", "% CD127+", "% KLRG1+", "% CD127+", "% KLRG1+", "% CD127+", 
"% KLRG1+", "% CD127+", "% KLRG1+", "% CD127+", "% KLRG1+"), 
    value = c(1, 28.7, 40.1, 47.4, 64.1, 69.9, 73.1, 79.42, 0.99, 
    21.72, 33, 56.6, 55.5, 82.9, 84.96, 86.7, 3.94, 43.4, 49.5, 
    60.8, 57.1, 69.8, 71.4, 77.72, 1, 20.56, 28.77, 35.1, 71.07, 
    71.2, 78.16, 84.04, 3.77, 56.9, 60.5, 66.5, 43.7, 50.36, 
    50.8, 51.8, 3.24, 58.2, 59.8, 70.8, 47.9, 58.5, 59.5, 61.3, 
    4.21, 62, 65.7, 73.8, 40, 51.5, 53.1, 55.69, 9.48, 41.7, 
    44, 63, 53.7, 57.31, 60.4, 60.8, 3.84, 34.1, 41.1, 53.2, 
    55.07, 55.3, 62.2, 76.6, NA, NA, NA, NA, NA, NA, NA, NA, 
    12.01, 18.5, 20.99, 66.39, 77.2, 85.6, 12.8, 31.3, 35.11, 
    59.8, 85.5, 89.7, 32.1, 33.3, 34.7, 63.2, 71.6, 80.5, 15.3, 
    17.02, 33.5, 65.54, 82.7, 85.8, 41.61, 51.3, 69.3, 39.81, 
    59, 62, 46.6, 52.1, 67.8, 39.5, 58.8, 66, 52.2, 52.9, 68.7, 
    46, 55.9, 61.6, 45.17, 59.9, 74.3, 31.87, 48.4, 51.2, 6.2, 
    56.34, 4.17, 70.85, 3.54, 59.89, 5.61, 49.71, 1.87, 77.09, 
    0.51, 86.05, 1.8, 80.69, 2.15, 79.43), Day = structure(c(1L, 
    2L, 3L, 4L, 4L, 3L, 2L, 1L, 1L, 3L, 4L, 2L, 2L, 4L, 1L, 3L, 
    1L, 3L, 2L, 4L, 4L, 2L, 3L, 1L, 1L, 3L, 4L, 2L, 4L, 2L, 3L, 
    1L, 1L, 3L, 2L, 4L, 4L, 1L, 2L, 3L, 1L, 3L, 2L, 4L, 4L, 2L, 
    3L, 1L, 1L, 3L, 2L, 4L, 4L, 3L, 2L, 1L, 1L, 3L, 4L, 2L, 2L, 
    1L, 4L, 3L, 1L, 2L, 3L, 4L, 1L, 4L, 3L, 2L, 2L, 3L, 4L, 1L, 
    2L, 3L, 4L, 1L, 3L, 2L, 4L, 3L, 2L, 4L, 2L, 3L, 4L, 3L, 2L, 
    4L, 2L, 3L, 4L, 3L, 2L, 4L, 2L, 3L, 4L, 3L, 4L, 2L, 3L, 2L, 
    4L, 3L, 2L, 4L, 3L, 2L, 4L, 3L, 2L, 4L, 3L, 2L, 4L, 3L, 2L, 
    4L, 3L, 2L, 4L, 3L, 2L, 4L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("8", "15", "22", 
    "30+"), class = "factor")), class = "data.frame", row.names = c(NA, 
-144L), .Names = c("Experiment", "Sample", "Genotype", "variable", 
"value", "Day"))

这是我使用 ... 制作的功能。

grouped.t.test <- function(dataset, subset.plot, comparison, ...)
  {
  group.by <- quos(...)
  if (is.null(subset.plot)){
    subset.plot <- dataset[['variable']]
    }
  filter(dataset, variable %in% subset.plot) %>%
    group_by(!!!group.by) %>%
    do(tidy(t.test(x = .$value[.[comparison] == levels(.[[comparison]])[1]],
                   y = .$value[.[comparison] == levels(.[[comparison]])[2]]))) %>%
    mutate(p.value.format = symnum(p.value, corr = FALSE, na = FALSE, cutpoints = c(0, 0.0001, 0.001, 0.01, 0.05, 1), symbols = c("****", "***", "**", "*", NA))) %>%
    arrange(!!!group.by)
  }
View(grouped.t.test(dataset = dataset, subset.plot = NULL, comparison = 'Genotype', variable, Day))

我希望能够将 ... 替换为一个参数（例如 group_vars）并像这样调用它：

View(grouped.t.test(dataset = dataset, subset.plot = NULL, comparison = 'Genotype', group_vars = c(variable, Day)))

这似乎不适用于 quos() 但我不明白为什么。能够使用多个独立使用的列表参数会很好（例如，创建一个参数“arrange.by”，它将是一个变量列表，用于在函数末尾进行排列。

我非常感谢任何帮助理解为什么这不起作用以及我可以做些什么！

【问题讨论】：

我对你的意思有点困惑。 c(variable, Day) 不会创建符号向量，它会尝试通过这些名称查找实际对象。您可以尝试在控制台输入c(variable, Day)。也许您的意思是您想传递 group_vars = c("variable","Day") 之类的参数，即作为字符？
我的理解是，使用 quosures 应该可以让你捕捉到函数的用户传递的内容，然后引用它。然后！！！应该在 group_by() 函数中取消引用它。这不是它的工作原理吗？
是的，但工作方式是使用...。您可以使用rlang::syms 之类的内容创建符号列表，但这对您的用户来说更有用。通常，您要么通过... 一次传递一个，要么作为字符向量在函数中自己将它们转换为符号。
我明白了。因此，如果我提供一个字符向量，并使用rlang::syms 创建一个符号列表，那么将其传递给group_by() 的正确方法是什么？谢谢！
看起来它适用于 !!! 取消引用。这是正确的实现吗？我收到警告消息：Vectorizing 'noquote' elements may not preserve their attributes using this.

标签： r dplyr tidyeval

【解决方案1】：

正如@lionel 所述——dplyr 的主要开发人员之一在此comment 中

您希望引用是外部的并由用户明确完成，而不是由您的函数隐式完成。为此，您可以要求您的用户引用 base::alist()、rlang::exprs() 或 dplyr::vars()

你可以为你的问题做这样的事情

grouped.t.test2 <- function(dataset, subset.plot, comparison, group_vars) {

  if (is.null(subset.plot)) {
    subset.plot <- dataset[['variable']]
  }

  filter(dataset, variable %in% subset.plot) %>%
    group_by(!!! group_vars) %>%
    do(tidy(t.test(x = .$value[.[comparison] == levels(.[[comparison]])[1]],
                   y = .$value[.[comparison] == levels(.[[comparison]])[2]]))) %>%
    mutate(p.value.format = symnum(p.value, corr = FALSE, na = FALSE, 
                                   cutpoints = c(0, 0.0001, 0.001, 0.01, 0.05, 1), 
                                   symbols = c("****", "***", "**", "*", NA))) %>%
    arrange(!!! group_vars)
}

grouped.t.test2(dataset = dataset, subset.plot = NULL, comparison = 'Genotype', 
               alist(variable, Day))

# or

grouped.t.test2(dataset = dataset, subset.plot = NULL, comparison = 'Genotype', 
               dplyr::vars(variable, Day))

# A tibble: 8 x 13
# Groups:   variable, Day [8]
  variable Day   estimate estimate1 estimate2 statistic p.value parameter
  <fct>    <fct>    <dbl>     <dbl>     <dbl>     <dbl>   <dbl>     <dbl>
1 % CD127+ 8        -3.24      1.66      4.90     -4.26 9.93e-4      12.6
2 % CD127+ 15      -24.4      31.1      55.5      -3.80 2.88e-3      11.2
3 % CD127+ 22      -22.1      27.4      49.5      -4.60 5.54e-4      12.5
4 % CD127+ 30+     -28.6      36.8      65.4      -5.23 1.36e-4      13.7
5 % KLRG1+ 8        23.8      81.2      57.4       9.79 3.11e-7      12.5
6 % KLRG1+ 15       16.5      73.7      57.2       3.78 2.08e-3      13.8
7 % KLRG1+ 22       20.9      70.1      49.2       4.44 4.82e-4      14.9
8 % KLRG1+ 30+      22.5      76.7      54.2       4.46 6.01e-4      13.4
# ... with 5 more variables: conf.low <dbl>, conf.high <dbl>,
#   method <fct>, alternative <fct>, p.value.format <chr>

【讨论】：