sparklyr 和基于标准评估 (SE) 的函数

Question

我正在尝试编写一个函数来执行 sdf_pivot() a 创建一个 Spark DataFrame，其列名包含原始变量或列的名称。

set.seed(80)
df <- data.frame(id = c(1:5),
                 var1 = sample(LETTERS[1:12], 5, replace = TRUE),
                 var2 = sample(LETTERS[13:16], 5, replace = TRUE))

ref <- copy_to(sc, df, "mytbl")
glimpse(ref)

Observations: 5
Variables: 3
$ id   <int> 1, 2, 3, 4, 5
$ var1 <chr> "F", "G", "J", "A", "H"
$ var2 <chr> "M", "O", "O", "O", "O"

这是var1 的预期结果，无需编写函数：

ref %>% 
  dplyr::select(id, var1) %>%
  dplyr::mutate(newvar1 = paste0("var1_",var1)) %>%
  sparklyr::sdf_pivot(formula = id ~ newvar1, fun.aggregate = "count") %>% 
  sparklyr::na.replace(0)

# Source:   table<sparklyr_tmp_56f96ab7d507> [?? x 6]
# Database: spark_connection
     id var1_A var1_F var1_G var1_H var1_J
  <int>  <dbl>  <dbl>  <dbl>  <dbl>  <dbl>
1     1      0      1      0      0      0
2     3      0      0      0      0      1
3     5      0      0      0      1      0
4     4      1      0      0      0      0
5     2      0      0      1      0      0

在我的一个函数版本下，当然不起作用，我也尝试过quote 和deparse，但我被mutate_ 和sdf_pivot 卡住了。

myPivotFunction <- function(sdf, varname, newvarname){

  mutate_op <- paste0(newvarname," = ", "var1_", varname)

  sdf %>% 
    dplyr::select_(.dots = list('id', varname)) %>%
    mutate_(.dots = setNames(newvarname, mutate_op)) %>%
    sparklyr::sdf_pivot(formula = id ~ newvar1, fun.aggregate = "count") %>% 
    sparklyr::na.replace(0)
}

Answer 1

一点点rlang 应该可以解决问题：

library(rlang)
library(glue)

myPivotFunction <- function(sdf, varname, newvarname){
  exprs <- c("id", glue('paste0("var1_", {varname})')) %>% 
    setNames(c("id", newvarname)) %>% 
    lapply(parse_quosure)

  sdf %>%
    transmute(!!! exprs) %>%
    sdf_pivot(
      formula = as.formula(glue("id ~ {newvarname}")), 
      fun.aggregate = "count") %>%
    na.replace(0)
}