通过 dbplyr/bigRquery 将 summarise() 调用中的分位数返回到 BigQuery SQL 数据库答案

【问题标题】：Return quantiles within a summarise() call through dbplyr/bigRquery to BigQuery SQL database通过 dbplyr/bigRquery 将 summarise() 调用中的分位数返回到 BigQuery SQL 数据库
【发布时间】：2020-07-24 19:39:09
【问题描述】：

我正在尝试获取分组 BigQuery 表中变量的分位数，但出现以下错误：

Error: Job 'xxxxx' failed
Syntax error: Expected end of input but got keyword WITHIN at [1:45] [invalidQuery]

Reprex 在下面。

# NOTE: for reprex to work, you must have BIGQUERY_TEST_PROJECT envvar set to name of project which has billing set up and to which you have write access

library(DBI)
library(bigrquery)
library(dplyr)

billing <- bq_test_project()

con <- dbConnect(
  bigrquery::bigquery(),
  project = "publicdata",
  dataset = "samples",
  billing = billing
)

natality <- tbl(con, "natality")
   
natality %>%
  group_by(year) %>%
  summarize(q25 = quantile(weight_pounds,0.25),
            q50 = median(weight_pounds),
            q75 = quantile(weight_pounds,0.75)
  )

任何人都知道一种解决方法，也许是通过在summarise() 调用中通过sql() 提供SQL 代码？

谢谢！

【问题讨论】：

您找到解决方法了吗？我面临着和你完全相同的问题。
中位数也给了我同样的错误......
@Ploulack 查看下面的答案

标签： r google-bigquery bigrquery

【解决方案1】：

一位同事通过在summarize() 调用中使用sql() 提供SQL 代码找到了答案：

# NOTE: for reprex to work, you must have BIGQUERY_TEST_PROJECT envvar set to name of project which has billing set up and to which you have write access

library(DBI)
library(bigrquery)
library(dplyr)

billing <- bq_test_project()

con <- dbConnect(
  bigrquery::bigquery(),
  project = "publicdata",
  dataset = "samples",
  billing = billing
)

natality <- tbl(con, "natality")
   
natality %>%
  group_by(year) %>%
  summarize(q25 = sql("approx_quantiles(weight_pounds,4)[offset(1)]"),
            q50 = sql("approx_quantiles(weight_pounds,2)[offset(1)]"),
            q75 = sql("approx_quantiles(weight_pounds,4)[offset(3)]")
  )

【讨论】：