如何在 R 中使用 tidyjson 处理嵌套的空 JSON 数组答案

【问题标题】：How to deal with nested empty JSON arrays with tidyjson in R如何在 R 中使用 tidyjson 处理嵌套的空 JSON 数组
【发布时间】：2019-11-01 21:05:42
【问题描述】：

我正在从 Salesforce 读取 JSON 对象。该对象是不规则的，因为有些嵌套数组是空的，有些不是。在 tidyjson 中如何处理？

我正在使用 R 中的 Salesforce 设置 API。目标是从 Salesforce 中获取有意义的数据以在 R 中处理。

json <- '
{
  "totalSize": [
    355710
  ],
  "done": [
    false
  ],
  "nextRecordsUrl": [
    "/services/data/v45.0/query/01gc000001L8zdkAAB-749"
  ],
  "records": [
    {
      "attributes": {
        "type": "Order_Line__c",
        "url": "/services/data/v45.0/sobjects/Order_Line__c/a0T1N000009aZ9lUAE"
      },
      "Id": "a0T1N000009aZ9lUAE",
      "Name": "OrderLine-1099369",
      "SO_Number_Formula__c": "548402-2.3",
      "Ship_From_Inventory__c": "XXX",
      "RMA_Number__c": "548402",
      "Part_Number__c": "01t1N00000JNeAQQA1",
      "Marketing_Part__c": "XXXXXXXXXXX",
      "Family__c": "XXXXXXXX",
      "Serial_Numbers__r": {
        "records": {}
      }
    },
    {
      "attributes": {
        "type": "Order_Line__c",
        "url": "/services/data/v45.0/sobjects/Order_Line__c/a0T1N000009aZ9mUAE"
      },
      "Id": "a0T1N000009aZ9mUAE",
      "Name": "OrderLine-1099370",
      "SO_Number_Formula__c": "962816-1.1",
      "Ship_From_Inventory__c": "XXX",
      "RMA_Number__c": "962816",
      "Part_Number__c": "01t1N00000JNc3qQAD",
      "Marketing_Part__c": "XXXXXXXXXX",
      "Family__c": "XXXXXXX",
      "RMA_Received_Date__c": "2019-02-18",
      "Serial_Numbers__r": {
        "totalSize": 1,
        "done": true,
        "records": [
          {
            "attributes": {
              "type": "Serial_Number__c",
              "url": "/services/data/v45.0/sobjects/Serial_Number__c/a0X1N00000NoyAjUAJ"
            },
            "Id": "a0X1N00000NoyAjUAJ",
            "Name": "SN217426",
            "Legacy_Line_Id__c": "962816SN217426",
            "Customer_Name__c": "XXXXXX",
            "Original_Shipment_Date__c": "2018-06-26",
            "Disposition__c": "Pending",
            "Status__c": "FailureVerification"
          }
        ]
      }
    }
  ]
}
'

mydata <- json %>% 
    as.tbl_json %>%
    enter_object("records") %>%
    gather_array() %>%
    spread_values(
      Id = jstring("Id"),
      Name = jstring("Name"),
      SO_Number_Formula = jstring("SO_Number_Formula__c"),
      Ship_From_Inventory = jstring("Ship_From_Inventory__c"),
      RMA_Number = jstring("RMA_Number__c"),
      Part_Number = jstring("Part_Number__c"),
      Marketing_Part = jstring("Marketing_Part__c"),
      Family = jstring("Family__c")) %>%
    enter_object("Serial_Numbers__r") %>%
    enter_object("records") %>%
    gather_ %>%
      spread_values(
    Id = jstring("Id"))

[records][Serial_Numbers__r][records] 中存在异常情况。在此示例中，第一次出现为空 {}，第二次出现非空。代码在执行gather_keys或gather_array时产生如下错误： gather_keys(.) 中的错误：1 条记录是值而不是对象 gather_array(.) 中的错误：1 条记录是值而不是数组

我认为这是由空数组 [records] 引起的。 Salesforce 输出中有很多这样的不规则性：有些记录有详细的嵌套数据，有些则没有。我该如何处理？

【问题讨论】：

标签： r salesforce-lightning jsonlite

【解决方案1】：

这是一个绝妙的问题，我们确实应该有一种更简洁的方式来处理。 enter_object() 在这些类型的情况下被证明是有问题的，您会根据不规则的 JSON 做法丢失记录。

我在这里提交了一个问题来跟踪改进：https://github.com/colearendt/tidyjson/issues/121

与此同时，我通常这样做的方式是根据描述它们的特征拆分记录。在这种情况下，您可以在父对象上使用gather_object() 以获得与enter_object() 相同的效果，然后使用filter / bind_rows 对行进行不同的处理。

理想情况下，bind_rows() 在这里的管道中会更好地工作...这是我希望看到的对 dplyr 的改进 (Issue here)！我很想知道这是否能解决您的问题！（另外，请记住 spread_all() 以简化某些列的指定，但代价是包的一些“猜测”！）。

  json <- '{
  "totalSize": [
    355710
  ],
  "done": [
    false
  ],
  "nextRecordsUrl": [
    "/services/data/v45.0/query/01gc000001L8zdkAAB-749"
  ],
  "records": [
    {
      "attributes": {
        "type": "Order_Line__c",
        "url": "/services/data/v45.0/sobjects/Order_Line__c/a0T1N000009aZ9lUAE"
      },
      "Id": "a0T1N000009aZ9lUAE",
      "Name": "OrderLine-1099369",
      "SO_Number_Formula__c": "548402-2.3",
      "Ship_From_Inventory__c": "XXX",
      "RMA_Number__c": "548402",
      "Part_Number__c": "01t1N00000JNeAQQA1",
      "Marketing_Part__c": "XXXXXXXXXXX",
      "Family__c": "XXXXXXXX",
      "Serial_Numbers__r": {
        "records": {}
      }
    },
    {
      "attributes": {
        "type": "Order_Line__c",
        "url": "/services/data/v45.0/sobjects/Order_Line__c/a0T1N000009aZ9mUAE"
      },
      "Id": "a0T1N000009aZ9mUAE",
      "Name": "OrderLine-1099370",
      "SO_Number_Formula__c": "962816-1.1",
      "Ship_From_Inventory__c": "XXX",
      "RMA_Number__c": "962816",
      "Part_Number__c": "01t1N00000JNc3qQAD",
      "Marketing_Part__c": "XXXXXXXXXX",
      "Family__c": "XXXXXXX",
      "RMA_Received_Date__c": "2019-02-18",
      "Serial_Numbers__r": {
        "totalSize": 1,
        "done": true,
        "records": [
          {
            "attributes": {
              "type": "Serial_Number__c",
              "url": "/services/data/v45.0/sobjects/Serial_Number__c/a0X1N00000NoyAjUAJ"
            },
            "Id": "a0X1N00000NoyAjUAJ",
            "Name": "SN217426",
            "Legacy_Line_Id__c": "962816SN217426",
            "Customer_Name__c": "XXXXXX",
            "Original_Shipment_Date__c": "2018-06-26",
            "Disposition__c": "Pending",
            "Status__c": "FailureVerification"
          }
        ]
      }
    }
  ]
}
'

  library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
  library(tidyr)
  library(tidyjson)
#> 
#> Attaching package: 'tidyjson'
#> The following object is masked from 'package:dplyr':
#> 
#>     bind_rows
#> The following object is masked from 'package:stats':
#> 
#>     filter

  prep_data <- json %>%
    as.tbl_json %>%
    enter_object("records") %>%
    gather_array() %>%
    spread_values(
      Id = jstring("Id"),
      Name = jstring("Name"),
      SO_Number_Formula = jstring("SO_Number_Formula__c"),
      Ship_From_Inventory = jstring("Ship_From_Inventory__c"),
      RMA_Number = jstring("RMA_Number__c"),
      Part_Number = jstring("Part_Number__c"),
      Marketing_Part = jstring("Marketing_Part__c"),
      Family = jstring("Family__c")) %>%
    enter_object("Serial_Numbers__r")

  # show that types are different
  prep_data %>%
    gather_object("key") %>%
    json_types() %>%
    select(key, type) %>%
    filter(key == "records")
#> # A tbl_json: 2 x 2 tibble with a "JSON" attribute
#>   `attr(., "JSON")`      key     type  
#>   <chr>                  <chr>   <fct> 
#> 1 "{}"                   records object
#> 2 "[{\"attributes\":..." records array

  # handle
  taller <- prep_data %>%
    gather_object("key") %>%
    json_types("type") %>%
    filter(key == "records")

  final <- tidyjson::bind_rows(
    taller %>% filter(type == "object"),
    taller %>% filter(type == "array") %>%
      gather_array("record_row") %>%
      spread_values(
        RecordId = jstring("Id")
      )
  )

  final %>% select(key, type, record_row, RecordId)
#> # A tbl_json: 2 x 4 tibble with a "JSON" attribute
#>   `attr(., "JSON")`      key     type   record_row RecordId          
#>   <chr>                  <chr>   <fct>       <int> <chr>             
#> 1 "{}"                   records object         NA <NA>              
#> 2 "{\"attributes\":{..." records array           1 a0X1N00000NoyAjUAJ

^{由reprex package (v0.3.0) 于 2020-03-15 创建}

【讨论】：

感谢 Cole 富有洞察力的评论。