在 bigquery 中创建一个 udf 以匹配数组输入

Posted 2023-03-25

【中文标题】在 bigquery 中创建一个 udf 以匹配数组输入

【英文标题】：Creating a udf in bigquery to match array inputs

【发布时间】：2020-06-18 21:47:02

【问题描述】：

我正在尝试创建一个 udf 以匹配 biqquery 中的数组，本质上如果 y 数组中有 x 数组的值，那么我希望结果为真。

例如match_rows([4,5,6], [5,6,7] 应该返回 true。

我已经把它写出来了，但我不断收到语法错误，而且我对我正在做的调试工作不够熟悉，所以希望有人能够对正在发生的事情有所了解。

具体错误 = No matching signature for function match_rows for argument types: ARRAY, ARRAY. Supported signature: match_rows(ARRAY, ARRAY) at [44:1]

CREATE TEMP FUNCTION match_rows(arr1 ARRAY<FLOAT64>, arr2 ARRAY<FLOAT64>)
  RETURNS BOOL
  LANGUAGE js AS 
  """
function findCommonElements2(arr1, arr2)  

    // Create an empty object 
    let obj = ; 

        // Loop through the first array 
        for (let i = 0; i < arr1.length; i++)  

            // Check if element from first array 
            // already exist in object or not 
            if(!obj[arr1[i]])  

                // If it doesn't exist assign the 
                // properties equals to the  
                // elements in the array 
                const element = arr1[i]; 
                obj[element] = true; 
             
         

        // Loop through the second array 
        for (let j = 0; j < arr2.length ; j++)  

        // Check elements from second array exist 
        // in the created object or not 
        if(obj[arr2[j]])  
            return true; 
         
     
    return false; 
 
""";

WITH input AS (
  SELECT STRUCT([5,6,7] as row, 'column2' as value) AS test
)
  SELECT
match_rows([4,4,6],[4,7,8]),
match_rows(test.row, test.row)

  FROM input ```

【参考方案1】：

您将 ARRAY<INT64> 传递给您的函数，而不是使用 ARRAY<FLOAT64>。因此，No matching signature 错误。为了解决这个错误，您可以将数组中的一个值分配为浮点数，语法如下：

WITH input AS (
#notice that the first element is a float and so the whole array is ARRAY<FLOAT64>
  SELECT STRUCT([5.0,6,7] as row, 'column2' as value) AS test
)
  SELECT
#the same as it was done above, first element explicitly as float
match_rows([4.0,4,6],[4,7,8]),
match_rows(test.row, test.row)

  FROM input

但是，我已经测试了您的函数，并且发现您的 javascript UDF 的语法不符合 documentation。语法应该如下：

CREATE TEMP FUNCTION multiplyInputs(x FLOAT64, y FLOAT64)
RETURNS FLOAT64
LANGUAGE js AS """
    //write directly your transformations here
    return x*y;
""";

注意，不需要指定function findCommonElements2(arr1, arr2)，你可以直接进入你的函数体，因为函数的名字是在声明CREATE TEMP FUNCTION之后定义的。

此外，我还发现您的函数没有返回所需的输出。出于这个原因，我编写了一个更简单的 JavaScript UDF，它返回你所期望的。下面是语法和测试：

CREATE TEMP FUNCTION match_rows(arr1 ARRAY<FLOAT64>, arr2 ARRAY<FLOAT64>)
  RETURNS BOOL
  LANGUAGE js AS 
  """
  //array of common elements betweent the two arrays
  var common_el= [];
  for(i=0;i < arr1.length;i++)
    for(j=0; j< arr2.length;j++)
      if(arr1[i] = arr2[j])
        //add to the common_el array when the element is present in both arrays
        common_el.push(arr1[i]);
      
    
  
  //if the array of common elements has at least one element return true othersie false 
  if(common_el.length > 0)return true;elsereturn false;
""";

WITH input AS (
  SELECT STRUCT([5.0,6,7] as row, 'column2' as value) AS test
)
  SELECT
match_rows([4.0,4,6],[4.0,5,7]) as check_1,
match_rows(test.row, test.row) as check_2

  FROM input#, unnest(test) as test

还有输出，

Row check_1 check_2
1   true    true

【参考方案2】：

以下是 BigQuery 标准 SQL

#standardSQL
CREATE TEMP FUNCTION match_rows(arr1 ANY TYPE, arr2 ANY TYPE) AS (
  (SELECT COUNT(1) FROM UNNEST(arr1) el JOIN UNNEST(arr2) el USING(el)) > 0
);
SELECT match_rows([4,5,6], [5,6,7])