通過customDimension的特定值進行選擇（BQ <> GA）

Question

我是使用BiqQuery（幾個星期的經驗）的新手，並試圖提高我的技能。我對用戶Willian Fuks發佈的 （Recreate GA Funnel on BigQuery）以下非常有趣的查詢有一個實用的問題。它是關於BigQuery中的GA數據以有效的方式重現漏斗的。

#standardSQL 
SELECT 
SUM((SELECT COUNTIF(eventInfo.eventAction = 'landing_page') FROM UNNEST(hits))) Landing_Page, 
SUM((SELECT COUNTIF(eventInfo.eventAction = 'model_selection_page') FROM UNNEST(hits) WHERE EXISTS(SELECT 1 FROM UNNEST(hits) WHERE eventInfo.eventAction = 'landing_page'))) Model_Selection 
FROM `64269470.ga_sessions_20170720` 

在該示例中，使用eventInfo.eventAction。我嘗試了幾件事情來使它與customDimension一起工作，但我失敗了。有誰知道我怎麼能重現查詢分段與customDimension而不是eventInfo.eventAction？

我曾與這樣的：

(SELECT MAX(IF(index=1,page1, NULL))FROM UNNEST(hits.customDimensions)) 

Answer 1

與customDimensions工作是更具挑戰性的一點，因爲這個領域也是一個數組類型（重複）。但實際上，主要區別在於需要另一個UNNEST操作。除此之外，這是相同的邏輯。

這裏有一些數據，向您展示：

#standardSQL 
WITH data AS(
    SELECT '1' AS fullvisitorid, 1 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(1 AS index, 'value1' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits UNION ALL 

    SELECT '1' AS fullvisitorid, 2 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits UNION ALL 

    SELECT '2' AS fullvisitorid, 1 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(1 AS index, 'value1' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits UNION ALL 

    SELECT '3' AS fullvisitorid, 1 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(3 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits UNION ALL 

    SELECT '3' AS fullvisitorid, 2 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(3 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(3 AS hitNumber, [STRUCT(3 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits UNION ALL 
    SELECT '4' AS fullvisitorid, 1 AS visitid, ARRAY<STRUCT< hitNumber INT64, customDimension ARRAY<STRUCT<index INT64, value STRING> > >> [STRUCT(1 AS hitNumber, [STRUCT(1 AS index, 'landing_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(2 AS hitNumber, [STRUCT(3 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension), 
                                      STRUCT(3 AS hitNumber, [STRUCT(3 AS index, 'model_selection_page' AS value), STRUCT(2 AS index, 'value2' AS value)] AS customDimension)] AS hits                                                                                                                                
) 

每個用戶（fullvisitorid）和每個會話（visitid）有其hits數組。注意我把它的每個命中分開了hitNumber，我發現它讓事情變得更容易理解。

下面的查詢計算其中索引1和值「LANDING_PAGE」一個customDimension發生會話總數，以及其中指數爲3和值是「model_selection_page」：

#standardSQL 
SELECT 
    SUM((SELECT 1 FROM UNNEST(hits), UNNEST(customDimension) WHERE index = 1 AND value = 'landing_page' LIMIT 1)) Landing_Page, 
    SUM((SELECT 1 FROM UNNEST(hits), UNNEST(customDimension) WHERE EXISTS(SELECT 1 FROM UNNEST(hits), UNNEST(customDimension) WHERE index = 1 AND value = 'landing_page') AND index = 3 AND value = 'model_selection_page' LIMIT 1)) Model_Selection 
FROM 
    data 

你可以玩的模擬數據以更好地理解這裏發生了什麼。簡而言之，請注意兩個UNNEST發生，首先獲取hits中的值，第二個獲取customDimension中的值。

領域Model_Selection是一個比較複雜的，因爲它首先必須評估維度「LANDING_PAGE」是否被解僱了，因爲你可以在這個表達式看到：

EXISTS(SELECT 1 FROM UNNEST(hits), UNNEST(customDimension) WHERE index = 1 AND value = 'landing_page') 

如果hits有某處「LANDING_PAGE」維度，那麼此表達式會在WHERE子句中返回True。

您也可以帶來的效果在用戶層面，像這樣：

#standardSQL 
SELECT 
    COUNT(DISTINCT (SELECT fullvisitorid FROM UNNEST(hits), UNNEST(customDimension) WHERE index = 1 AND value = 'landing_page' LIMIT 1)) Landing_Page, 
    COUNT(DISTINCT (SELECT fullvisitorid FROM UNNEST(hits), UNNEST(customDimension) WHERE EXISTS(SELECT 1 FROM UNNEST(hits), UNNEST(customDimension) WHERE index = 1 AND value = 'landing_page') AND index = 3 AND value = 'model_selection_page' LIMIT 1)) Model_Selection 
FROM 
    data 

當你正在學習的BigQuery，我建議用模擬擺弄數據和測試每一步觀察輸出。你可以玩UNNEST，運行一些查詢來測試它的輸出等，以便更好和更深入地理解如何使用這些技術。