Comment faire groupe par le champ répété dans BigQuery

Question

En BigQuery, j'ai créé une table avec le schéma ci-dessous

id     INTEGER NULLABLE  
visits    INTEGER NULLABLE  
dimensions   RECORD REPEATED  
dimensions.value STRING 
dimensions.key  STRING 

Comment obtenir une somme (visites) en regroupant les périphériques et les valeurs étatiques?

données de l'exemple:

{"id": 1, visits: 100, "dimensions": [{"key":"device","value":"mobile"}, {"key":"state","value":"CA"}]} 
{"id": 1, visits: 500, "dimensions": [{"key":"device","value":"desktop"}, {"key":"state","value":"CA"}]} 
{"id": 1, visits: 200, "dimensions": [{"key":"device","value":"mobile"}, {"key":"state","value":"NY"}]} 
{"id": 2, visits: 100, "dimensions": [{"key":"device","value":"mobile"}, {"key":"state","value":"CA"}]} 
{"id": 2, visits: 500, "dimensions": [{"key":"device","value":"desktop"}, {"key":"state","value":"CA"}]} 
{"id": 2, visits: 200, "dimensions": [{"key":"device","value":"mobile"}, {"key":"state","value":"NY"}]} 
{"id": 2, visits: 780, "dimensions": [{"key":"device","value":"desktop"}, {"key":"state","value":"NY"}]} 

Je veux id, dispositif, état, somme (visites) dans la sortie.

Je pourrais faire un groupe en utilisant une seule dimension avec la requête ci-dessous, mais je ne sais pas comment le faire pour plusieurs dimensions.

SELECT id,d.value, sum(visits) FROM dataset.tabe_name,UNNEST(dimensions) as d where d.key = "device" group by id, d.value LIMIT 1000 

Et est-il également possible d'écrire une requête générique lorsque les valeurs clés ne sont pas connues à l'avance?

Answer 1

est ci-dessous pour BigQuery standard SQL

#standardSQL 
SELECT 
    id, 
    (SELECT value FROM UNNEST(dimensions) WHERE key = "device") AS device, 
    (SELECT value FROM UNNEST(dimensions) WHERE key = "state") AS state, 
    SUM(visits) AS visits 
FROM `dataset.tabe_name` 
GROUP BY id, device, state 
LIMIT 1000 

Vous pouvez essayer/jouer avec des données factices de votre exemple ci-dessous

#standardSQL 
WITH data AS (
    SELECT 1 AS id, 100 AS visits, ARRAY<STRUCT<key STRING, value STRING>>[("device", "mobile"), ("state", "CA")] AS dimensions UNION ALL 
    SELECT 1, 100, [STRUCT<key STRING, value STRING>("device", "mobile"), ("state", "CA")] UNION ALL 
    SELECT 1, 500, [STRUCT<key STRING, value STRING>("device", "desktop"), ("state", "CA")] UNION ALL 
    SELECT 1, 200, [STRUCT<key STRING, value STRING>("device", "mobile"), ("state", "NY")] UNION ALL 
    SELECT 2, 100, [STRUCT<key STRING, value STRING>("device", "mobile"), ("state", "CA")] UNION ALL 
    SELECT 2, 500, [STRUCT<key STRING, value STRING>("device", "desktop"), ("state", "CA")] UNION ALL 
    SELECT 2, 200, [STRUCT<key STRING, value STRING>("device", "mobile"), ("state", "NY")] UNION ALL 
    SELECT 2, 780, [STRUCT<key STRING, value STRING>("device", "desktop"), ("state", "NY")] 
) 
SELECT 
    id, 
    (SELECT value FROM UNNEST(dimensions) WHERE key = "device") AS device, 
    (SELECT value FROM UNNEST(dimensions) WHERE key = "state") AS state, 
    SUM(visits) AS visits 
FROM data 
GROUP BY id, device, state 
-- ORDER BY id, device, state