在 BigQuery 中搜索 ARRAY of STRUCT 的条件

Question

I have a problem in BigQuery related to ARRAY and STRUCT. I have a data structure like this:

    SELECT 
       1 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data
       UNION ALL SELECT 2 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name)] AS data
       UNION ALL SELECT 3 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
       UNION ALL SELECT 4 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
       UNION ALL SELECT 5 AS id, [STRUCT('d' AS name, 'd1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data

Now I want to pick a rows by a condition on the name. For example I want only the rows where name = 'a' AND name = 'b' For the example above only id 1 should be returned as a correct answer.

If I flatten the array using UNNEST and try to run this query I get empty results:

WITH sequences AS
  (SELECT 
   1 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data
   UNION ALL SELECT 2 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name)] AS data
   UNION ALL SELECT 3 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 4 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 5 AS id, [STRUCT('d' AS name, 'd1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data)
     
SELECT sequences.id
        
FROM sequences

WHERE EXISTS (SELECT * FROM UNNEST(data) AS x WHERE x.name = 'a' AND x.name = 'b' )

I have tried to use JOIN to get the correct results but using JOIN gave me double results. For example this query returned me id's 1,2,3,4

WITH sequences AS
  (SELECT 
   1 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data
   UNION ALL SELECT 2 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name)] AS data
   UNION ALL SELECT 3 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 4 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 5 AS id, [STRUCT('d' AS name, 'd1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data),
     
flat AS (SELECT sequences.id, data.*
        
FROM sequences, UNNEST(data) AS data)

SELECT 
 f1.id
FROM flat f1
JOIN flat f2 USING(id)
WHERE ( f1.name = 'a' AND f2.name = 'b') 

How can I pick the rows where name only equals to 'a' and 'b'

Answer 1

Assumming names are distinct within the data

#standardSQL
SELECT *
FROM `project.dataset.table` t
WHERE (
  SELECT 
    COUNTIF(name IN ('a', 'b')) = 2
    AND COUNT(name) = 2
  FROM t.data
)

Answer 2

(After writing, I found it not very different from Mikhail Berlyant's answer)

CREATE TEMP FUNCTION has_names(data ANY TYPE, names ARRAY<STRING>)
AS ((
    SELECT ARRAY_LENGTH(data) = COUNT(*) FROM UNNEST(data) AS x WHERE x.name IN UNNEST(names) 
));

WITH sequences AS
  (SELECT 
   1 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data
   UNION ALL SELECT 2 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name)] AS data
   UNION ALL SELECT 3 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('c' AS name, 'c1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 4 AS id, [STRUCT('a' AS name, 'a1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name), STRUCT('d' AS name, 'd1' AS last_name)] AS data
   UNION ALL SELECT 5 AS id, [STRUCT('d' AS name, 'd1' AS last_name), STRUCT('b' AS name, 'b1' AS last_name)] AS data)
     
SELECT sequences.id
FROM sequences
WHERE has_names(data, ['a', 'b']);