在 hive/impala 中运行多个 sql 查询以测试通过或失败

Question

I am running 100 queries (test cases) to check for data quality in hive/impala. The majority of the queries check for null values based on some conditions. I am using conditional aggregation to count the trivial test cases like below. I want to add a more complex query condition to this type of check. I also would like to see the counts if there are nulls.

I want to know how to incorporate the more complex query and also add a count if there are nulls present. Expected output below.

What I have so far:

SELECT (CASE WHEN COUNT(*) = COUNT(car_type) THEN 'PASS' ELSE 'FAIL' END) as car_type_test,
       (CASE WHEN COUNT(*) = COUNT(car_color) THEN 'PASS' ELSE 'FAIL' END) as car_color_test,
       (CASE WHEN COUNT(*) = COUNT(car_sale) THEN 'PASS' ELSE 'FAIL' END) as car_sale_test       
FROM car_data;

More complex type query to add:

SELECT Count(*), 
       car_job 
FROM   car_data 
WHERE  car_job NOT IN ( "car_type", "car_license", "car_cancellation", 
                        "car_color", "car_contract", "car_metal", "car_number" ) 
        OR car_job IS NULL 
GROUP  BY car_job

Example expected output:

car_type_test  car_color_test  car_sale_test  car_job_test
PASS           PASS             PASS           FAIL
                                               102

Answer 1

I would recommend putting this on one row instead of two:

SELECT (CASE WHEN COUNT(*) = COUNT(car_type) THEN 'PASS'
             ELSE REPLACE('FAIL ([n])', '[n]', COUNT(*) - COUNT(car_type))
        END) as car_type_test,
       (CASE WHEN COUNT(*) = COUNT(car_color) THEN 'PASS'
             ELSE REPLACE('FAIL ([n])', '[n]', COUNT(*) - COUNT(car_color))
        END) as car_color_test,
       (CASE WHEN COUNT(*) = COUNT(car_sale) THEN 'PASS'
             ELSE REPLACE('FAIL ([n])', '[n]', COUNT(*) - COUNT(car_sale))
        END) as car_sale_test       
FROM car_data;

Answer 2

If there is an option to have output look more like a table, rather than a 100-column (one per test case?) something, then solution may be easier.

Test name         Test results    Extra info
car_type_test     PASS               
car_color_test    PASS
car_sale_test     PASS
car_job_test      PASS            102

For example, you can build a UNION of all your queries, provided they conform to the same schema.

SELECT * FROM (
  SELECT 
    'car_type_test' `Test name`,
    CASE WHEN COUNT(*) = COUNT(car_type) THEN 'PASS'
         ELSE 'FAIL' 
    END  `Test result`,
    '' `Extra info`
  FROM car_data
  UNION ALL
  ...
  UNION ALL
  SELECT 
     'car_job_test' `Test name`,
     CASE WHEN count(*) > 0 THEN 'FAIL'
          ELSE 'PASS' 
     END `Test result`,
     collect_list(cast(count(*) as string)  `Extra info`
  FROM   car_data 
  WHERE  car_job NOT IN ( "car_type", "car_license", "car_cancellation", 
                        "car_color", "car_contract", "car_metal", "car_number" ) 
        OR car_job IS NULL 
  GROUP  BY car_job
) TESTS;