爲bigquery生成嵌套的JSON

Question

我在Bigquery中創建了一個嵌套的表。輸入需要是JSON。我需要輸入以下示例：

{「store_nbr」：「1234」，「sls_dt」：「2014-01-01 00:00:00」，「Items」：[{「sku」：「 3456「，」sls_amt「：」9.99「，」discounts「：[{」disc_nbr「：」1「，」disc_amt「：」0.99「}，{」disc_nbr「：」2「，」disc_amt「 }]

我有很大的查詢扁平表。我已經閱讀了嵌套可以幫助重複json的旋轉。我在另一個線程中看到了一個很好的例子。

Nest multiple repeated fields in BigQuery

但這只是一個級別嵌套。我需要做兩個級別的「項目」 - >「折扣」。任何建議表示讚賞。

Answer 1

試試下面
可以打磨的細微差別，但總體應該給你良好的開端！

SELECT * 
FROM JS( 
    (// input table 
    SELECT store_nbr, sls_dt, NEST(CONCAT(STRING(item_sku), '|', STRING(sls_amt), '|', STRING(discounts))) AS items 
    FROM (
     SELECT store_nbr, sls_dt, item_sku, sls_amt, GROUP_CONCAT(CONCAT(STRING(disc_nbr), ',', STRING(disc_amt)), ';') AS discounts 
     FROM 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 3456 AS item_sku, 9.99 AS sls_amt, 1 AS disc_nbr, 0.99 AS disc_amt), 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 3456 AS item_sku, 9.99 AS sls_amt, 2 AS disc_nbr, 1.00 AS disc_amt), 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 2345 AS item_sku, 7.99 AS sls_amt, 1 AS disc_nbr, 0.59 AS disc_amt), 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 4567 AS item_sku, 7.99 AS sls_amt, 1 AS disc_nbr, 0.59 AS disc_amt), 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 4567 AS item_sku, 7.99 AS sls_amt, 2 AS disc_nbr, 0.69 AS disc_amt), 
     (SELECT 1234 AS store_nbr, "2014-01-01 00:00:00" AS sls_dt, 4567 AS item_sku, 7.99 AS sls_amt, 3 AS disc_nbr, 0.79 AS disc_amt), 
     (SELECT 2345 AS store_nbr, "2014-01-02 00:00:00" AS sls_dt, 3456 AS item_sku, 9.99 AS sls_amt, 1 AS disc_nbr, 0.99 AS disc_amt), 
     (SELECT 2345 AS store_nbr, "2014-01-02 00:00:00" AS sls_dt, 3456 AS item_sku, 9.99 AS sls_amt, 2 AS disc_nbr, 1.00 AS disc_amt), 
     (SELECT 2345 AS store_nbr, "2014-01-02 00:00:00" AS sls_dt, 4567 AS item_sku, 7.99 AS sls_amt, 1 AS disc_nbr, 0.59 AS disc_amt), 
     GROUP BY store_nbr, sls_dt, item_sku, sls_amt 
    ) GROUP BY store_nbr, sls_dt 
), 
    store_nbr, sls_dt, items, // input columns 
    "[ // output schema 
    {'name': 'store_nbr', 'type': 'INTEGER'}, 
    {'name': 'sls_dt', 'type': 'STRING'}, 
    {'name': 'items', 'type': 'RECORD', 
    'mode': 'REPEATED', 
    'fields': [ 
     {'name': 'sku', 'type': 'STRING'}, 
     {'name': 'sls_amt', 'type': 'FLOAT'}, 
     {'name': 'discounts', 'type': 'RECORD', 
     'mode': 'REPEATED', 
     'fields': [ 
     {'name': 'disc_nbr', 'type': 'INTEGER'}, 
     {'name': 'disc_amt', 'type': 'FLOAT'} 
     ]  
     }]  
    }]", 
    "function(row, emit) { // function 
    var items = []; 
    for (var i = 0; i < row.items.length; i++) { 
     x = row.items[i].split('|'); 
     var discounts = []; 
     y = x[2].split(';'); 
     for (var j = 0; j < y.length; j++) { 
     discount = y[j].split(','); 
     discounts.push({disc_nbr:parseInt(discount[0]), disc_amt:parseFloat(discount[1])}) 
     } 
     items.push({sku:x[0], sls_amt:parseFloat(x[1]), discounts: discounts}); 
    }; 
    emit({ 
     store_nbr: row.store_nbr, 
     sls_dt: row.sls_dt, 
     items: items 
     }); 
    }" 
) 

結果如下

與預期的模式