非結構化資料

BigQuery 入門

Matt Forrest

Field CTO

為何非結構化資料很重要

非結構化資料範例：

/*--------------+-----------------------------------------*
 | customer_id  | emails                                  |
 +--------------+-----------------------------------------+
 | 12345        | ['[email protected]', '[email protected]']   |
 *--------------+-----------------------------------------*/

ARRAY

SELECT ARRAY
  (SELECT 'bigquery' UNION ALL
   SELECT 'analytics' UNION ALL
   SELECT 'sql') AS new_array;


/*------------------------------------*
 | new_array                           |
 +------------------------------------+
 | ['bigquery', 'analytics', 'sql']   |
 *-----------------------------------*/

SELECT ARRAY
  (SELECT 'bigquery' UNION ALL
   SELECT 'analytics' UNION ALL
   SELECT 'sql')[1] as result;


/*---------------*
 | result          |
 +---------------+
 | 'analytics'   |
 *---------------*/

STRUCT

SELECT
  STRUCT

  <skill string, learning bool>
  ('big query', true)    
  as skills;


/*--------------+--------------*
 | skills.skill | learning     |
 +--------------+------------- +
 | 'big query'  | true         |
 *--------------+--------------*/

SELECT
  STRUCT
  <skill string, learning bool>

  ('big query', true).skill    
  as key;


/*--------------+
 | key.         |
 +--------------+
 | 'big query'  |
 *--------------*/

ARRAY_LENGTH 與 ARRAY_CONCAT

SELECT
  ARRAY_LENGTH(
    [
      '[email protected]', 
      '[email protected]'
    ]) as len;

/*------*
 | len  |
 +------+
 | 2    |
 *------*/

SELECT
  ARRAY_CONCAT(
    ['one'], ['two']) as new_array;

/*-------------------*
 | new_array         |
 +-------------------+
 | ['one', 'two']    |
 *-------------------*/

UNNEST

將 email 資料展開（UNNEST）：

SELECT 
    * 
FROM
    UNNEST(['[email protected]', '[email protected]']) as emails;

/*-------------------+
 | emails            |
 +-------------------+
 | '[email protected]' |
 | '[email protected]'  |
 *-------------------*/

UNNEST 搭配 STRUCT

範例資料－ARRAY 中的 STRUCT：

[
    {'big query': true},
    {'sql': true}
]

SELECT 
    my_skills.skill, 
    my_skills.learned
FROM 
    UNNEST(skills) as my_skills

/*--------------+--------------*
 | skill        | learned      |
 +--------------+------------- +
 | 'big query'  | true         |
 +--------------+------------- +
 | 'sql'        | true         |
 *--------------+--------------*/

SEARCH

在 email 資料上使用搜尋：

SELECT 
    SEARCH(['[email protected]', '[email protected]'], 'gmail.com') as results;

/*----------*
 | results  |
 +----------+
 | true     |
 *----------*/

非結構化資料小抄

ARRAY

類似有序列表
以 0 作為起始索引存取值（例如：my_array[0]）

STRUCT

類似 JSON 或字典
可含多種型別與任意結構
該欄所有列的結構必須相同

ARRAY_CONCAT/ARRAY_LENGTH

串接多個陣列並取得陣列長度

UNNEST

將 ARRAY 資料攤平

SEARCH

在 ARRAY 或 STRUCT 中搜尋符合的值

一起來練習吧！

BigQuery 入門