sql Treasure Data&（一部Presto）の知っておいて便利だった关数ref：https：//qiita.com/inuscript/items/51b9b3c58260d823

Posted 2021-05-13

SELECT *
FROM access TABLESAMPLE BERNOULLI(1)

SELECT 
  some_name,
  map_keys(
    histogram(some_type)
  )
FROM foo
GROUP BY some_name

SELECT TD_SESSIONIZE(time, 3600, ip_address) as session_id
FROM ...
ORDER BY ip_address, time 
-- ↑ソートしないと正しく出ない！

SELECT 
  *
  TD_PARSE_AGENT(user_agent)['ua_family'] AS family,
  TD_PARSE_AGENT(user_agent)['ua_major'] AS version
FROM
  access
WHERE TD_PARSE_AGENT(user_agent) IN ['pc', 'smartphone']

WITH a AS (
  SELECT 
    TD_PARSE_AGENT(user_agent)['ua_family'] AS family,
    TD_PARSE_AGENT(user_agent)['ua_major'] AS version
  FROM 
    access
),
b AS (
  SELECT
    if(family = 'IE', family || version, family) AS browser
  FROM a
)
SELECT * FROM b

SELECT
  TD_PARSE_AGENT(user_agent)['os_family']
  TD_PARSE_AGENT(user_agent)['os_major']
  TD_PARSE_AGENT(user_agent)['os_minor']
  TD_PARSE_AGENT(user_agent)['ua_family']
  TD_PARSE_AGENT(user_agent)['ua_major']
  TD_PARSE_AGENT(user_agent)['ua_minor']
  TD_PARSE_AGENT(user_agent)['device']
FROM access

SELECT *
FROM access
WHERE 
  url_extract_host(referer) = "google.com"
  -- 正規表現でやるとこんな感じ
  -- regexp_like(referer, '.*google.com/.*')

SELECT
  :
WHERE 
  -- 1日間のデータなら -1d。一ヶ月なら-30dとか。
  TD_TIME_RANGE(time,
    TD_TIME_ADD(TD_SCHEDULED_TIME(), '-1d', 'JST'), 
    null,
    'JST'
  )