More than 3 years have passed since last update.

DatabricksAdvent Calendar 2021

@iizukak09(i k)

dbt with Databricksで始めるDataOps(2)

Posted at 2021-12-19

こちらの記事の続きになります。

検証の流れ

環境準備
dbtプロジェクトの作成
dbtのモデル作成と実行
より複雑なモデルの作成と実行 ←この記事はここから
dbtのテスト実行

4. より複雑なモデル作成と実行

1. Databricks SQLでのテーブル準備

まずは、Databricks SQLでテーブルとデータを準備します。

DROP TABLE IF EXISTS zzz_game_opponents;
DROP TABLE IF EXISTS zzz_game_scores;
DROP TABLE IF EXISTS zzz_games;
DROP TABLE IF EXISTS zzz_teams;

CREATE TABLE zzz_game_opponents (
  game_id INT,
  home_team_id INT,
  visitor_team_id INT
) USING DELTA
;

INSERT INTO zzz_game_opponents
  VALUES
    (1, 1, 2),
    (2, 1, 3),
    (3, 2, 1),
    (4, 2, 3),
    (5, 3, 1),
    (6, 3, 2)
;

CREATE TABLE zzz_game_scores (
  game_id INT,
  home_team_score INT,
  visitor_team_score INT
) USING DELTA
;

INSERT INTO zzz_game_scores
  VALUES
    (1, 4, 2),
    (2, 0, 1),
    (3, 1, 2),
    (4, 3, 2),
    (5, 3, 0),
    (6, 3, 1)
;

CREATE TABLE zzz_games (
  game_id INT,
  game_date DATE
) USING DELTA
;

INSERT INTO zzz_games
  VALUES
    (1, '2020-12-12'),
    (2, '2021-01-09'),
    (3, '2020-12-19'),
    (4, '2021-01-16'),
    (5, '2021-01-23'),
    (6, '2021-02-06')
;

CREATE TABLE zzz_teams (
  team_id INT,
  team_city VARCHAR(15)
) USING DELTA
;

INSERT INTO zzz_teams
  VALUES
    (1, "San Francisco"),
    (2, "Seattle"),
    (3, "Amsterdam")
;

2. dbtのモデル作成

modelsディレクトリ配下に以下のようなzzz_game_details.sqlを作成します。

-- 各試合詳細情報を提供するテーブルを作成する
-- スキーマは以下の通り
-- game_id int -- ゲームID
-- home varchar(15) -- ホームチーム
-- visitor varchar(15) -- アウェイチーム
-- home_score int -- ホームチームの点数
-- visitor_score int -- アウェイチームの点数
-- winner varchar(15) -- 勝利チーム
-- date date -- 試合日

{{ config(
  materialized='table',
  file_format='delta'
) }}

with all_ids as (
  -- Step 1 of 4: Combine data from various tables (for example, game and team IDs, scores, dates).
  select
    go.home_team_id,
    g.game_id,
    gs.home_team_score as home_score,
    go.visitor_team_id,
    gs.visitor_team_score as visitor_score,
    g.game_date
  from
    zzz_games as g,
    zzz_game_opponents as go,
    zzz_game_scores as gs
  where
    g.game_id = go.game_id and
    g.game_id = gs.game_id
), visitor_ids as (
  -- Step 2 of 4: Replace the home team IDs with their actual city names.
  select
    game_id,
    t.team_city as home,
    home_score,
    visitor_team_id,
    visitor_score,
    game_date
  from
    all_ids,
    zzz_teams as t
)
select
  game_id,
  home,
  t.team_city as visitor, -- Step 4 of 4: Replace the visitor team IDs with their city names.
  home_score,
  visitor_score,
  -- Step 3 of 4: Display the city name for each game's winner.
  case
    when
      home_score > visitor_score
        then
          home
    when
      visitor_score > home_score
        then
          t.team_city
  end as winner,
  game_date as date
from
  visitor_ids,
  zzz_teams as t
where
  visitor_ids.visitor_team_id = t.team_id
order by game_date desc
;

同様に、zzz_win_loss_records.sqlも作成します。

-- チームごとの勝敗数サマリのビューを作成します。

with sub as (
  -- Step 1 of 2: Determine the winner and loser for each game.
  select
    game_id,
    winner,
    case
      when
        home = winner
          then
            visitor
      else
        home
    end as loser
  from {{ ref('zzz_game_details') }}
)
-- Step 2 of 2: Calculate the number of wins and losses for each team.
select
  winner as team,
  count(winner) as wins,
  -- Each team played in 4 games.
  (4 - count(winner)) as losses
from 
  sub
group by winner
order by wins desc

以下dbtコマンドを実行してテーブルとビューのマテリアライズします。

dbt run --model models/zzz_game_details.sql models/zzz_win_loss_records.sql

00:00:00  Running with dbt=1.0.0
00:00:00  Found 7 models, 4 tests, 0 snapshots, 0 analyses, 190 macros, 0 operations, 0 seed files, 0 sources, 0 exposures, 0 
metrics
00:00:00
00:00:00  Concurrency: 1 threads (target='dev')
00:00:00
00:00:00  1 of 2 START table model dbt_schema.zzz_game_details............................ [RUN]
00:00:00  1 of 2 OK created table model dbt_schema.zzz_game_details....................... [OK in 5.81s]
00:00:00  2 of 2 START view model dbt_schema.zzz_win_loss_records......................... [RUN]
00:00:00  2 of 2 OK created view model dbt_schema.zzz_win_loss_records.................... [OK in 1.47s]
00:00:00
00:00:00  Finished running 1 table model, 1 view model in 14.66s.
00:00:00
00:00:00  Completed successfully
00:00:00
00:00:00  Done. PASS=2 WARN=0 ERROR=0 SKIP=0 TOTAL=2

■zzz_game_detailテーブル

■zzz_win_loss_recordsビュー

5. dbtのテスト実行

dbtのテストは、singular testsという定型の(パラメータを含まない)SELECT文によるワンオフなテストと、
generic testsと呼ばれるパラメータ化されたクエリによるテストの2種類があります。
(generic testsの詳細例)

singular testsは、testsディレクトリ配下にSQL文を配置して定義します。
generic testsは、testブロックでパラメータされたSQL文を、tests/genericかmacrosディレクトリに配置して
さらに、modelsディレクトリ等にYAMLファイルで呼び出して利用します。
いずれのテストでもSELECT文の結果が0件であれば合格(pass)と判定します。

generic testsの定義例

{% test not_null(model, column_name) %}

    select *
    from {{ model }}
    where {{ column_name }} is null

{% endtest %}

また、generic testsには、以下4つの組み込みテストがあり、すぐ利用できるようになっています。

unique: 重複する値が存在しない
not_null: NULL値が存在しない
accepted_value: リストで提供される値のみ取る
relationships: 参照先テーブルの参照カラムに値が存在する（参照整合性）

1. generic tests

以下schema.ymlをmodelsディレクトリ配下に配置します。
testsの項目にgeneric testsの名前を指定することで呼び出しを行えます。

version: 2

models:
  - name: zzz_game_details
    columns:
      - name: game_id
        tests:
          - unique
          - not_null
      - name: home
        tests:
          - not_null
          - accepted_values:
              values: ['Amsterdam', 'San Francisco', 'Seattle']
      - name: visitor
        tests:
          - not_null
          - accepted_values:
              values: ['Amsterdam', 'San Francisco', 'Seattle']
      - name: home_score
        tests:
          - not_null
      - name: visitor_score
        tests:
          - not_null
      - name: winner
        tests:
          - not_null
          - accepted_values:
              values: ['Amsterdam', 'San Francisco', 'Seattle']
      - name: date
        tests:
          - not_null
  - name: zzz_win_loss_records
    columns:
      - name: team
        tests:
          - unique
          - not_null
          - relationships:
              to: ref('zzz_game_details')
              field: home
      - name: wins
        tests:
          - not_null
      - name: losses
        tests:
          - not_null

2. singular tests

testsディレクトリにテスト用のSQLファイルを配置します。

zzz_game_details_check_dates.sql
日付の範囲が正しいかチェックする

-- This season's games happened between 2020-12-12 and 2021-02-06.
-- For this test to pass, this query must return no results.

select date
from {{ ref('zzz_game_details') }}
where date < '2020-12-12'
or date > '2021-02-06'

zzz_game_details_check_scores.sql
点数が負値を取ることがないことをチェックする

-- This sport allows no negative scores or tie games.
-- For this test to pass, this query must return no results.

select home_score, visitor_score
from {{ ref('zzz_game_details') }}
where home_score < 0
or visitor_score < 0
or home_score = visitor_score

zzz_game_details_check_records.sql
勝敗数は0-4の範囲であることをチェックする

-- Each team participated in 4 games this season.
-- For this test to pass, this query must return no results.

select wins, losses
from {{ ref('zzz_win_loss_records') }}
where wins < 0 or wins > 4
or losses < 0 or losses > 4
or (wins + losses) > 4

3. `dbt test`コマンド

dbt testコマンドでテストを実行します。
何もオプションを指定しない場合は、すべてのテストを実行しますが、--selectオプションなどでテスト対象をフィルタすることも可能です。

モデルに「zzz_win_loss_records」を指定、テストタイプはsingular testsのみを指定
（zzz_game_details_check_records.sqlのみ実行される）

dbt test --select zzz_win_loss_records,test_type:singular
00:00:00  Running with dbt=1.0.0
00:00:00  Found 7 models, 23 tests, 0 snapshots, 0 analyses, 191 macros, 0 operations, 0 seed files, 0 sources, 0 exposures, 0 metrics
00:00:00
00:00:00  Concurrency: 1 threads (target='dev')
00:00:00
00:00:00  1 of 1 START test zzz_game_detals_check_records................................. [RUN]
00:00:00  1 of 1 FAIL 2 zzz_game_detals_check_records..................................... [FAIL 2 in 1.72s]
00:00:00
00:00:00  Finished running 1 test in 5.84s.
00:00:00
00:00:00  Completed with 1 error and 0 warnings:
00:00:00
00:00:00  Failure in test zzz_game_detals_check_records (tests\zzz_game_detals_check_records.sql)
00:00:00    Got 2 results, configured to fail if != 0
00:00:00
00:00:00    compiled SQL at target\compiled\dbt_demo_project\tests\zzz_game_detals_check_records.sql
00:00:00
00:00:00  Done. PASS=0 WARN=0 ERROR=1 SKIP=0 TOTAL=1

zzz_win_loss_recordsビューは以下のような結果を返すので、lossesが0未満の値を含むので失敗します。

次にzzz_game_detailsテーブルのgeneric testsを実行してみます。
以下コマンドでzzz_win_loss_recordsモデルの参照整合性テストを含む12のgeneric testsが実行されます。

dbt test --select zzz_game_details,test_type:generic

00:00:00  Running with dbt=1.0.0
00:00:00  Found 7 models, 23 tests, 0 snapshots, 0 analyses, 191 macros, 0 operations, 0 seed files, 0 sources, 0 exposures, 0 metrics
00:00:00  
00:00:00  Concurrency: 1 threads (target='dev')
00:00:00  
00:00:00  1 of 12 START test accepted_values_zzz_game_details_home__Amsterdam__San_Francisco__Seattle [RUN]
00:00:00  1 of 12 PASS accepted_values_zzz_game_details_home__Amsterdam__San_Francisco__Seattle [PASS in 1.15s]
00:00:00  2 of 12 START test accepted_values_zzz_game_details_visitor__Amsterdam__San_Francisco__Seattle [RUN]
00:00:00  2 of 12 PASS accepted_values_zzz_game_details_visitor__Amsterdam__San_Francisco__Seattle [PASS in 1.15s]
00:00:00  3 of 12 START test accepted_values_zzz_game_details_winner__Amsterdam__San_Francisco__Seattle [RUN]
00:00:00  3 of 12 PASS accepted_values_zzz_game_details_winner__Amsterdam__San_Francisco__Seattle [PASS in 1.03s]
00:00:00  4 of 12 START test not_null_zzz_game_details_date............................... [RUN]
00:00:00  4 of 12 PASS not_null_zzz_game_details_date..................................... [PASS in 1.03s]
00:00:00  5 of 12 START test not_null_zzz_game_details_game_id............................ [RUN]
00:00:00  5 of 12 PASS not_null_zzz_game_details_game_id.................................. [PASS in 0.99s]
00:00:00  6 of 12 START test not_null_zzz_game_details_home............................... [RUN]
00:00:00  6 of 12 PASS not_null_zzz_game_details_home..................................... [PASS in 1.01s]
00:00:00  7 of 12 START test not_null_zzz_game_details_home_score......................... [RUN]
00:00:00  7 of 12 PASS not_null_zzz_game_details_home_score............................... [PASS in 1.01s]
00:00:00  8 of 12 START test not_null_zzz_game_details_visitor............................ [RUN]
00:00:00  8 of 12 PASS not_null_zzz_game_details_visitor.................................. [PASS in 0.99s]
00:00:00  9 of 12 START test not_null_zzz_game_details_visitor_score...................... [RUN]
00:00:00  9 of 12 PASS not_null_zzz_game_details_visitor_score............................ [PASS in 1.07s]
00:00:00  10 of 12 START test not_null_zzz_game_details_winner............................ [RUN]
00:00:00  10 of 12 PASS not_null_zzz_game_details_winner.................................. [PASS in 2.20s]
00:00:00  11 of 12 START test relationships_zzz_win_loss_records_team__home__ref_zzz_game_details_ [RUN]
00:00:00  11 of 12 PASS relationships_zzz_win_loss_records_team__home__ref_zzz_game_details_ [PASS in 1.02s]
00:00:00  12 of 12 START test unique_zzz_game_details_game_id............................. [RUN]
00:00:00  12 of 12 FAIL 6 unique_zzz_game_details_game_id................................. [FAIL 6 in 1.02s]
00:00:00
00:00:00  Finished running 12 tests in 17.01s.
00:00:00
00:00:00  Completed with 1 error and 0 warnings:
00:00:00
00:00:00  Failure in test unique_zzz_game_details_game_id (models\schema.yml)
00:00:00    Got 6 results, configured to fail if != 0
00:00:00
00:00:00    compiled SQL at target\compiled\dbt_demo_project\models\schema.yml\unique_zzz_game_details_game_id.sql
00:00:00
00:00:00  Done. PASS=11 WARN=0 ERROR=1 SKIP=0 TOTAL=12

You get articles that match your needs
You can efficiently read back useful information
You can use dark theme

What you can do with signing up