Query analytics for the day-to-day developer with MySQL 8or around 100MB/table Now supports terabyte-sized databases Supports SQL standards as new as SQL 2016 ... available (Window

Query analytics for the day-to-day developer with MySQL 8.0

Gabriela D'Ávila Ferrara @gabidavila Developer Advocate @ Google Cloud ^(gabi|gabby)\.dev$

I try to solve all my problems with a single SQL query

Is it always pretty?

FizzBuzz

WITH RECURSIVE fizz_buzz (sequence, modulo_3, modulo_5) AS ( SELECT 1, CAST('' AS CHAR(4)), CAST('' AS CHAR(5)) UNION ALL SELECT sequence + 1, IF(MOD(sequence + 1, 3) = 0, 'Fizz', ''), IF(MOD(sequence + 1, 5) = 0, 'Buzz', '') FROM fizz_buzz WHERE sequence < 100 )

SELECT IF( CONCAT(modulo_3, modulo_5) = '', sequence, CONCAT(modulo_3, modulo_5)) AS fizzbuzz FROM fizz_buzz;

Brief History

● Created to handle from up to 10 to 100M rows or around 100MB/table

● Now supports terabyte-sized databases

● Supports SQL standards as new as SQL 2016

● But... some stuff from SQL 2003 just became available (Window Functions, CTEs)

MySQL 8.0.16

Be warned!

Don’t do this in production!

● Just because you can, doesn’t mean you should!

● The cost of analytical queries are too high ● Run on a replica if you must to ● Remember this is just to help you in an

emergency

Demo

Considerations

How often you write in raw SQL…

● A CRUD operation? ● A DML operation? ● Create/edit a Function? ● Create/edit a Procedure? ● A View?

Have you ever…

● Generated reports using a scripting language? (Python, PHP)

● Did an ETL? ● Synced data across different types of

databases? (i.e. full-text search)

Scenario

Let’s say you own a store

🏬 💵 💻

Tables: •products •users •orders •order_items

And you want several reports

● Most expensive order per user

● The highest price each product was ever sold and what was the date

of that ● The monthly amount

sold in a year together with the growth

Subqueries

SHOW CREATE TABLE `users`;

CREATE TABLE `users` ( `id` bigint(20) UNSIGNED NOT NULL AUTO_INCREMENT, `username` varchar(255) NOT NULL, `created_at` timestamp NOT NULL, `updated_at` timestamp NOT NULL )

SELECT * FROM `users`;

Who never did this?

SELECT users.id, users.username, (SELECT id FROM orders WHERE users.id = user_id ORDER BY total LIMIT 1) AS order_id, (SELECT total FROM orders WHERE users.id = user_id ORDER BY total LIMIT 1) AS order_total FROM users ORDER BY users.id LIMIT 10;

The most expensive order for each user

LATERAL

SELECT users.id, users.username, total_orders.id AS order_id, total_orders.total AS order_total FROM users, LATERAL( SELECT id, total FROM orders WHERE users.id = user_id ORDER BY total LIMIT 1 ) AS total_orders ORDER BY users.id LIMIT 10;

Demo

OLTP vs OLAP

OLTP: Online Transaction Processing

● Transactional ● Single-source of truth ● Normalized (usually) ● Examples: MySQL, PostgreSQL, SQL

Server, etc.

OLAP: Online Analytical Processing

● Aggregations ● Star-Schema, Cube ● Forecasts ● Examples: Redshift, Pentaho*

Query analytics? 🤔

Analytical clause, up to 4 parts

● Analytical function ● Partitioning Clause ● Order by Clause ● Windowing Clause

Window Functions

What they do?

● Allows to analyze the rows of a given result set

● Can behave like a GROUP BY without changing

the result set

● Allows you to use a frame to "peek" OVER a

PARTITION of a window

Window Functions

● Examples:

● Enumerate rows - ROW_NUMBER()

● Show Aggregated sums - SUM()

● Rank results - RANK(), DENSE_RANK()

● Look at neighboring rows - LEAD(), LAG()

SHOW CREATE TABLE `orders`;CREATE TABLE `orders` ( `id` bigint(20) UNSIGNED NOT NULL AUTO_INCREMENT, `user_id` bigint(20) UNSIGNED NOT NULL, `status` varchar(20) NOT NULL DEFAULT 'new', `total` decimal(10,2) NOT NULL, `created_at` datetime NOT NULL, `updated_at` datetime NOT NULL )

SELECT … FROM `orders` WHERE … ORDER BY created_at

+---------+---------+-----------+---------------------+ | id | user_id | status | created_at | +---------+---------+-----------+---------------------+ | 6534954 | 654321 | canceled | 2011-08-21 19:06:59 | | 6534949 | 654321 | canceled | 2013-11-16 09:02:19 | | 6534953 | 654321 | completed | 2015-10-28 02:21:30 | | 6534951 | 654321 | new | 2018-01-05 17:12:23 | | 6534952 | 654321 | new | 2018-07-19 04:23:41 | | 6534955 | 654321 | new | 2018-11-12 05:37:48 | | 6534950 | 654321 | pending | 2018-12-20 06:11:23 | +---------+---------+-----------+---------------------+

SELECT … FROM `orders` WHERE … ORDER BY created_at

+---------+---------+-----------+---------------------+---------------------+---------------------+ | id | user_id | status | previous_order | created_at | next_order | +---------+---------+-----------+---------------------+---------------------+---------------------+ | 6534954 | 654321 | canceled | NULL | 2011-08-21 19:06:59 | 2013-11-16 09:02:19 | | 6534949 | 654321 | canceled | 2011-08-21 19:06:59 | 2013-11-16 09:02:19 | 2015-10-28 02:21:30 | | 6534953 | 654321 | completed | 2013-11-16 09:02:19 | 2015-10-28 02:21:30 | 2018-01-05 17:12:23 | | 6534951 | 654321 | new | 2015-10-28 02:21:30 | 2018-01-05 17:12:23 | 2018-07-19 04:23:41 | | 6534952 | 654321 | new | 2018-01-05 17:12:23 | 2018-07-19 04:23:41 | 2018-11-12 05:37:48 | | 6534955 | 654321 | new | 2018-07-19 04:23:41 | 2018-11-12 05:37:48 | 2018-12-20 06:11:23 | | 6534950 | 654321 | pending | 2018-11-12 05:37:48 | 2018-12-20 06:11:23 | NULL | +---------+---------+-----------+---------------------+---------------------+---------------------+

Previous and Next orders | LAG and LEAD

SELECT id, user_id, status, LAG(created_at) OVER(ORDER BY created_at) AS previous_order, created_at, LEAD(created_at) OVER(ORDER BY created_at) AS next_order FROM orders WHERE user_id = 654321 ORDER BY created_at LIMIT 10;

Break down

windowfunction

column

# rows preceding

LAG(created_at, 1) OVER (ORDER BY created_at)

Repetition?

SELECT id, user_id, status, LAG(created_at) OVER(ORDER BY created_at) AS previous_order, created_at, LEAD(created_at) OVER(ORDER BY created_at) AS next_order FROM orders WHERE user_id = 654321 ORDER BY created_at LIMIT 10;

Named Windows!

SELECT id, user_id, status, LAG(created_at) OVER(dates) AS previous_order, created_at, LEAD(created_at) OVER(dates) AS next_order FROM orders WHERE user_id = 654321 WINDOW dates AS (ORDER BY created_at) ORDER BY created_at LIMIT 10;

Demo

CTE Common Table Expressions

CTE

Common Table Expressions

● Similar to CREATE [TEMPORARY] TABLE

● Doesn’t need CREATE privilege

● Can reference other CTEs (if those are already defined)

● Can be recursive

● Easier to read

Recursive CTE

● Useful with hierarchical data

● The Recipe is:

● Base query comes first

● Second query comes after an UNION statement

● And the stop condition should be on the recursive call

FizzBuzz

WITH RECURSIVE fizz_buzz (sequence, modulo_3, modulo_5) AS ( SELECT 1, CAST('' AS CHAR(4)), CAST('' AS CHAR(5)) UNION ALL SELECT sequence + 1, IF(MOD(sequence + 1, 3) = 0, 'Fizz', ''), IF(MOD(sequence + 1, 5) = 0, 'Buzz', '') FROM fizz_buzz WHERE sequence < 100 )

SELECT IF( CONCAT(modulo_3, modulo_5) = '', sequence, CONCAT(modulo_3, modulo_5)) AS fizzbuzz FROM fizz_buzz;

SELECT * FROM `store`.`categories`;

+------+-------------------------------+--------------------+ | id | name | parent_category_id | +------+-------------------------------+--------------------+ | 1 | Animal | 0 | | 2 | Plant | 0 | | 3 | Dog | 1 | | 4 | Cat | 1 | | 5 | Tulip | 10 | | 6 | West Highlander White Terrier | 12 | | 7 | Lettuce | 11 | | 8 | Sunflower | 10 | | 10 | Flowers | 2 | | 11 | Veggies | 2 | | 12 | Terrier | 3 | +------+-------------------------------+--------------------+ 11 rows in set (0.00 sec)

Demo

Recursive CTE

WITH RECURSIVE tree (depth_level, node, path, node_id) AS ( SELECT 1, CAST('root' AS CHAR(255)), CAST('root' AS CHAR(65535)), 0 ) SELECT * FROM tree;

Recursive CTE

WITH RECURSIVE tree (depth_level, node, path, node_id) AS ( SELECT 1, CAST('root' AS CHAR(255)), CAST('root' AS CHAR(65535)), 0 UNION ALL SELECT tree.depth_level + 1, categories.name, CONCAT_WS('/', tree.path, categories.name), categories.id FROM tree INNER JOIN categories

ON tree.node_id = categories.parent_category_id WHERE tree.depth_level < 5 ) SELECT * FROM tree ORDER BY path;

Thank you!

● Twitter: @gabidavila ● Co-host at GCP Podcast: gcppodcast.com ● Website: gabi.dev