https://github.com/airbytehq/sql-levels-explained

explaining sql levels based on one meme
https://github.com/airbytehq/sql-levels-explained

sql

Last synced: 3 days ago
JSON representation

explaining sql levels based on one meme

• Host: GitHub
• URL: https://github.com/airbytehq/sql-levels-explained
• Owner: airbytehq
• License: mit
• Created: 2022-08-17T16:07:03.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2023-08-27T13:09:30.000Z (about 1 year ago)
• Last Synced: 2024-08-05T15:07:18.280Z (3 months ago)
• Topics: sql
• Homepage: https://twitter.com/largedatabank/status/1559651463919452161
• Size: 12.7 KB
• Stars: 97
• Watchers: 20
• Forks: 9
• Open Issues: 7
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# SQL-Levels-Explained

Explaining sql levels based on one meme:



If you're replicating SQL databases, whether MySQL, Postgres, or SQL Server, [check out Airbyte](https://www.theseattledataguy.com/what-is-airbyte-and-why-you-should-use-it/)!

## Level 0

- Concepts

  - ORMs

  - data types

  - foreign keys

  - indexes

- SQL: Just go through https://mode.com/sql-tutorial/

  - SELECT/INSERT/UPDATE/DELETE

    - SELECT DISTINCT

  - ORDER BY

  - GROUP BY

  - LIMIT/OFFSET

  - CREATE TABLE

  - JOIN

## Level 1

- Concepts

  - ACID

  - transactions

  - keyset pagination

  - normal forms

  - computed columns

  - stored columns

  - inverted indexes

  - window functions

- SQL

  - outer joins

  - ORDER BY in aggregates

  - CTEs

  - query plans and EXPLAIN

## Level 2

- Concepts

  - Connection pools

  - plan hints

  - cursors

  - MVCC garbage collection

  - recursive CTEs

  - LATERAL joins

- Insights

  - ORMs create bad queries: *explanation needed*

  - optimizers don't work without table statistics: *explanation needed*

  - there are no non-nullable types: *explanation needed*

  

## Level 3

- Concepts

  - MERGE

  - COUNT(*) vs COUNT(1): https://svenweller.wordpress.com/2016/10/11/otn-apprecition-day-the-otn-forum/

  - isolation levels

  - zigzag join

  - grouping sets, cube, rollup

  - write skew

  - partial indexes

  - sharding

  - phantom reads

  - triggers

- Insights

  - serializable restarts require retry loops on all statements: *explanation needed*

  - generator functions zip when cross joined: *explanation needed*

  

## Level 4

- Concepts

  - SELECT FOR UPDATE

  - denormalization

  - transaction contention

  - sargability

  - star schemas

  - utf8mb4

- Insights

  - Ascending Key Problem: *explanation needed*

  - Ambiguous network errors: *explanation needed*

  - NULLs in CHECK constraints are truthy: *explanation needed*

  

## Level 5

- Concepts

  - DEFERRABLE INITIALLY IMMEDIATE

  - MATCH PARTIAL foreign keys

  - EXPLAIN approximates SELECT COUNT(*)

  - causual reverse

- Insights

  - TPCC requires wait times: *explanation needed*

  - cost models don't reflect reality: *explanation needed*

  - 'null'::jsonb IS NULL = false: *explanation needed*

## Level 6

  

- Concepts

  - database cracking: *explanation needed*

  - learned indexes: *explanation needed*

  - XTID exhaustion: *explanation needed*

  - Worst Case Optimal Join: *explanation needed*

  - Volcano model: *explanation needed*

- Battle Scars

  - Vectorized doesn't mean SIMD: *explanation needed*

  - join ordering is NP hard: *explanation needed*

  - NULLs are equal in DISTINCT but inequal in UNIQUE: *explanation needed*

## Level 7

- dee and dum: https://blog.jooq.org/creating-tables-dum-and-dee-in-postgresql/

- the Halloween problem: https://en.wikipedia.org/wiki/Halloween_Problem

- allballs: https://www.postgresql.org/message-id/24526.1106597936%40sss.pgh.pa.us

- fsyncgate: *explanation needed*

- Every SQL operator is actually a JOIN: *explanation needed*

- SERIAL is non-transactional: *explanation needed*

- NULL (?): *explanation needed*