My favorite change is that MariaDB got an improved implementation of the WHERE year(...) = ? rewriting. use-the-index-luke.com/blog/2024-12...

In picture: the green check mark used to be yellow, with the "a" footnote: Unconditionally rewritten — even if only a matching function-based index exists

I wish I was running my business in a garage. I'd have more space there.

Related: New mugs arrived: use-the-index-luke.com/shop#mugplus...

I think the reason for that is the legal dispute that caused the SEQUEL→SQL renaming. The reunion paper (pic) doesn't say it explicitly, but I'd guess that the removing the vowels trick requires a different pronunciation too (but that is just my interpretation).

www.mcjones.org/System_R/SQL...

The article just says how it is typically pronounced — but doesn't say what is correct.

I, for myself, refer to the standard itself. Part 1 has a clear answer (picture).

Big Data was yesterday. Today we've got Colossal Data!

www.commitstrip.com/en/2014/06/0...

Just because the comic is old doesn't mean it's wrong.

TIL: MySQL's TIMESTAMP doesn't even provide the full 32bit epoch as it starts one second after 1970-01-01 00:00:00Z.

dev.mysql.com/doc/refman/9...

(1) It’s slow; (2) is gives you wrong results. What else do you need?

The main problem is that the number of rows you have seen in the past cannot be used to find the spot where you stopped over the last time. Since then, new rows might have arrived.

JSON support in SQL is not about storing JSON documents¹.

While storing JSON in SQL tables is possible and sometimes quite useful, the true power of JSON in SQL comes from the ability to *process* JSON within SQL.

Just scheduled the first round of trainings 2025.

Now also offering a US shift (9am pacific).

Also pimped my concurrency training. Read more at my website:

winand.at/newsletter/2...

For those of you who like stickers.

I'll ship mine to you—totally free. Coasters will also be in the package.

use-the-index-luke.com/shop

SQL's data types CHAR and VARCHAR are about characters. Sounds like a no-brainer, doesn't it?

Well, a consequence is that the number in the parenthesis refer to characters, not bytes.

Unfortunately, some SQL engines don't get that right—causing major confusion among users.

SQL's FILTER clause should be easy to implement for every engine. Why is it still not supported widely?

Sure it can be emulated easily—and ugly. The value on readability (and learning) is huge.

Please vendors: Just get it done.

modern-sql.com/feature/filter

For this WHERE, what index to create?

WHERE subsidiary_id = 42
AND last_name LIKE '%INA%'

Is (subsidiary_id) enough or should it be (subsidiary_id, last_name)?

A decade ago I prepared an animation for that.

use-the-index-luke.com/sql/clusteri...

😱

Sneak peek. First here on Blue Sky — the others have to wait (probably some month).

Can you guess about which SQL-Feature this chart is about?

Last orders please!

The next "SQL Reloaded" training has some seats left. The training covers a lot what you already know, but much deeper than you every thought about it.

Online, as a series of six two-hour units.

Starting Jan 9th.
More info: winand.at/sql-training...

On the way to pgconf.eu.

Also: I hope to win the ”most outdated profile picture“ competition.