Our ex-employee wrote an amazing SQL SELECT-query consisting of 6449 characters. It has 11 JOINS and takes a solid minute to execute.

The table it fetches from has 16 records and the SQL query returns 46857 records and it was production code lmao

Wtf? 16 records? That doesn‘t even justify a DB!

Databases are surprisingly fast

And then people do god forsaken things like this and then they aren't

@Lensflare They've been waiting for the 17th record to insert since 2018.

@Lensflare it doesn't even justify saying the word DB in the convo.

Cache that shit and work with it.

@Lensflare 16 records in that table of a new feature that is in prod but a few customers have access to it.

is he trying to code using SQL

has he tried just using the CPU

cloud computing oficiados I swear

What- What does it even do?!

Joins against one table with 16 records eh? Impressive!

The only stat that doesn’t completely knock me on my ass is the 11 joins. The rest are absolutely unbelievable.

@jestdotty I have a longstanding argument in favour of key-value DBs; they force all decisions onto the compute node so storage is reducible to a pile of network and storage hardware and the smallest OS that can run the DB. Sure an individual task is drastically slower, but they can be parallelized much better. Also I think dealing with databases as data you put in an index rather than indexes you create over data builds character. It's not entirely serious either way but I think it would be interesting to see the kind of practices that evolve when you execute _zero custom logic_ on the storage nodes.

Are you storing in JSON or something and the poor DB has to parse it?

@AlgoRythm Around 5k characters is just stuff like "status"."id" AS "status_id". The word " AS " is found 100 times in the query.

I'm not a back-end developer but that looks like shit.

@Ranchonyx It retrieves data about conversations, related tickets, priorities, categories, ...

Front-end colleague fixed the query and executation time dropped to 0.085 seconds. It now runs around 835x faster and the output is correct.

@wojtek322 aliases are dead common in ETL, less common but still commonplace in regular application use. It’s a code smell, well-built applications might have none at all.

Why does this sound familiar?