How Modern SQL Databases Come up with Algorithms that You Would Have Never Dreamed Of by Lukas Eder

SQL is the only ever successful, mainstream, and general-purpose 4GL (Fourth Generation Programming Language) and it is awesome!
With modern cost based optimisation, relational databases like Oracle, SQL Server, PostgreSQL finally keep up to the promise of a powerful declarative programming model by adapting to ever changing productive data without performance penalties. Thousand-line-long, complex SQL statements can be run in far below a millisecond against billion-row strong tables if database developers know their ways around the SQL language - and the best news is: It's not that hard!
In this talk, I'll show how the SQL database will constantly outperform any hand written data retrieval algorithm - or in other words - how SQL, being a logic language, is the best language for business logic.

[@danieldiaz6025]

This is a great talk. I learned new things about window functions from the query at 18:36 (it seems that the OVER clause has an implicit "RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW", which is possibly the default behaviour when you include an ORDER BY).

[@yankhonski9708]

The talk is great!
I think each tool has its purpose.
SQL is used for data manipulation in relational databases.
Java is used for building server side business logic, REST APIs.
Javascript is for UI side (mostly).
I could also use Kafka streams to solve this problem: each time I receive a payment event, I store it into the database and also I can process it separately. It is useful for large scale applications with millions of writes and replications.

[@AndrewBrownK]

I love this talk so much, it never gets old

[@mesutkayman]

Excellent talk

[@kamillatosinski3054]

I died when the ISS part came in :D

[@magfal]

27:20 the select syntax is an early approximation of natural language parsing.
Think of it like how one would ask a warehouse employee to go fetch something for you in a complex set of storage locations, assuming that the person has little knowledge of the contents of the storage locations.
Retrieve these items from the following locations but only the ones that match these criteria.

[@tomknud]

Great discussion, thanks

[@gggrow]

10 minutes in, I haven't learned that SQL is awesome yet but I have learned that Java is horrible

[@Technomancr]

I dunno, his arguments are kinda flawed. Like he spends 5 minute talking about getters/setters/hash/equals, which no one ever writes by hand. They just use lombok for. And then 30 seconds later he uses lombok in another place. Why not just use lombok in the first example? Oh right, he's writing bad java code on purpose to try and make a point.

[@illyam689]

He is obviously exaggerating things

[@coolcoolj0102]

Still really useful presentation to watch on 2019 : )

[@hantt]

pants are always greener in SQL

[@seonwookim9595]

Love the talk, learned a lot, great presentation

[@donbeave]

Amazing talk!

[@perrar1na]

Thank you

[@lts8683]

Very useful

[@kingluke2]

So glad I gave up object-oriented programming years ago. Creates so much unnecessary complexity.

[@andy12829]

Java Champion telling about SQL 😎

[@wotwot6868]

If you go even further. You'll end up with Clojure.

[@lukas_eder]

Indeed, Clojure lends itself to really elegant programming styles. I've seen a talk about Datalog and I was thrilled. Very beautiful!

[@AndrewBrownK]

And lose type safety, no thanks

[@133289ify]

@@AndrewBrownK f*ck types ;)

[@coolcoolj0102]

may I know why?

[@gnatinator]

8:20 ... jfc java. Pretty contrived example, though-- most languages aren't this bad.

[@clray123]

annotatiomania.com/ is the shit... annotations and the resulting "google it up, copy paste and maybe it will work" instead of understanding code is the worst thing that has happened to Java during its entire history.. what was once a single language has been turned into 100 different fucked up dialects.

[@clray123]

He's exaggerating a bit about Java, but his point is true. Can't count time when I've been facepalming myself because of all this boilerplate bullshit and HighBrowTheoryOfHowThingsShouldBeDone when all you need is to shuffle around some fucking bits of data (which happens to be what 99% of webapps consist of).
And yes my advice to Java developers is throw away your fugly Hibernate whenever you can, hardcode the simple SQL statements and be done with the bs.

[@JerreMuesli]

+1 for flaming Lombok

[@coolcoolj0102]

lombok is useful. He just flamed Hibernate annotation style

[@133289ify]

tldw: java proxy bean factories suck

[@igo_rs]

Using the worst java architecture for comparison is pathetic.

[@sergeibatiuk3468]

Wow, these gifs are so annoying

[@ShekharGulati7]

Bad talk. Talk about SQL please. Looks like we are learning Java

[@thrael3]

SQL database is a liablity. It is number one reason why we have to duplicate expensive production database setup in test and catch frequent problems with incorrect execution plans.

[@lukas_eder]

Yes, sometimes those execution plans don't behave. Sometimes. I'm sure that, given your opinion, you have carefully evaluated how often your hand written algorithms would not behave correctly if you *didn't* use SQL, and how much time you'd spend on fixing that, instead? SQL works 99.9% of the time and we spend some time fixing that remaining 0.1% (arbitrary numbers of course). How would that compare to hand written algorithms? And how much would it cost to write them?
But I'm just repeating my talk :)