Stringly typed

Kinda started using that with polars, force the scan/read schema as string, and down the road cast as the necessary type when needed. I’m calling it “just in time” data cleaning.

Everithing is bytes.

some of you have never programmed in tcl and it shows

Remember Tcl

Simply make everything an array of bytes

God, I’m so over SQL.

It’s great, but it is so old and shows it. Feels like 99% of my SQL queries are just cheese.

Works though, and quick.

I opened a TCL script once. It’s use of uplevel scared me. I’ve never dared to return since.

For those who don’t know: uplevel is a command that goes up one level of the stack frame, and then executes code there. A function can therefore execute code in its callers stack frame.

Oh, you worked at Oracle by any chance?

SQL is the only bedrock in my entire career. Its the one thing that has stayed relevant.

SQL is great but when you start having issues processing what is actually going on, its fine to pull out what you need and throw another language on top (python, C#, etc…etc…). Getting it to work slow is one step in making it fast again.

Which language can nil an int?

Just cast it. /s

Yeah it’s curious that it hasn’t really undergone some major changes or had some major challengers (except NoSQL I guess).

It is also the bash approach, isn’t it?!

Its been a while but yeah NoSQL was the closest.

I remember a good 4-5 years where developers all around me were using couchdb, mongodb, and a host of others. mostly json in <-> json out kind of systems. And VERY hard to maintain after the initial TODO. I remember so much debugging and finding out old records didnt have a way to deal with changes in the “tables” or equivalents. It was maddening.

Dont get me wrong, it did create some really awesome specialty tools but you cant really get around ACID compliance when dealing with databases.

I think SQL has some awesome properties that keep it going:

1. Most major distributions are rock solid stable.
2. Its optimized and fast for data.
3. Its understandable to many types of industries. Software development is only the start.
4. Its integrated with everything already. So ODBCs can just plug and play most of the time.
5. Its the devil we know. ACID, transactions, etc… are all things we know about and are proven to work very well. Definitly when you need to MAKE SURE a thing made its way into the system.

I have. I quickly learned not to.

Tk is overlooked, though. It’s not pretty, and its approach is archaic, but it’s one of the few GUI toolkits that Just Works on every platform I tried it on with minimum fuss.

Yeah 100% with you, had this mongo database where the first entry was like a description, the nr 2 and on the actual data. I mean if there were a description… Sometes 2 descriptions…

Why oh why.

And for sure SQL is kind of the cement of DB today, don’t get me wrong, I like that what I learned yesterday actually still works, I’m just pondering the fact that it is so.

Maybe SQL isn’t the hip language so people doesn’t try to reinvent it all the time

I fucking love tcl

I took great pains last week to convert a big python project to make it typed. (shoutout to MonkeyType)

It’s so much nicer to develop now…

It has though

Window functions were an addition, but more recently struct, json, and array fields with native support. Pipe syntax is getting multiple implementations.

Match recognize is a whole new standard abstraction of window functions.

Union by name is being added (fuck union by position).