This is possibly one of the more cursed single sentences I've ever seen in a job posting

SpookJ 👻

@ireneista @xgranade @aud @glyph my direct supervisor in grad school wrote F95 like it was a second language, ended up with a parallel track of development of a Runge-Kutta solver that would do the same thing as the "normal" C++ code I was trying to press into the task we needed.

My stance is that FORTRAN is Good, Actually. The design scope was the right size to begin with, and there's just no arguing with what it takes to compete with LAPACK et al.

Fun to write, though? Absolutely not.

Xandra Granade 🏳️‍⚧️

@SnoopJ @ireneista @aud @glyph If it was zero-based, had a sensible syntax for declaring function parameters, didn't have the weird function/subroutine split, and didn't hard crash if you add print statements inside a function called from a print statement, then I'd likely even agree.

There's a lot to like about FORTRAN, but it just has too many papercuts stemming from being one of the first languages period, imho.

SpookJ 👻

@xgranade did you ever run into Forthon?

That was another Grad School Adventure, and my encounter with it was what taught me that Python 2 lets you mutate the values of True, False.

@ireneista @aud @glyph

Asta [AMP]

@[email protected] @[email protected] @[email protected] @[email protected] sounds like we need to write FORCIS, then, as a... actually nevermind.

Despite not knowing FORTRAN, I have to agree with the assessment that it is good; there is a reason we continue to use FORTRAN libraries wrapped up in other codes.

#chapelLang (geez Audrey shut up about Chapel) used to default to indexing by 1 (it now uses 0 indexing, as it well should) but also would let you define the index however the fuck you wanted. I think it still does. You want -10 to -40? It's yours, baby. Do it. Who can stop you?

Xandra Granade 🏳️‍⚧️

@SnoopJ @ireneista @aud @glyph No, but I was excited about FORTRESS for a long time. Sun's attempt to modernize FORTRAN, and it included so many good ideas, like you could declare which ring and field axioms a data type obeyed, which would then automatically generate unit tests and give the compiler extra assumptions.

Each program also had a canonical, unambiguous transformation to a LaTeX document, making it easy to embed FORTRESS in papers.

Irenes (many)

@aud @SnoopJ @glyph @xgranade no, no, the problem is that it's FORTRAN when it should be BYTRAN.........

Irenes (many)

@xgranade @SnoopJ @aud @glyph whoa. ring and field axioms. that is a good idea.

SpookJ 👻

@xgranade @ireneista @aud @glyph neat, TIL

Xandra Granade 🏳️‍⚧️

@SnoopJ @ireneista @aud @glyph It got abandoned in very early alpha versions, but it was fucking cool in how it was designed.

Xandra Granade 🏳️‍⚧️

@ireneista @aud @SnoopJ @glyph POLYENBY (am I doing this right?)

Anyway, of all things, Visual fucking Basic let you redefine array bases on a compilation-unit level, which created no small amount of bugs.

Asta [AMP]

@[email protected] @[email protected] @[email protected] @[email protected] I wish literally anyone except for Oracle had purchased Sun.

Xandra Granade 🏳️‍⚧️

@ireneista @SnoopJ @aud @glyph IKR? I keep thinking that really just needs to be in any new language. generalize it so that you can just list parameterized invariants. But that just gets back to hypothesis and similar libraries... the ring and field axioms were even cooler than that for what information they gave the compiler.

Kevin Riggle

@SnoopJ @xgranade @aud What I really want is the LLM to translate idiomatic COBOL to idiomatic Java and then the transpiler and some formal methods to prove that the idiomatic Java is identical to the idiomatic COBOL, but that sounds like hard work and real computer science and software engineering so it’s probably not what they’re doing

Asta [AMP]

@[email protected] @[email protected] @[email protected] @[email protected] oh Jesus. Chapel just basically exposed the range map to the user, should you wish it, for any individual array. I think you could still compile assuming the default index started at 1 as an option but that was just during the change; I don’t think you could arbitrarily change them all at the compile level. Oooof.

Irenes (many)

@xgranade @SnoopJ @aud @glyph yeah absolutely

Asta [AMP]

@[email protected] @[email protected] @[email protected] yeah, that actually sounds like it would be useful and work! So definitely not what they’re doing.

Xandra Granade 🏳️‍⚧️

@ireneista @SnoopJ @aud @glyph Slide 116 or so of https://web.archive.org/web/20060819201513/http://research.sun.com/projects/plrg/PLDITutorialSlides9Jun2006.pdf has a good summary.

_NetNomad

@SnoopJ @aud @glyph @xgranade most of our codebase where i work is COBOL and every few years we start talking to a new company that claims they can translate it into Java, but none have actually been able to deliver. my conspiracy theory is that any companies still left with mainframes and COBOL are far past the point of being able to transition from that, and these COBOL-away companies make all their money doing long expensive consultations before saying it won't work and finding another sucker. it's Lucy-pulling-the-football-out-from-under-Charlie-Brown-as-a-Service, or LptfofuCBaaS

Xandra Granade 🏳️‍⚧️

@aud @SnoopJ @glyph @ireneista Oh no, it was slightly worse than that even still. Any file in your project could have a compiler directive that changed the array base globally for the project. It wasn't even an option in your project, that might make some sense!

Asta [AMP]

@[email protected] @[email protected] @[email protected] @[email protected] okay! Having not just focused on the LaTeX part this time but the ring and field one… that is… I want that.

Am I correct in thinking that would mean, like, compiler created operator overloads/implementations for arbitrary types for which you could define what the result must be??