rebanevapustus

> Brazil One of the most violently unequal countries on the planet. Very proficient however in virtue signalling.

Not true.

There has to be some change in the code, and they will not share the same semantics (and perhaps won't work when retractions/deletions also appear whilst streaming). And let's not even get to the leaky abstractions for good performance (watermarks et al).

They did, and their product is great.

It is the only database/query engine that allows you to use the same SQL for both batch and streaming (with UDFs).

I have made an accessible version of a subset of Differential Dataflow (DBSP) in Python right here: https://github.com/brurucy/pydbsp

DBSP is so expressive that I have implemented a fully incremental dynamic datalog engine as a DBSP program.

Think of SQL/Datalog where the query can change in runtime, and the changes themselves (program diffs) are incrementally computed: https://github.com/brurucy/pydbsp/blob/master/notebooks/data...

That's Feldera for you. It has a full blown SQL Api.

Materialize is cool as well, but is sorta hostile to self-hosting.

What does Arroyo have that Materialize or Feldera do not?

It seems that Arroyo is a strict subset of those.

There is no independent component of a country. It is very naive to claim so.

https://www.cnnbrasil.com.br/politica/por-decisao-de-moraes-...

What kind of non-authoritarian country arrests people for merely cursing at politicians on twitter?

Moreover, what kind of non-authoritarian country issues hundreds of thousands of rulings by its Supreme Court?

The Brazilian Supreme Court is an unelected entity that has complete control over the country, and firmly issues unappealable censorship arrests.

There is absolutely nothing this tyrannical in almost any western democracy, sans the UK.

It's cool that "we" have a payment system, however, I would never be comfortable using something whose people in charge are those that keep us (not me particularly because I've gladly left the country almost a decade ago) in this misery.

I use Crypto for everything you've mentioned. It's instant, almost free, and alexandre(he deserves a lowercased a) can't take my money if he feels that writing his name in lowercase makes me unworthy of my civil rights.

The Brazilian government is a *very* corrupt authoritarian oligarchy. I would take any US company over that any day.

It's naive to believe that the Brazilian government has the population's best interest at heart.

It is a Russian-style aristocracy with zero freedom of speech. I genuinely believe that Lula as a "people's person" cared for the "people" at some point, but that's long gone.

Brazil is a failed state with 52(!) GNI. That level of inequality signals a gargantuan failure of the state.

My hope for the country is that is gets split at some point, since as it is, it only "works" for a very small minority.

> The current descent into a quasi-fascist state isn't enticing either. > In the meantime, I guess I'll learn some basic Mandarin and spend more time in China.

What?

As somebody who is Brazilian, I gotta tell you.

Comparing the US with Brazil, especially with respect to some violence-adjacent statistic, is absurd in a way I don't think anybody from anywhere aside from perhaps South Africa could grasp big is your privilege.

This is 100% it.

D-Wave is great at solving issues that directly map to what it is physically solving i.e approximating the state of lowest energy of some entangled (sparse) lattice with fixed topology.

The point is, what if certain problems can be mapped to it such that business value could come out of it?

So far this hasn't been a thing, but at least it can do something non-trivial. There's no other quantum computing device that is as close to attaining real-world usefulness than D-Wave.

They did refuse to take it down.

You have no idea how many emails were exchanged, attempts at explaining them what a commit log is, what verified commits are...to just refuse to do anything until Stanford weighed in.

It is incredibly condescending to say "arXiv is not a place for a copyright strike". as if I decided to start one.

If somebody gets robbed in front of a place of worship/whatever, would you scold the victim by saying that's not the place to get robbed?

Incredible.

I have zero compassion for ArXiv.

Allow me to share a horror story:

I was the victim of a pretty bizarre super in-your-face academic theft. Someone snooped a half-finished paper draft of mine off GitHub and...actually got it published in ArXiv and a "real" journal: https://forbetterscience.com/2023/10/30/stephensons-alternat...

In spite of having a full commit log (with GitHub verified commits!!!) of both the code AND the paper, both ArXiv and the journal didn't seem to care or bother at all.

I went all the way to contacting Stanford, the institution that the thief falsely pretended to be affiliated with, to get them to help me with this.

Stanford contacted ArXiv, and ArXiv then: 1. Removed the thief's upload: https://arxiv.org/abs/2307.14810 2. Allowed the thief to copyright strike MY (!!!!) own research: https://arxiv.org/abs/2308.04214

How does this make any sense? You remove somebody's stolen content, and then allow the thief to copyright strike it? what the fuck...

The eytzinger layout cannot be used for anything dynamic. B-Trees are dynamic structures.

Thanks for clarifying it!

I checked some of your videos, and I couldn't quite grasp how much of (capital) DDflow do you borrow.

What does "our model does support cycles and iteration" mean? Do these also have the incremental semantics that guarantee bounds over the size of updates?

Electric Clojure v3 cannot be "Differential Dataflow" because it *seems* to be restricted to DAGs only. Differential Dataflow is highly expressive, it supports iteration.

In this implementation streams are not vectors nor lists. They are ordered hash maps.

Incremental Dataflow for the Masses sounds much better :)

https://github.com/brurucy/pydbsp/issues/6 it's been answered! (to the best of my knowledge)

Not so sure. I did not find the eurorack module they reference there, even on the analogue computer's manufacturer website.

That's why embedded Datalog is so good. You can delegate all that to the host language.

Datalog does not use backtracking, and is getting ever increasingly more popular.

See: - The fastest non-incremental embedded Datalog engine https://github.com/s-arash/ascent - The state-of-the-art non-embedded and non-incremental Datalog engine https://github.com/knowsys/nemo - A python library that contains an embedded incremental Datalog engine https://github.com/brurucy/pydbsp - A Rust library that provides a embedded incremental Datalog engine over property graphs https://github.com/brurucy/materialized-view

That's how it's done in the rust stdlib alternative https://github.com/brurucy/indexset

Faster reads, slower inserts, but then you get the capability of indexing by position in (almost) O(1). In regular B-Trees this can only happen in O(n).

It does not. The example I give on the README is only meant to show how easy it is to use it to "streamify" regular relational Dataframe operations.

He managed to DMCA strike the corrected version of the paper: https://arxiv.org/abs/2308.04214

We then contacted Stanford, and somehow got their legal team to message Arxiv telling them that Matthew is a fraudster.

In spite (!!) of Arxiv removing his stolen version thanks to Stanford's email: https://arxiv.org/abs/2307.14810

They refused to lift the obviously spurious DMCA claim on mine.

After almost a year, with an unsurmountable amount of open-source evidence and the thief having had every single paper he has ever written retracted for fraud!! the best the journal did is to add a notice: https://www.mdpi.com/2674-113X/2/3/20

Arxiv cared even less. They allowed the thief to DMCA strike me multiple times. He even managed to take down the real version of the paper by claiming that it was his: https://arxiv.org/abs/2308.04214

> Did you end up publishing your work eventually

No. When I tried to do so, I was actually rejected from a conference because their plagiarism detecting system labelled that I was trying to publish something that already been published (what was stolen).

It was very traumatic.

Big fan of Feldera here.

I would advise everybody to stay clear of anything that isn't Feldera or Materialize. Nobody aside from these guys have a IVM product that is grounded on proper theory.

If you are interested in trying out the theory (DBSP) underneath Feldera, but in Python, then check this out: https://github.com/brurucy/pydbsp

It works with pandas, polars...anything.

I was the victim of a pretty bizarre super in-your-face academic theft.

Someone snooped a half-finished draft of mine off GitHub and...actually got it published in a real journal: https://forbetterscience.com/2024/05/29/who-are-you-matthew-...

In spite of having a full commit log (with GitHub verified commits!!!) of both the code AND the paper, both arxiv and the journal didn't seem to care or bother at all.

Anyhow, I highly recommend reading the for better science blog. It's incredible how rampant fraud truly is. This applies to multiple nobel prize winners as well. It's nuts.

I was the victim of a pretty bizarre super in-your-face academic theft. Someone snooped a half-finished draft of mine off GitHub and...actually got it published in a real journal.

https://forbetterscience.com/2023/10/30/stephensons-alternat...

My institution completely brushed it off as a no problem, and said that it wasn't worth pursuing it at all.

It was very traumatising (and still is). I lost all faith in academia.

There is no such thing as "a more academic form of Datalog". There is only one Datalog.

Both Datascript and Datomic support recursion. In SICP it too talks about how rules can be defined recursively.

If there is no recursion, then it's just SQL without negation.

What makes it Datalog is the recursive part.

Plus, the "triples" part of what you call "Datalog", as it is explained on Datomic's documentation, is nothing related to Datalog at all. It's just a convenient format to ensure that all data is in a RDF-inspired format. Semantics aside, it makes it easier to build indexes and for it to be performant, compared to multi-relation unbounded-arity Datalog.