[qsort]

As much as SQL could be better, it's a very strong local optimum. There is no credible replacement whatsoever for what it does.

It's true people try to replace it, but most solutions are of rather questionable quality. I personally argue against ORMs in pretty much any situation.

> SQL is basically a DSL for data processing

Yeah and what's wrong with that? SQL databases power everything, including data sources with billions of records that serve millions of customers, with hundreds of developers/DBAs working on them. Does that not count as programming in the large?

[samatman]

> There is no credible replacement whatsoever for what it does.

I'm compelled to agree with this under protest. There was a language called Dataphor based on D and the Third Manifesto, it's unfortunately hard to find even sample code any longer, but it would be a strict improvement on SQL.

Did I mention you can't even find a corpus? Good luck running any implementation on a modern system.

I'm reasonably content writing SQL, but I know a syntax with the same power but lacking several disadvantages is possible, and I'd rather use a mature implementation of that instead, if I were able.

[qsort]

The worst part of SQL is indeed how hard it is to compose. Something like D, where you would keep the relational model but add user-defined types would indeed be a strict improvement.

The reason why I'm skeptical of most proposed alternatives is that I'm fundamentally in your exact same position: I know SQL isn't perfect but I'm reasonably content with it; and they invariably all end up throwing away the "good parts" of SQL (relational model, declarative, easy stuff is easy).

The first step of an hypothetical solution that replaces SQL isn't "SQL sucks", it's "SQL is extremely good at what it does but has problems that are only fixable with a new language".

[cmrdporcupine]

The Dataphor stuff was all open sourced I believe. But it's in a C#/MS ecosystem, and tied into their overall application builder tool etc. Not really a standlone product.

There's other "D" type projects. Rel is one. Don't have URL handy but it's fairly easy to find. They publish their grammar and example code.

None of these tools have ever matured or become popular. I have theories why, we could discuss forever.

I think the issue has more to do with two things: a) most people don't understand the relational model fully, so they have no idea what they're missing b) new databases (and existing) simply cannot afford to rock the boat here because they need customers. And in terms of the architecture of the DB system, the SQL parser is one of the lower effort items (when compared to query planner, optimization, storage impl and storage optimization, replication, etc.) So why invest there for little value? Customers aren't asking for it.

The company I'm contracting for right now has some interest in working in this space.

Date & Darwin did good work with "The Third Manifesto" but it went almost completely ignored, and the tone and target of it may have been off. Frustratingly we went through a phase where SQL went out of style and then back into style ("NewSQL") and so there may have been a window missed there where alternative query languages (but still based on the relational model) could have risen. But instead "NoSQL" was too interested in jettisoning the relational model along with SQL (mostly I would argue because they don't understand it).

There has been some recent rise in Datalog implementations in the Clojure community. That is interesting, though not strictly as an alternative to SQL.

[tkot]

>As much as SQL could be better, it's a very strong local optimum

I mostly agree, though I think that SQL's popularity comes mostly from network effects.

> Yeah and what's wrong with that?

Nothing is wrong wrong with that. I just think that it's reasonable to interpret the words "programming in the large" as "being usable as general purpose programming language" (which I admit is not a very strict definition) whereas SQL (or XSLT or awk or bash...) seem well suited for certain niches (important niches nonetheless).

[carapace]

Prolog is a better RM language than SQL. (Codd had a language called Alpha but that's about all I know about it.) It's not a credible replacement for all use cases, but it's pretty capable.