trealla-js VS the-power-of-prolog

I think Prolog really shines as an embedded query engine (I know this is old and it's been removed since). It's perfect for declarative configuration, very easy to write powerful queries once you wrap your head around it.

The Yarn constraints plugin also used (Tau) Prolog, although it looks like it's in the process of being replaced with JS, which makes me a bit sad. The reasoning is here: https://github.com/yarnpkg/berry/issues/1276. Seems like the biggest issue is lack of a nice dev environment. I maintain the Trealla Prolog Wasm port (npm package 'trealla') and I hope some day to use it for a VSCode extension or LSP or something to provide a nice dev experience. Performance has also been cited as an issue[1] but Trealla is quite fast and I expect it could easily handle a complex Yarn workspace with tons of facts. If this sounds like something you'd be interested in helping me with, feel free to contact me or make an issue/discussion here: https://github.com/guregu/trealla-js

[1]: https://github.com/yarnpkg/berry/issues/4079#issuecomment-10...

port a C or Rust project to wasm and make a JS package for it (trealla-js is an example of such a package, porting a C Prolog interpreter to wasm)

Hey, this is mine. Thanks for submitting it. I'll answer some questions.

> Why?

I ported Trealla Prolog to WASM (WASI) and I was looking for something useful to test it against. I found Spin, which can run WASM+CGI, and landed on this. Making this project exposed a number of bugs in my port that have now been fixed, so consumers of more useful projects[1][2] benefit as well. Also, PHP style templates are just fun! There's something valuable to just being able to shove a little bit of code inside some HTML and get it up on the internet.

I started my webdev journey with PHP many many years ago, and it's nice to revisit it from a different perspective. I don't use the real (elephant) PHP anymore, but I've gained a newfound appreciation for how fun its quick & dirty development style is.

I hope this project can serve as an example of how to use Prolog for fun things. It does showcase some of the cooler dynamic aspects of the language, and the PHP parsing code is like 10 lines of DCG.

> Is it a joke?

Yes and no. The name is certainly a joke. I was pondering what 'Prolog on Rails' might be and thought calling it PHP would be funny. This led to the PHP-style templates which were quick to implement and pretty powerful. Despite the humorous presentation, it does actually work.

> Can you use Prolog for web services?

Yes! For example, SWI has a mature HTTP package: https://www.swi-prolog.org/pldoc/doc_for?object=section(%27p.... It's used to power SWISH, an online Prolog code sharing thing: https://swish.swi-prolog.org/

> Next steps?

I would like to support persistence somehow. I think it'd be really cool if you could use Prolog's dynamic database[3] as a persistent store. Spin has components for Postgres and Redis so it shouldn't be too hard to implement, but I lose the WASI compatibility if I do that... which means I can't use the binary from WAPM, etc.

I would also like to experiment with running Trealla on Cloudflare Workers. I have another project, worker-prolog[4], which uses Tau Prolog (a Prolog written in Javascript) on Workers.

On a somewhat related note, I've also been playing around with Cosmopolitan libc[5]. I got Trealla to compile to an APE executable but there's some issues with the embedded Prolog libraries getting garbled, so I need to improve my GDB skills and figure out what's going on there.

Finally, I'd like to say thanks to Andrew Davison (@infradig on GitHub), the author of Trealla Prolog, for letting me add WASM support to his project and helping me with lots of things. For example, PHP led to Andrew implementing improvements for using DCGs to parse Prolog terms, which is now super fast[6]!

[1]: https://github.com/guregu/trealla-js

[2]: https://github.com/trealla-prolog/go

[3]: https://www.swi-prolog.org/pldoc/man?predicate=assertz/1

[4]: https://github.com/guregu/worker-prolog

[5]: https://justine.lol/cosmopolitan/

[6]: https://github.com/trealla-prolog/trealla/issues/53

With trealla-prolog/go on the backend and trealla-js on the frontend, you can share the same validation code.

I found async generators to be handy while porting a C Prolog interpreter to WASM/JS. A `for await` loop is a natural way to iterate over query results, and as a bonus you can use `finally` within the iterator to automatically clean up resources (free memory). There are ways for determined users to leak (manually iterating and forgetting to call `.return()`) but I've found that setting a finalizer on a local variable inside of the generator seems to work :-). I can't say manual memory management is a common task in JS but it did the trick.

The generator in question, which is perhaps the gnarliest JS I've ever written: https://github.com/guregu/trealla-js/blob/887d200b8eecfca8ed...

Prolog is excellent for bikeshedding, in fact that might be its strongest axis. It starts with everything you get in a normal language such as naming things, indentation, functional purity vs side effects, where to break code into different files and builds on that with having your names try to make sense in declarative, relational, logical and imperative contexts, having your predicates (functions) usable in all modes - and then performant in all modes - having your code be deterministic, and then deterministic in all modes. Being 50 years old there are five decades of learning "idiomatic Prolog" ideas to choose from, and five decades of footguns pointing at your two feet; it has tabling, label(l)ing, SLD and SLG resolution to choose from. Built in constraint solvers are excellent at tempting you into thinking your problem will be well solved by the constraint solvers (it won't be, you idiot, why did you think that was a constraint problem?), two different kinds of arithmetic - one which works but is bad and one which mostly works on integers but clashes with the Prolog solver - and enough metaprogramming that you can build castles in the sky which are very hard to debug instead of real castles. But wait, there's more! Declarative context grammars let you add the fun of left-recursive parsing problems to all your tasks, while attributed variables allow the Prolog engine to break your code behind the scenes in new and interesting ways, plenty of special syntax not to be sneezed at (-->; [_|[]] {}\[]>>() \X^+() =.. #<==> atchoo (bless you)), a delightful deep-rooted schism between text as linked lists of character codes or text as linked lists of character atoms, and always the ISO-Standard-Sword of Damocles hanging over your head as you look at the vast array of slightly-incompatible implementations with no widely accepted CPython-like-dominant-default.

Somewhere hiding in there is a language with enough flexibility and metaprogramming to let your meat brain stretch as far as you want, enough cyborg attachments to augment you beyond plain human, enough spells and rituals to conjour tentacled seamonsters with excellent logic ability from the cold Atlantic deeps to intimidate your problem into submission.

Which you, dear programmer, can learn to wield up to the advanced level of a toddler in a machine shop in a mere couple of handfuls of long years! Expertise may take a few lifetimes longer - in the meantime have you noticed your code isn't pure, doesn't work in all modes, isn't performant in several modes, isn't using the preferred idiom style, is non-deterministic, can't be used to generate as well as test, falls into a left-recursive endless search after the first result, isn't compatible with other Prolog Systems, and your predicates are poorly named and you use the builtin database which is temptingly convenient but absolutely verboten? Plenty for you to be getting on with, back to the drawing boar...bikeshed with you.

And, cut! No, don't cut; OK, green cuts but not red cuts and I hope you aren't colourblind. Next up, coroutines, freeze, PEngines, and the second 90%.

Visit https://www.metalevel.at/prolog and marvel as a master deftly disecting problems, in the same way you marvel at Peter Norvig's Pytudes https://github.com/norvig/pytudes , and sob as the wonders turn to clay in your ordinary hands. Luckily it has a squeaky little brute force searcher, dutifully headbutting every wall as it explores all the corners of your problem on its eventual way to an answer, which you can always rely on. And with that it's almost like any other high level mostly-interpreted dynamic programming / scripting language.

There is also AI that isn't machine learning. One could use formal logic to state rules and facts about the world and infer things from that. This sounds attractive but the main issue is that you need to build and maintain all of this knowledge. Most oldschool AI falls into this category. There's also fun programming languages like Prolog that are deep into this school: https://www.metalevel.at/prolog

There's a nice book[1][2] about Prolog, with modern characteristics. Moreover, there are things like ProbLog[3] and DeepProbLog[4] that allow you to use probabilistic reasoning and power of machine learning. I am personally looking forward for Scryer Prolog[5] to achieve its goals.

[1] https://www.metalevel.at/prolog

[2] https://github.com/triska/the-power-of-prolog

[3] https://github.com/ML-KULeuven/problog

[4] https://github.com/ML-KULeuven/deepproblog

[5] https://github.com/mthom/scryer-prolog

I had the same troubles until I encountered Markus Triska's modern perspective on revitalizing Prolog: https://www.metalevel.at/prolog.

The Power of Prolog [0] is a fantastic blog/video series covering everything from basic syntax, theoretical basis, modern features and idiomatic constructs.

I highly recommend it if you want to get the gist of Prolog and its modern features.

If you want a tour of Prolog, you can watch the video with that name [1].

[0]: https://www.metalevel.at/prolog

[1]: https://youtu.be/8XUutFBbUrg

I've gone through The Art of Prolog, most of The Power of Prolog, and a good chunk of the P-99 problems, and I have to say I'm simultaneously fascinated by and sceptical of Prolog. For some problem domains, implicit search is a very desirable property, and I can definitely see Prolog shining in that case. There are also many desirable properties and possibilities that are often reiterated, but concrete examples of how they would work are often missing. It comes down to: how does "production Prolog" look? A talk on Strange Loop by Michael Hendricks on exactly that topic was really helpful (especially w.r.t. some useful tools and libraries: func and yall are really great, and I still need to check mavis), but it still leaves me wondering on a couple of things.

Pretty much every Prolog book is quite good, but if you have the money or a local library with a copy, I really like Programming in Prolog by Clocksin, or Art of Prolog by Stering and Shapiro. If you want to follow a web resource, the standard suggestion is Markus Triska's The Power of Prolog.