Timeouts on calls are, as the OP mentions, a thing in Erlang. Inter-process and inter-computer calls in QNX can optionally time out, and this includes all system calls that can block. Real-time programs use such features. Probably don't want it on more than that. It's like having exceptions raised in things you thought worked.
- Capabilities
They've been tried at the hardware level, and IBM used them in the System/38, but they never caught on. They're not really compatible with C's flat memory model, which is partly they fell out of fashion. Capabilities mean having multiple types of memory. Might come back if partially-shared multiprocessors make a comeback.
- Production-Level Releases
That's kind of vague. Semantic versioning is a related concept. It's more of a tooling thing than a language thing.
- Semi-Dynamic Language
I once proposed this for Python. The idea was that, at some point, the program made a call that told the system "Done initializing". After that point, you couldn't load more code, and some other things that inhibit optimization would be prohibited. At that point, the JIT compiler runs, once. No need for the horrors inside PyPy which deal with cleanup when someone patches one module from another.
Guido didn't like it.
- Value Database
The OP has a good criticism of why this is a bad idea. It's an old idea, mostly from LISP land, where early systems saved the whole LISP environment state. Source control? What's that?
- A Truly Relational Language
Well, in Python, almost everything is a key/value store. The NoSQL people were going in that direction. Then people remembered that you want atomic transactions to keep the database from turning to junk, and mostly backed off from NoSQL where the data matters long-term.
- A Language To Encourage Modular Monoliths
Hm. Needs further development. Yes, we still have trouble putting parts together. There's been real progress. Nobody has to keep rewriting Vol. I of Knuth algorithms in each new project any more. But what's being proposed here?
- Modular Linting
That's mostly a hack for when the original language design was botched. View this from the point of the maintenance programmer - what guarantees apply to this code? What's been prevented from happening? Rust has one linter, and you can add directives in the code which allow exceptions. This allows future maintenance programmers to see what is being allowed.
> But what's being proposed here?
I read it, and to me it seems like they're worried about the wrong things. As I understand it, they're worried about the difficulty and hassle of calling unrelated code in the monolith, and proposing things that would make it easier. But that's wrongheaded. Monoliths don't suffer because it's too hard to reuse functionality. They suffer because it's too easy. Programmers create connections and dependencies that shouldn't exist, and the monolith starts to strangle itself (if you have good tests) or shake itself to pieces (if you don't) because of unnecessary coupling. You need mechanisms that enforce modularity, that force programmers to reuse code at designated, designed module interfaces, not mechanisms that make it easier to call arbitrary code elsewhere in the monolith.
In my opinion, a great deal of the success of microservices is due to the physical impossibility of bypassing a network API and calling the code of another service directly. Because of this, programmers respect the importance of designing and evolving APIs in microservices. Essentially, microservices enforce modularity, forcing programmers to carefully design and evolve the API to their code, and this is such a powerful force for good design that it makes microservices appealing even when their architectural aspects aren't helpful.
A language that made it possible to enforce modularity in a monolith as effectively as it is enforced in microservices would make monoliths a no-brainer when you don't need the architectural aspects of microservices.
That's why they introduced a module system for Java in Java 9. It sounded pretty cool to me when they announced it, but it was a bit too late to make much difference in the Java ecosystem (Java 9 came out 21 years after Java 1) and I haven't heard much about it since then.
Besides, it makes sense to encapsulate at different scopes - a class, a package, a module, a library/application. These are "abstractions" on different levels, semantics, constraints.
C is not really dependent on a flat memory model - instead, it models memory allocations as separate "objects" (quite reniniscent of "object orientation in hardware" which is yet another name for capabilities), and a pointer to "object" A cannot be offset to point into some distinct "object" B.
> A Truly Relational Language
This is broadly speaking how PROLOG and other logic-programming languages work. The foundational operation in such languages is a knowledge-base query, and "relations" are the unifying concept as opposed to functions with predefined inputs and outputs.
(This is one of those times where the C memory model as described in the spec is very different from the mental PDP-11 that C programmers actually use to reason about)
> Well, in Python, almost everything is a key/value store.
Why would that be anywhere near an adequate substitute? KV stores are not relational, they don't support relational algebra. KV stores in PLs are common as dirt, so if they were relevant to the question of ending relations in a language I think the author would have noticed.
Symbolics Genera can save (incremental and complete) images (-> "Worlds"). The image tracks all the sources loaded into it. The sources/files/docs/... of the software is stored on a central (or local) file server.
I can for example start an initial world and load it with all the wanted software in the various versions I want. Maybe I save a new world from that.
I can also start an pre-loaded world and incrementally update the software: write patches, create new minor/major versions, load patches and updates from the central server, install updates from distributions, ... Maybe save new worlds.
The "System Construction Tool" tracks what code is loaded in what version from where.
I found this description amusing because all modern memory safe languages have capabilities, and they all have multiple types of memory: that's what an object is! Memory safety partitions memory into different types, and object references are capabilities!
What languages do next is where they break the security properties of capabilities: they add "ambient authority" and "rights amplification". Quick primer:
Ambient authority is basically the same problem as globally mutable state. Globally mutable state impedes modular reasoning about code, but if that state also carries authority to do something dangerous in the real world, like launch missiles, then it also impedes modular reasoning about security for the exact same reasons.
Rights amplification is the ability to turn a reference to an object with little to no authority, into a reference to an object with more authority. File.Open is the quintessential example, where you can turn an immutable string that conveys no authority, into a file handle to your root file system!
File.Open is also typically accessible to all code, meaning it's also ambient authority. This classical file API is completely bonkers from a security perspective.
So we already have capabilities, what we really need to do is to stop adding all of this insecurity! The developers of the E language actually showed that this could be done by making a capability secure subset of Java called Joe-E, which removed the parts of Java and the standard library that exposed ambient authority or rights amplification patterns. Most Java code could run unmodified.
And as for whether capability security will ever be adopted elsewhere, it already has been! WASM's WASI has capability security in its core design, because capability security is exactly what you need for good isolation and virtualization, which are the core properties WASM needs.
- Everything except C now has standard strings, not just arrays of characters. Almost all languages now have some standard way to do key/value sets. What else ought to be standard?
-- Arrays of more than one dimension would be helpful for numerical work. Most languages descended from C lack this. They only have arrays of arrays. Even Rust lacks it. Proposals run into bikeshedding - some people want rectangular slices out of arrays, which means carrying stride info around.
-- Standard types for 2, 3 and 4-element vectors would help in graphics work. There are too many different implementations of those in most language and too much conversion.
Things to think about:
- Rust's ownership restrictions are harsh. Can we keep the safety and do more?
-- The back-reference problem needs to be solved somehow. Back references can be done with Rc and Weak, but it's clunky.
-- Can what Rust does with Rc, RefCell, and .borrow() be checked at compile time? That allows eliminating the run-time check, and provides assurance that the run-time check won't fail. Something has to look at the entire call tree at compile time, and sometimes it won't be possible to verify this at compile time. But most of the time, it should be.
-- There's a scheme for ownership where there's one owning reference and N using references. The idea is to verify at compile time that the using references cannot outlive the owning one. Then there's no need for reference counds.
-- Can this be extended to the multi-thread case? There have been academic demos of static deadlock detection, but that doesn't seem to have made it into production languages.
-- A common idiom involves things being owned by handles, but also indexed for lookup by various keys. Dropping the handle drops the object and removes it from the indices. Is that a useful general purpose operation? It's one that gets botched rather often.
-- Should compilers have SAT-solver level proof systems built in?
-- Do programs really have to be in monospaced fonts? (Mesa on the Alto used the Bravo word processor as its text editor. Nobody does that any more.)
-- There's async, there are threads, and there are "green threads", such as Go's "goroutines". Where's that going?
-- Can we have programs which run partly in a CPU and partly in a GPU, compiled together with the appropriate consistency checks, so the data structures and calls must match to compile?
-- How about "big objects?" These are separately built program components which have internal state and some protection from their callers. Microsoft OLE did that, some .dll files do that, and Intel used to have rings of protection and call gates to help with that, hardware features nobody used. But languages never directly supported such objects.
So there are a few simple ideas to think about.
Zig's strings are null-terminated arrays of bytes. It's an unfortunate mistake from an otherwise interesting language.
Source: https://stackoverflow.com/questions/36647286/are-c-multidime...
Although at least with 2d arrays I prefer to just use a 1d array and index it with [x * width + y], because one problem with multidimensional arrays in C is they need multiple allocations/frees.
edit: Isn't it just:
float (*arr)[m][n] = malloc(sizeof(*arr));But it seems to work, which I didn't expect[0]. It also seems like you have to de-reference it to use it which is weird.
In your code link you over allocate memory, sizeof *arr is enough and you need to dereference like with (*arr)[i][j]. You need to dereference it because it is a pointer to an array, if you dereference you get an array. You can also let the first dimensions decay then it looks like:
double (*arr)[m] = malloc(n * sizeof *arr);
arr[i][j] = ...
I think that character strings should not be restricted to (or generally expected to be) Unicode (although using Unicode and other character sets will still be possible).
I also think that key/value lists should allow any or most types as keys, including references to objects. (PostScript allows any type to be used as keys except strings (they are converted to names if you use them as keys) and nulls.)
I think that big integers (which can have a program-specified limited length in programming languages with typed variables) and arrays of fixed-length records (which C already has; JavaScript has typed arrays which is a more limited implementation of this) are also things that would be helpful to include as standard.
> Arrays of more than one dimension would be helpful for numerical work.
I agree with this too; it is a good idea.
> Standard types for 2, 3 and 4-element vectors would help in graphics work.
This is probably helpful, too, although they can be used for stuff other than graphics work as well.
> Do programs really have to be in monospaced fonts?
No, but that is due to how it is displayed and is not normally a feature of the program itself. Many people including myself do use monospace fonts, but this should not usually be required.
> There's async, there are threads, and there are "green threads", such as Go's "goroutines". Where's that going?
I had read about "green threads" and I think that it is a good idea.
> How about "big objects?" These are separately built program components which have internal state and some protection from their callers. Microsoft OLE did that, some .dll files do that, and Intel used to have rings of protection and call gates to help with that, hardware features nobody used. But languages never directly supported such objects.
I also think it is sensible to have components that can be separated from the callers, and that operating system support (and perhaps hardware support) for such thing might be helpful. I would design a programming language for such an operating system that would directly support such objects.
https://www.languagesforsyste.ms/publication/fearless-concur...
> -- Should compilers have SAT-solver level proof systems built in?
They already do. The exhaustivity checker in Rust (and functional languages) is equivalent in power to SAT.
Of course not. I've been using a proportional font for at least 10 years and I'm still in business working on code bases shared with developers using monospaced fonts. Both work, none disturb the other, proportional is easier to read as any book can demonstrate. Alignment doesn't matter much.
About capabilities, I think that capabilities should be a feature of the operating system, although hardware support would be helpful. However, I think that it could be done with tagged memory, without necessarily needing multiple types of memory, and programming languages such as C could still be capable of using them (although some things might not work as it would be expected on other computers, e.g. if you try to copy a reference to a capability into a memory area that is expected to be a number and then try to perform arithmetic on that number, the program is likely to crash even if the result is never dereferenced).
However, my idea also involves "proxy capabilities" too, so that you can effectively make up your own capabilities and other programs receive them without necessarily knowing where they came from (this allows supporting many things, including (but not limited to) many of the idea of "divergent desktop" of Arcan).
If anything many "modern" low code SaaS products are much worse in this regard, than what Lisp and Smalltalk have been offering for years.
They simply assert "twiddling a run-time variable for debugging in your staging environment can propagate straight into a bug on production".
As-if straight into production without re-testing.
> Source control? What's that?
"ENVY/Manager augments this model by providing configuration management and version control facilities. All code is stored in a central database rather than in files associated with a particular image. Developers are continuously connected to this database; therefore changes are immediately visible to all developers."
https://www.google.com/books/edition/Mastering_ENVY_Develope...
~
1992 Product Review: Object Technology’s ENVY Developer
http://archive.esug.org/HistoricalDocuments/TheSmalltalkRepo...
He proposes that there is a need for a way to connect modules, i.e. dependency injection, without the modules having explicit knowledge of each other, with compile-time verification that the modules being connected are compatible, without the interface song and dance.
I think squeak had Monticello for source control with their image based approach almost 20+ years ago and there was something else for smalltalk in the '80s too.
But yeah people like text and hate images, and I believe Pharo switched back to some git integration.
"ENVY/Manager augments this model by providing configuration management and version control facilities. All code is stored in a central database rather than in files associated with a particular image. Developers are continuously connected to this database; therefore changes are immediately visible to all developers."
https://www.google.com/books/edition/Mastering_ENVY_Develope...
~
1992 Product Review: Object Technology’s ENVY Developer
http://archive.esug.org/HistoricalDocuments/TheSmalltalkRepo...
https://cuis-smalltalk.github.io/TheCuisBook/The-Change-Set....
I am also agreeing that relational approach to in-memory data is a good, efffective thought.
I recently compiled some of my C code with the sqlite database and I'm preparing to think how the SQL model of my standard code could be used as the actual implementation language of in memory operations.
Instead of writing the hundredth loop through objects I just write a SQL query instead with joining with seeing the internal data representation of the software as an information system instead of bespoke code.
I was hoping to make it possible to handle batches of data and add parallelism because arrays are useful when you want to parallelise.
I was thinking, wouldn't it be good if you could write your SQL queries in advance of the software and then parse them and then compile them to C code (using an unrolled loop of the SQLite VM) so they're performant. (For example, instead of a btree for a regular system operation, you can just use a materialised array a bit like a filesystem so you're not rejoining the same data all the time)
I was thinking of ways of representing actors somehow communicating by tables but I do not have anything concrete for that.
DataDraw is an ultra-fast persistent database for high performance programs written in C. It's so fast that many programs keep all their data in a DataDraw database, even while being manipulated in inner loops of compute intensive applications. Unlike slow SQL databases, DataDraw databases are compiled, and directly link into your C programs. DataDraw databases are resident in memory, making data manipulation even faster than if they were stored in native C data structures (really). Further, they can automatically support infinite undo/redo, greatly simplifying many applications.
> I just write a SQL query instead with joining with seeing the internal data representation of the software as an information system instead of bespoke code
This sounds very similar to how CINQ's macro-based implementation performs relational optimizations on top of regular looking Clojure code (whilst sticking to using a single language for everything).
I really like what PRQL [1] did, at least it makes table operations easily chainable. Another one that comes to mind is Datomic [2].
[2]: https://docs.datomic.com/peer-tutorial/query-the-data.html
data Person = User ?except .password ?select .name, .id + Customer ?select .name, .customer_id AS .id
(Funny after years of searching I found somebody who taught me how to do really complex modelling in OWL DL but from reading the literature I'm pretty sure the average PhD or prof in the field has no idea.)
Is there any resource you could recommend for that?
[0] Recent discussion https://news.ycombinator.com/item?id=42577736
"RM: What was key to SQL becoming the standard language for relational databases in the mid- 1980s? Was all down to good marketing?
CJD: In other words, why did SQL became so popular? Especially given all its faults? Well, I think this is rather a sorry story. I said earlier that there has never been a mainstream DBMS product that’s truly relational. So the obvious question is: Why not? And I think a good way for me to answer your questions here is to have a go at answering this latter question in their place, which I’ll do by means of a kind of Q&A dialog. Like this:
Q:
Why has no truly relational DBMS has ever been widely available in the marketplace?
A:
Because SQL gained a stranglehold very early on, and SQL isn’t relational.
Q:
Why does SQL have such a stranglehold?
A:
Because SQL is “the standard language for RDBMSs.”
Q:
Why did the standard endorse SQL as such and not something else-something better?
A:
Because IBM endorsed SQL originally, when it decided to build what became DB2. IBM used to be more of a force in the marketplace than it is today. One effect of that state of affairs was that-in what might be seen as a self-fulfilling prophecy-competitors (most especially Relational Software Inc., which later became Oracle Corp.) simply assumed that SQL was going to become a big deal in the marketplace, and so they jumped on the SQL bandwagon very early on, with the consequence that SQL became a kind of de facto standard anyway.
Q:
Why did DB2 support SQL?
A:
Because (a) IBM Research had running code for an SQL prototype called System R and (b) the people in IBM management who made the decision to use System R as a basis on which to build DB2 didn’t understand that there’s all the difference in the world between a running prototype and an industrial strength product. They also, in my opinion, didn’t understand software (they certainly didn’t understand programming languages). They thought they had a bird in the hand.
Q:
Why did the System R prototype support SQL?
A:
My memory might be deficient here, but it’s my recollection that the System R implementers were interested primarily in showing that a relational-or “relational”-DBMS could achieve reasonable performance (recall that “relational will never perform” was a widely held mantra at the time). They weren’t so interested in the form or quality of the user interface. In fact, some of them, at least, freely admitted that they weren’t language designers as such. I’m pretty sure they weren’t all totally committed to SQL specifically. (On the other hand, it’s true that at least one of the original SQL language designers was a key player in the System R team.)
Q:
Why didn’t “the true relational fan club” in IBM-Ted and yourself in particular-make more fuss about SQL’s deficiencies at the time, when the DB2 decision was made?
A:
We did make some fuss but not enough. The fact is, we were so relieved that IBM had finally agreed to build a relational-or would-be relational-product that we didn’t want to rock the boat too much. At the same time, I have to say too that we didn’t realize how truly awful SQL was or would turn out to be (note that it’s much worse now than it was then, though it was pretty bad right from the outset). But I’m afraid I have to agree, somewhat, with the criticism that’s implicit in the question; that is, I think I have to admit that the present mess is partly my fault."
Postgres was "truly relational" for a significant portion of its life before finally losing the battle with the SQL virus. There is probably no DMBS more widely available. Granted, it wasn't widely used until the SQL transition.
> SQL isn’t relational.
This is key. Relations are too complicated for the layman, the one who is paying for it, to understand. Tables are more in tune to what is familiar to them. The hardcore math/software nerds might prefer relationality, but they aren't the ones negotiating multi-million dollar contracts with Oracle/IBM.
I remember when Postgres moved to SQL. People started billing it as being Oracle, but free. That got non-technical manager attention. Without that marking success appealing to the layman, I expect nobody would be using it today.
It was a breakthrough system is almost all aspects. It defined what dbs would look like both internally and externally for decades.
In many ways SQL is like C: good for what authors wanted it to be but severe consequences much later.
But the history doesn't care. SQL (and C) still have many-many years ahead of them.
Some of the other points resonate with me. I think sensible dynamic scoping would be an easy way to do dependency injection. Together with something like linear types you could do capabilities pretty smoothly, I think. No real reason why you couldn’t experiment with some persistent storage as one of these dependencies, either. Together with a good JIT story would make for a good, modular environment.
AFAIK it doesn’t have closures (it’s too C-like) so you need to use methods for all your (implicit) interfaces, but that’s okay…
I think the “exemplars” could be automatically yoinked from documentation and tests and existing usage of the function in the code base. Work needs to be done on the IDE front to make this accessible to the user.
> Smalltalk and another esoteric programming environment I used for a while called Frontier had an idea of a persistent data store environment. Basically, you could set global.x = 1, shut your program down, and start it up again, and it would still be there.
Frontier! I played with that way back when on the Mac. Fun times.
But as for programming language with integrated database... MUMPS! Basically a whole language and environment (and, in the beginning, operating system) built around a built-in global database. Any variable name prefixed with ^ is global and persistent, with a sparse multi-dimensional array structure to be able to organize and access the variables (e.g. ^PEOPLE(45,"firstname") could be "Matthew" for the first name of person ID 45). Lives on today in a commercial implementation from Intersystems, and a couple Free Software implementations (Reference Standard M, GT.M, and the GT.M fork YottaDB). The seamless global storage is really nice, but the language itself is truly awful.
In fact, it does this during a "preinit" stage that runs immediately after compilation. Once all preinit code finishes executing, the compiled image is overwritten with the updated state. The language includes a "transient" keyword to permit creating objects that should not be stored.
This same mechanism permits in-memory snapshots, which are used for the game's UNDO feature. No need to rewind or memento-ize operations, just return to a previous state.
It's not a general-purpose mechanism. After all, the language is for building games with multiple player-chosen save files, and to permit restarting the game from a known Turn 0 state.
Save image... for short-term convenience; build clean every week from archived text files.
----
1984 "Smalltalk-80 The Interactive Programming Environment" page 500
"At the outset of a project involving two or more programmers: Do assign a member of the team to be the version manager. … The responsibilities of the version manager consist of collecting and cataloging code files submitted by all members of the team, periodically building a new system image incorporating all submitted code files, and releasing the image for use by the team. The version manager stores the current release and all code files for that release in a central place, allowing team members read access, and disallowing write access for anyone except the version manager."
https://rmod-files.lille.inria.fr/FreeBooks/TheInteractivePr...
The best Smalltalk these days is GlamorousToolkit: https://gtoolkit.com/
It has a sort of git in it, so you can easily "rollback" your image to previous states. So going back and forth in history is trivial.
Sounds very interesting. Does it support multi-developer merging and/or rebasing of changes?
That is, the VM is mapped to a backing file, changes persisted automatically, no "saving", limited by drive space (which, nowadays, is a lot). But nowadays we also have vast memory space to act as a page cache and working memory.
My contrived fantasy use case was having a simple array name "mail", which an array containing all of my email messages (in email object, of course). Naturally as you get more mail, the array gets longer. Also, as you delete mail, then the array shifts. It's no different, roughly, than the classic mbox format, save it's not just text, its objects.
You can see if you delete a email, from a large (several GBs), there would be a lot of churn. That implies maybe it's not a great idea to use that data structure, but that's not the point. You CAN use that data structure if you like (just like you can use mbox if you like).
Were it to be indexed, that would be done with parallel data structures (trees or hashes or whatever).
But this is all done automagically. Just tweaks to pages in working memory backed by the disk using the virtual memory manager. Lots and lot of potential swapping. C'est la vie, no different from anything else. This what happens when you map 4TB into a 16GB work space.
The problem with such a system, is how fragile is potentially is. Corrupt something and it happily persists that corruption, wrecking the system. You can't reboot to fix it.
Smalltalk suffers from that today. Corrupt the image (oops, did I delete the Object become: method again?), and its gone for good. This is mitigated by having backup images, and the changelist to try to bring you back to the brink but no further.
I'm guessing a way to do that in this system is to use a copy on write facility. Essentially, snapshot the persistent store on each boot (or whatever), and present a list of previous snapshot at start up.
Given the structure of a ST VM you'd like to think this is not that dreadful to work up. I'd like to think a paper napkin implementation PoC would be possible, just to see what it's like. One of those things were the performance isn't really that great, but the modern systems are so fast, we don't really notice it in human terms.
But I do think it would be interesting.
> oops, did I delete the Object become: method again?), and its gone for good.
And then you admit actually it's not gone for good because if you created the method that will be recorded in the changes.log file and if it was a provided method that will still be in the provided sources file.
https://cuis-smalltalk.github.io/TheCuisBook/The-Change-Log....
But as you said, the language itself is almost unbearable to use.
Functions are so ubiquitous we forget what they really are: a type of guarantee about the conditions under which the code within will run. Those guarantees include the availability of arguments and a place to put the return value (on the stack).
One of the key guarantees about sync functions is the call structure: one thread of execution will be in one function and one function only at any point during the program; the function will only be exited on return (or exception, or panic) or call of another function; and all the local data will be available only for the duration of that function call.
From that perspective, async functions are a _weakening_ of the procedural paradigm where it is possible to "leave behind" an instruction pointer and stack frame to be picked up again later. The ability to suspend execution isn't an additional feature, it's a missing guarantee: a generalisation.
There is always an interplay between expressiveness and guarantees in programming languages. Sometimes, it is worth removing a guarantee to create greater expressiveness. This is just an example of that.
I mentioned exceptions earlier — it's no wonder that exceptions and async both get naturally modelled in the same way (be it with monads or algebraic effects or whatever). They are both examples of weakening of procedural guarantees. Exceptions weaken the guarantee that control flow won't exit a function until it returns.
I think the practical ramifications of this are that languages that want async should be thinking about synchronous functions as a special case of suspendable functions — specifically the ones that don't suspend.
As a counterpoint, I can imagine a lot of implementation complexities. Hardware is geared towards the classical procedural paradigm, which provides an implementation foundation for synchronous procedures. The lack of that for async can partially explain why language authors often don't provide a single async runtime, but have this filled in by libraries (I'm thinking of Rust and Kotlin here).
The "sync guarantees" don't really exist either. If you have a(); b(); the compiler may very well reorder them to b(); a(); and give you similar issues as async. It may elide a() entirely (and reclaim the call structures), or the effects of a() might not be visible to b() yet. Synchronous functions also can and do suspend with all the associated issues of async. That comes up frequently in cryptography, kernel, and real time code.
A grammar is basically a class (or role/trait), and they can contain regexes (and regular methods). Those regexes have backtracking control (for simple tokens, you don't want to to try parse a string any other way than the first, obvious match).
This makes it much easier to write composable and understandable parsers.
I know that, technically, you could do that in a library, but somehow that's never the same; if it's not baked into the language, the hurdle to introduce another dependency is always looming, and then if there's more than one such library, the parsers aren't composable across libraries and so on.
I really wish more people would use Raku :-) Because it is all built in there already! :-)
I've added this feature to QuickJS [1] and it works quite well in Sciter as persistent Storage [2] mechanism, used in Sciter.Notes [3] for example.
let storage = Storage.open(filename);
let persistentData = storage.root;
if( !persistentData ) storage.root = persistentData = {... initial storage structure ...};
= Semi-Dynamic Language
De-facto we already use similar approach quite a while. In form of GPU shaders or WebAssembly. The solution is not in making script JIT friendly (that is against its nature) but with an option to use native/compiled/loadable modules written in languages that were designed to be compileable from the beginning.
My Sciter, as an embeddable engine, is an example of such environment. Native host application exposes native functions/classes to HTML/CSS/JS engine that implements UI layer of the application. UI is dynamic (fluid,styleable,etc.) by nature while application backend (a.k.a. business logic layer) is more static and linear by nature.
[1] https://gitlab.com/c-smile/quickjspp
You could write something similar to NSUserDefaults in any language for any computing environment, but that's not what the author is talking about. He is talking about the same basic concept, yes, but where it exists within in the language itself, not something implemented on top.
On other side persistence in QuickJS is more than that. Essentially it is a NoSQL DB integrated into the language and its runtime. For example you can write
let uname = root.users[2].firstName;
And also, in QuickJS, Storage is not reading whole DB in memory but fetches/unloads data on demand transparently for the user. You can think about content of DB as about genuine language data structure with root at storage.root
For “capabilities” and “A Language To Encourage Modular Monoliths”, I like the idea of a capability-secure module system. Something like ML’s signatures and functors, but modules can’t import, they only get access to the arguments passed into a functor. Everything is dependency injection. The build system determines which modules are compiled with which dependencies (which functors are passed which arguments).
An existing “semi-dynamic language” is CLOS, the Common Lisp object system. Its metaobject protocol is designed so that there are clear points when defining or altering parts of the object system (classes, methods, etc.) at which the result is compiled, so you know when you pay for being dynamic. It’s an interesting pre-Self design that doesn’t rely on JITs.
WRT “value database”, a friend of mine used to work for a company that had a Lisp-ish image-based geospatial language. They were trying to modernise its foundations by porting to the JVM. He had horror stories about their language’s golden image having primitives whose implementation didn’t correspond to the source, because of decades of mutate-in-place development.
The most common example of the “value database” or image-based style of development is in fact your bog standard SQL database: DDL and stored procedures are very much mutate-in-place development. We avoid the downsides by carefully managing migrations, and most people prefer not to put lots of cleverness into the database. The impedance mismatch between database development by mutate-in-place and non-database development by rebuild and restart is a horribly longstanding problem.
As for “a truly relational language”, at least part of what they want is R style data frames.
Most languages have a while construct and a do-while.
while(condition){block};
do{block}while(condition);
...
start:
condition
branch-if-false > end
block
branch-always > start
end:
...
...
start:
block
condition
branch-if-true > start
...
while(true){
prepare;
if(!check) break;
process
}
do{prepare}while(condition){process}
The best thing is that this construct can be optimized in assembly perfectly:
...
jump-always > start
after:
process
start:
prepare
condition
branch-if-true > after
...Example:
(loop repeat 8
;; init x, then decide how to "prepare" next iteration
for x = 0 then (+ (* 2 x) 1)
collect x)
(0 1 3 7 15 31 63 127)
(loop repeat 8
for x = 0 then (+ (* 2 x) 1)
;; but stop if x gets too big
while (< x 100)
collect x)
(0 1 3 7 15 31 63)
---
(prog1 foo bar*)
Useful if `foo` is the condition and you need to perform some change to it immediately after, eg:
(while (prog1 (< next prev) (setq prev next)) ...)
(prog2 foo bar baz*)
Might be what GP wants. `foo` is the preparation, `bar` is the condition`, and `baz` can be some post-condition mutation on the compared value. Not too dissimilar to
for (pre, cond, post) {}
(while (prog2 pre cond post) ...)
(progn foo*)
`progn` is similar to repeated uses of the comma operator in C, which GP has possibly overlooked as one solution.
while (prepare, condition) { process } for ([]{
/*prepare*/
}(); /*condition*/;) {
/*body*/
}
Let's rewrite the loop this way, with 'break' expanded to 'goto':
while (true) {
prepare...
if (!cond) goto exitpoint;
process...
}
exitpoint:
goto entrypoint;
do {
process...
entrypoint:
prepare...
} while(cond);
If we want to add the same kind of syntactic sugar for the jump-in case, the landing point must be explicit (no way for the compiler to guess it), so the only one we can make implicit is the starting point, that is where the 'do' is.
So we need: a new statement, let's call it 'entry', that is the dual of 'break' and a new semantic of 'do' to not start the loop at the opening '{' but at 'entry'.
do {
process...
entry;
prepare...
} while (cond);
loop
Get(Current_Character);
exit when Current_Character = '*';
Echo(Current_Character);
end loop;
do {
Get(Current_Character);
if (Current_Character == '*') break;
print(Current_Character);
} while (true);
loop
prepare;
while check;
process;
end;
do {
let value = prepare();
} while (value.is_valid) {
process(value);
}
Your are right about the word double braces, but I can't think of an alternate syntax other than just removing the braces around the while. But in that case it may seem odd to have a keyword that can only be used inside a specific block...wich is basically a macro for a if(.)break; Maybe I'm too used to the c/java syntax, maybe with a different way of defining blocks?
As long as it's documented and expected, it's not weird.
The scope then is the whole "do-while" statement, not the brace.
Hard to express without primitives to indicate that, maybe.
do {
let value = prepare();
while (value.is_valid) {
process(value);
} values := []string{"hello", "world"}
for v, value := range values {
fmt.Printf("%s", value);
}
inter {
fmt.Printf(",")
}
before {
fmt.Printf("[")
}
after {
fmt.Println("]")
}
empty {
fmt.Println("(nothing found)")
}
Though, you still need all that to be nested inside the loop (a closure, more generally.)
from <initialization statements> until <termination condition> loop <group of statements> end
repl_prompt="${repl_prompt-repl$ }"
while
printf "%s" "$repl_prompt"
read -r line
do
eval "$line"
done
echo
This should also be doable in languages where statements are expressions, like Ruby, Lisp, etc.
Here's a similar Ruby repl:
while (
print "repl> "
line = gets
) do
result = eval line
puts result.inspect
end
putsSidenote: I wasn't aware that shell allows for multiple lines, good to know!
Something like
do{
raw_data=produce.get()
data=parse(raw_data)
}while(data!=STOP){
process(data)
}
A do-while here also has the same issue, but in this case after you check the condition you can't do any processing afterwards (unless you move the check and process into a single check_and_process method...which you can totally do but again the idea is to not require it)
do {
offset = tell(filehandle)
} while(line = readline(filehandle))
{
print "line starts at offset: ", offset
}Look at C: it has 5 loop-related keywords (4.5; 'break' is two-timing in 'switch') yet it is still not enough.
https://docs.python.org/3/reference/compound_stmts.html
See sections 8.3, the for statement, and 8.2, the while statement.
Scala for example has a `breakable {}` block that lets you indicate where you should land after a `break`
breakable {
while condition {
// loop body
if otherCondition then break;
// rest of the body
}
// body of pythonic else block
} // you land here if you break
I find this so incredibly useful, that I miss it from other languages.
Something related that I've noticed with OO languages such as Java is that it tends to result in "ceremony" getting repeated n-times for processing n objects. a well-designed begin-process-end syntax for function calls over iterables would be amazing. This could apply to DB connection creation, security access checks, logging, etc...
while { prepare; condition }
do { process }
Scala has a lot of syntax goodies although some stuff is exotic. For example to have a 'break' you need to import it and indicate where from exactly you want to break out of.
We're working on a language with some of these ideas:
Object capabilities, async calls as easy as sync calls, modular monoliths, and (eventually) unified logging.
None of the relational language features though.
Feedback appreciated!
I am interested in the "No Hidden I/O" section. Can you store, say, the system value (?) to a list? Because I'm afraid this is not enough to properly track side-effects.
Unfortunately I think capabilities/an effect system is required for that (so that any point that actually uses a side effect, taints its whole call tree - the same way async/checked exceptions do, both of which are specific Effects.
The end result is something that looks a lot like distributed, persistent, transactional memory. Rather than explicit interactions with a database, local variable writes to your state are transactionally persisted if a method call succeeds, even across process/machine boundaries. And that benefits point 7, because transactional method calls compose across team/application boundaries.
[1] Loosen Up The Functions [3] Production-Level Releases [5] Value Database [7] A Language To Encourage Modular Monoliths
https://learn.microsoft.com/en-us/azure/azure-functions/dura...
Because multiple Azure entity functions are not updated transactionally, you are essentially always implementing the saga pattern: you have to worry about cleaning up after yourself in case of failure.
In Reboot, transactional function calls automatically roll back all state changes if they fail, without any extra boilerplate code. Our hypothesis is that that enables a large portion of an application to skip worrying about failure entirely.
Code that has side-effects impacting the outside world can be isolated using our workflow mechanism (effectively durable execution), which can themselves be encapsulated inside of libraries and composed. But we don't think that that is the default mode that developers should be operating in.
> Code that has side-effects impacting the outside world can be isolated using our workflow mechanism (effectively durable execution
Sounds very interesting!
I have been thinking about something like this for a new PL, and many kinds of side effect can actually be reversed, as if it never happened.
I have also read that exceptions can complicate control flow, disallowing some optimizations - but if they are transactional, then we can just add their reverse to the supposedly already slow error path, and enjoy our performance boost!
I've also always thought Joe Armstrong's (RIP) thought of "why do we need modules" is really interesting, too. There's a language I've seen posted on HN here a couple times that seems to go in that approach, with functions named by their normalized hash contents, and referred to anywhere by that, but I can't seem to remember what it's called right now. Something like "Universe" I think?
[0] https://github.com/Tablam/TablaM [1] https://www.scattered-thoughts.net [2] https://erlang.org/pipermail/erlang-questions/2011-May/05876...
Would you still have distinct activation records per call or forfeit the ability to have reentrant functions and recursion?
That's one of the main reasons to move to dynamic (as in a call stack) allocation of your activation records versus a single static allocation per function.
It would be awkward to work with since you'd have to be aware of all of the eventual caller scopes rather than your current local scope when defining it.
I suppose it would be like macros instead of true functions.
PIC16 has a hardware stack of only a few return addresses, and therefore imposes the "all register allocation is static" + "no recursion" that you're asking for.
Many such programming environments need strict control over stack sizes to avoid any possibility of stack overflow.
I had a similar notion a few years back, thinking about a somewhat wider range of "scoped guarantees". The compiler would compute things such as the maximum stack usage of a function, and this would "roll up" to call sites automatically. This could also be used to enforce non-usage of certain dangerous features such as locks, global flags, or whatever.
A lot of "staged" programs can be thought of as semi dynamic as well, even things like C++ template expansion or Zig comptime: run some logic up front, freeze it, then run the rest of the application later
To me , this idea seems so so insane (especially for things like extraction , like you start extracting a zip on one device and it can be partially extracted and then you can partially extract it on the other) (yes sure , you could loop over each file and have a list of files currently unzipped and rather unzip the file which hasn't been unziped yet)
But Imagine if the file to be extracted is a singular file in zip (like 100 gig file)
I don't know , I have played this with criu and it had worked. Qemu can also work. But this idea is cool
Instead of using a default storage where entropy can hit , I would personally like it if the values were actually stored in sqlite and combined with Truly Relational Language maybe as well (but it doesn't truly require you to learn sqlite)
I had posted this on one of hackernews this as well and theoretically its possible with the brainfu* in sqlite intepreter that I had found. But I don't know.... If anybody knows of a new language / a method for integrating this in new languages , it would be pretty nice.
Another cool way is using encore in golang or typescript and then hosting the aws stack yourself or running encore locally I am not sure)
There is also sst framework which can allow to be run on docker and also https://github.com/vercel/fun
Capabilities and dynamic scoping for "modularisation" nicely lead to implicit variables instead of truly global dynamically scoped variables. Implicit variables also probably work well to implement effect systems which means well behaved asyncs.
Edit: other features I want:
- easy embedding in other low level languages (c++ specifically)
- conversely, easy to embed functions written in another language (again c++).
- powerful, shell-like, process control system (including process trees and pipelines), including across machines.
- built-in cooperative shared memory concurrency, and preemptive shared nothing.
- built-in distributed content addressed store
I guess I want Erlang :)
The best I've managed is a functional language that allows for map, filter, and reduce, but forbids recursion or any other looping or infinite expansion in usercode.
The pitch is that this kind of language could be useful in contexts where you're executing arbitrary code provided by a potentially malicious third party.
One problem is that, while non-Turing-completeness can be helpful for maintainability, it's not really sufficient for security. Starlark programs can still consume exponential amounts of time and memory, so if you run an adversary's Starlark program without sandboxing it, you're just as vulnerable to denial-of-service attacks as you'd be with a Turing-complete language. The most common solution is sandboxing, wherein you terminate the program if it exceeds time or memory limits; however, once you have that, it's no longer necessary for the language to not be Turing-complete, so you might as well use a popular mainstream language that's easy to sandbox, like JavaScript.
One other intriguing option in the space is CEL (https://cel.dev), also designed at Google. This targets use cases like policy engines where programs are typically small, but need to be evaluated frequently in contexts where performance matters. CEL goes beyond non-Turing-completeness, and makes it possible to statically verify that a program's time and space complexity are within certain bounds. This, combined with the lack of I/O facilities, makes it safe to run an adversary's CEL program outside a sandbox.
If you want to limit the runtime, you need to apply a timeout.
That's because the language has a notion of `total` functions [2], and only `total` functions can be used for computing type signatures. These `total` functions must terminate in finite time and not crash. AFAIK, they aren't Turing complete, but they're still pretty expressive. `partial`, Turing complete functions are still allowed at runtime (outside of type signatures) [1].
If I understand correctly, running Idris in the type-checking mode (`--check`) should give you what you want.
[1]: https://cs.stackexchange.com/a/23916/159425
[2]: https://idris2.readthedocs.io/en/latest/tutorial/typesfuns.h...
"Semi-dynamic" is one of the most common architectures there is for large & complex systems. AAA games are usually written in a combination of C++ and a scripting language. GNU Emacs is a Lisp application with a custom interpreter that is optimized for writing a text editor. Python + C is a popular choice as well as Java + Groovy or Clojure, I've even worked with a Lua + FORTRAN system.
I also think "parsers suck". It should be a few hundred lines at most, including the POM file, to add an "unless" statement to the Java compiler. You need to (1) generate a grammar which references the base grammar and adds a single production, (2) create a class in the AST that represents the "unless" statement and (3) add an transformation that rewrites
unless(X) {...} -> if(!X) {...}
var statement = <<<SELECT * FROM page where id=:pageId>>>;
The counter to this you hear is that compile time is paramount and there's a great case for that in large code bases. (I had a system with a 40 minute build) Yet there's also a case that people do a lot of scripty programming and trading compile time for a ergonomics can be a win (see Perl and REBOL)
I think one goal in programming languages is to bury Lisp the way Marc Anthony buried Caesar. Metaprogramming would be a lot more mainstream if it was combined with Chomksy-based grammars, supported static typing, worked with your IDE and all that. Graham's On Lisp is a brilliant book (read it!) that left me disappointed in the end because he avoids anything involving deep tree transformations or compiler theory: people do much more advanced transformations to Java bytecodes. It might be easier to write those kind of transformations if you had an AST comprised of Java objects instead of the anarchy of nameless tuples.+
This is (essentially) Crystal lang's type system. You end up with semantic analysis/compilation taking a significant amount of time, longer than other comparable languages, and using a lot of resources to do so.
I can’t tell you how much time I have wasted on broadcasting bugs, where operations “work” but they aren’t doing what I want them to.
Jax can do this but no one uses Jax because of other reasons.
What we learned from Prolog is mostly that starting from an exponentially-complex primitive and then trying to beat it into submission doesn't work at scale. Relational DBs don't have that problem. They do go n-squared and n-cubed and so forth easily, but there are lots of solutions to that as well.
I helped on a language called Eve about 10 years ago. A truly relational language was exactly what that language was supposed to be, or at least that's what we were aiming at as a solution for a user-centric programming language.
The language we came up with was sort of like Smalltalk + Prolog + SQL. Your program was a series of horn clauses that was backed by a Entity-Attribute-Value relational database. So you could write queries like "Search for all the clicks and get those whose target is a specific id, then as a result create a new fact that indicates a a button was pressed. Upon the creation of that fact, change the screen to a new page". We even played around with writing programs like this in natural language before LLMs were a thing (you can see some of that here https://incidentalcomplexity.com/2016/06/10/jan-feb/)
Here's a flappy bird game written in that style: https://play.witheve.com/#/examples/flappy.eve
It's very declarative, and you have to wrap you brain around the reactivity and working with collections of entities rather than individual objects, so programming this way can be very disorienting for people used to imperative OOP langauges.
But the results are that programs are much shorter, and you get the opportunity for really neat tooling like time travel debugging, where you roll the database back to a previous point; "what-if" scenarios, where you ask the system "what would happen if x were y" and you can potentially do that for many values of y; "why not" scenarios, where you ask the system why a value was not generated; value providence, where you trace back how a value was generated... this kind tooling that just doesn't exist with most languages due to how they languages are built to throw away as much information away as possible on each stage of compilation. The kind of tooling I'm describing requires keeping and logging information about your program, and then leveraging it at runtime.
Most compilers and runtimes throw away that information as the program goes through the compilation process and as its running. There is a cost to pay in terms of memory and speed, but I think Python shows that interpretation speed is not that much of a barrier to language adoptions.
But like I said, that was many years ago and that team has disbanded. I think a lot of what we had in Eve still hasn't reached mainstream programming, although some of what we were researching found its way into Excel eventually.
> Loosen Up The Functions... Capabilities... Production-Level Releases... Semi-Dynamic Language... Modular Monoliths
I really like where the author's head at, I think we have similar ideas about programming because I've been developing a language called Mech that fits these descriptors to some degree since Eve development was shut down.
https://github.com/mech-lang/mech
So this language is not supposed to be relational like Eve, but it's more like Matlab + Python + ROS (or Erlang if you want to keep it in the languages domain).
I have a short 10 min video about it here: https://www.hytradboi.com/2022/i-tried-rubbing-a-database-on... (brief plug for HYTRADBOI 2025, Jamie also worked on Eve, and if you're interested in the kinds of thing the author is, I'm sure you'll find interesting videos at HYTRADBOI '22 archives and similarly interested people at HYTRADBOI '25), but this video is out of date because the language has changed a lot since then.
Mech is really more of a hobby than anything since I'm the only one working on it aside from my students, who I conscript, but if anyone wants to tackle some of these issues with me I'm always looking for collaborators. If you're generally interested in this kind of stuff drop by HYTRADBOI, and there's also the Future Of Coding slack, where likeminded individuals dwell: https://futureofcoding.org. You can also find this community at the LIVE programming workshop which often coincides with SPLASH: https://liveprog.org
Re-reading the eve website now, with 10+ years more experience and understanding of languages I'm really astounded at how brilliant Eve was, and how far ahead of it's time it was (and still is). Also at how rare it is to have any revolutionary ideas in modern programming language design make it out of theory in contemporary times. There were many radical ideas in the 60 and 70s, but so much now is incremental.
It's a shame Eve couldn't continue, just to see what it would've become and the influence it would have had on language expectations. Really cool stuff in there. While not likely, I hope someone picks up those ideas and continues them.
Did the effort just run out of funding? Or did it hit a stumbling block?
As for continuing the ideas, I'm putting a lot of what we learned into Mech. I know that Chris and Josh went to work at Relational AI, but I'm not sure exactly what work they got up to there. But Chris recently posted about generative AI and how it goes back to things we were thinking about in 2015 with respect to Eve: https://x.com/ibdknox/status/1630548754238435330
So in that sense generative AI can be a vector for some of the ideas in Eve coming to the mainstream.
For what I was using it for at the time (~3 years ago when I used it seriously) it offered performance close to the compiled orbital analysis code we had (in more conventional languages, Fortran and C) but with the flexibility for developing models of Python and other dynamic/interactive languages. An excellent tradeoff: very small performance cost for better interactivity and flexibility.
In your informed opinion, how would it make sense to be thrilled thinking about (not just a semi-dynamic APL (=S-DAPL?) as above but) designing
a APL-kith for widespread adoption by BOTH (demoscenic)gamecoders & Analysts (the academic variety)???
[0] https://www.mcmillen.dev/language_checklist.html
(Note in particular that the very APL inspired Wolfram has monopoly with physicists BUT does nothing for engineers
https://www.stephenwolfram.com/media/physics-whiz-goes-into-...
1988
>“But we tricked him, so to speak,” says Nobelist Murray Gell-Mann, who helped to bring Wolfram west. “We gave him a Ph.D.” )
I'm wondering if languages could provide some kind of meta information, hooks or extension points, which could be used to implement big ideas on top. These big ideas could then be reused and modified depending on the needs of the project.
I wonder if "Programming languages seem to have somewhat stagnated to me.", a sentiment I share, is just me paying less attention to them or a real thing.
I feel like in the next few years in languages will be things like Rust descendants where people with experience in using Rust want to keep what works for them but scales back some of the rigidity in favour of pragmatism.
It's also with noting that there are existing languages that are also changing over time. Freepascal has developed a lot of features over the years that make it fairly distant from original Pascal. More recent languages like Haxe are still developing into their final form. TypeScript has gone from a language that provided a tangible solution to an existing problem to a quagmire of features that I'd rather not have.
All in all a very enjoyable post.
It's done in a way that allows for a lot of subtlety, too. Basically you can use "dynamic" in lieu of most type annotations, and what this does is make any dispatch (in a broad sense - this includes stuff like e.g. overload resolution, not just member dispatch) on that particular value dynamic, but without affecting other values involved in the expression.
It specifically allows for multiple dispatch that is not even available in most dynamic languages not called lisp!
That said, here's some of the feedback of the type you said you didn't want >8)
(1) Function timeouts. I don't quite understand how what you want isn't just exceptions. Use a Java framework like Micronaut or Spring that can synthesize RPC proxies and you have things that look and work just like function calls, but which will throw exceptions if they time out. You can easily run them async by using something like "CompletableFuture.supplyAsync(() -> proxy.myCall(myArgs))" or in Kotlin/Groovy syntax with a static import "supplyAsync { proxy.myCall(myArgs) }". You can then easily wait for it by calling get() or skip past it. With virtual threads this approach scales very well.
The hard/awkward part of this is that APIs are usually defined these days in a way that doesn't actually map well to standard function calling conventions because they think in terms of POSTing JSON objects rather than being a function with arguments. But there are tools that will convert OpenAPI specs to these proxies for you as best they can. Stricter profiles that result in more idiomatic and machine-generatable proxies aren't that hard to do, it's just nobody pushed on it.
(2) Capabilities. A language like Java has everything needed to do capabilities (strong encapsulation, can restrict reflection). A java.io.File is a capability, for instance. It didn't work out because ambient authority is needed for good usability. For instance, it's not obvious how you write config files that contain file paths in systems without ambient authority. I've seen attempts to solve this and they were very ugly. You end up needing to pass a lot of capabilities down the stack, ideally in arguments but that breaks every API ever designed so in reality in thread locals or globals, and then it's not really much different to ambient authority in a system like the SecurityManager. At least, this isn't really a programming language problem but more like a standard library and runtime problem.
(3) Production readiness. The support provided by app frameworks like Micronaut or Spring for things like logging is pretty good. I've often thought that a new language should really start by taking a production server app written in one of these frameworks and then examining all the rough edges where the language is mismatched with need. Dependency injection is an obvious one - modern web apps (in Java at least) don't really use the 'new' keyword much which is a pretty phenomenal change to the language. Needing to declare a logger is pure boilerplate. They also rely heavily on code generators in ways that would ideally be done by the language compiler itself. Arguably the core of Micronaut is a compiler and it is a different language, one that just happens to hijack Java infrastructure along the way!
What's interesting about this is that you could start by forking javac and go from there, because all the features already exist and the work needed is cleaning up the resulting syntax and semantics.
(4) Semi-dynamic. This sounds almost exactly like Java and its JIT. Java is a pretty dynamic language in a lot of ways. There's even "invokedynamic" and "constant dynamic" features in the bytecode that let function calls and constants be resolved in arbitrarily dynamic ways at first use, at which point they're JITd like regular calls. It sounds very similar to what you're after and performance is good despite the dynamism of features like lazy loading, bytecode generated on the fly, every method being virtual by default etc.
(5) There's a library called Permazen that I think gets really close to this (again for Java). It tries to match the feature set of an RDBMS but in a way that's far more language integrated, so no SQL, all the data types are native etc. But it's actually used in a mission critical production application and the feature set is really extensive, especially around smooth but rigorous schema evolution. I'd check it out, it certainly made me want to have that feature set built into the language.
(6) Sounds a bit like PL/SQL? I know you say you don't want SQL but PL/SQL and derivatives are basically regular programming languages that embed SQL as native parts of their syntax. So you can do things like define local variables where the type is "whatever the type of this table column is" and things like that. For your example of easily loading and debug dumping a join, it'd look like this:
DECLARE
-- Define a custom record type for the selected columns
TYPE EmpDept IS RECORD (
name employees.first_name%TYPE,
salary employees.salary%TYPE,
dept departments.department_name%TYPE
);
empDept EmpDept;
BEGIN
-- Select columns from the joined tables into the record
SELECT e.first_name, e.salary, d.department_name INTO empDept
FROM employees e JOIN departments d ON e.department_id = d.department_id
WHERE e.employee_id = 100;
-- Output the data
DBMS_OUTPUT.PUT_LINE('Name: ' || empDept.name);
DBMS_OUTPUT.PUT_LINE('Salary: ' || empDebt.salary);
DBMS_OUTPUT.PUT_LINE('Department: ' || emptDebt.name);
END;
(7) I think basically all server apps are written this way in Java, and a lot of client (mobile) too. It's why I think a language with integrated DI would be interesting. These frameworks provide all the features you're asking for already (overriding file systems, transactions, etc), but you don't need to declare interfaces to use them. Modern injectors like Avaje Inject, Micronaut etc let you directly inject classes. Then you can override that injection for your tests with a different class, like a subclass. If you don't want a subtyping relationship then yes you need an interface, but that seems OK if you have two implementations that are really so different they can't share any code at all. Otherwise you'd just override the methods you care about.
Automatically working out the types of parameters sounds a bit like Hindley-Milner type inference, as seen in Haskell.
(8) The common way to do this in the Java world is have an annotation processor (compiler plugin) that does the lints when triggered by an annotation, or to create an IntelliJ plugin or pre-canned structural inspection that does the needed AST matching on the fly. IntelliJ's structural searches can be saved into XML files in project repositories and there's a pretty good matching DSL that lets you say things like "any call to this method with arguments like that and which is inside a loop should be flagged as a warning", so often you don't need to write a proper plugin to find bad code patterns.
I realize you didn't want feedback of the form "but X can do this already", still, a lot of these concepts have been explored elsewhere and could be merged or refined into one super-language that includes many of them together.
In Common Lisp there are tricks you can pull like declaring functions in a lexical scope (using labels or flet) to remove their lookup overhead. But CL is generally fast enough that it doesn't really matter much.
I seem to remember that Common Lisp compilers are allowed to do this for functions that are in the same file even if they are not declaimed inline. If A and B are in the same file, and B is not declaimed notinline (the opposite of inline), then A can be translated to assume the B definition.
So all your helper functions in a Lisp module are allowed to be called more efficiently, not having to go through the function binding of the symbol.
For instance, in Zig, you can see that a function allocates memory (capability) because it requires you to pass an allocator that it can call!
I'd like to see if others are more creative than me!
The trick is making sure that that object is the Only possible way to do the thing. And making more features like that, for example Networking, or File I/O, etc
Perhaps programming language is not the right abstraction to implement capability and need strong support from hardware and OS. OS with CPU based memory segmentation is an old idea probably worth re-exploring.
Implementing capability in the programming languages constructs will only increase cognitive overload and it will not be helpful for the programmer productivity [1].
b) Semi-Dynamic Language:
Dynamic language is what we want but static language is what we need [TM]. Instead of making dynamic language more static why not make static language more dynamic?
I think D language is moving the right direction with the default GC, RDMD based scriting, CTFE, and Variant based standard library features [2].
c) A Truly Relational Language:
Relational is only but one of the techniques of data processing, and other popular ones are spreadsheet, graph and matrices.
Rather than constraint programming languages with relational native constructs, better to provide generic mechanisms with fundamental constructs for data with associative array algebra [2].
d) Value Database:
This very much relates with (c) and can be excellent by-products and side-effects solution of it.
e) A Language To Encourage Modular Monoliths:
I think this is best point and idea from the entire article but as it rightly pointed out this is mainly architecture problem and programming language plays as a supporting role. It's the same as the Internet is based on packet switching rather than circuit switching, regardless of the RFCs and standards are being implemented in any languages.
However, for OS architecture with regard to Linus vs Tanembaum debate, modular monolithic is what we have now and the most popular in the form Linux, Windows and some say MacOS, and together they cover more than 99% of our OSes.
[1] Cognitive load is what matters:
https://news.ycombinator.com/item?id=42489645
[2] std.variant:
https://dlang.org/phobos/std_variant.html
[3] Mathematics of Big Data: Spreadsheets, Databases, Matrices, and Graphs:
https://mitpress.mit.edu/9780262038393/mathematics-of-big-da...
- I like the idea of a multiparadigm programming language (many exists) but where you can write part of the code in a different language, not trying to embed everything in the same syntax. I think in this way you can write code and express your ideas differently.
- A [social] programming language where some variables and workflows are shared between users [1][2].
- A superreflective programming language inspired by Python, Ruby, and others where you can override practically everything to behave different. For example, in Python you can override a function call for an object but not for the base system, globals() dict cannot be overriden. See [3]. In this way you save a lot of time writing a parser and the language basic logic.
- A declarative language to stop reinventing the wheel: "I need a website with a secure login/logout/forgot_your_password_etc, choose a random() template". It doesn't need to be in natural language though.
[1] https://blog.databigbang.com/ideas-egont-a-web-orchestration...
[2] https://blog.databigbang.com/egont-part-ii/
[3] https://www.moserware.com/2008/06/ometa-who-what-when-where-...
Modern web frameworks are getting pretty declarative. If you want a basic web app with a log in/out page that's not hard to do. I'm more familiar with Micronaut than Spring but you'd just add:
micronaut.security.authentication=cookie
https://micronaut-projects.github.io/micronaut-security/late...
Yes, SQL or a global spreadsheet. I would say that it is like SQL plus a DAG or, we can imagine an aggregation of SQLs. The interesting thing is that parts of the global system are only recalculated if there is a change, like in a spreadsheet.
> a shared RDBMS where everyone has a user account and can create tables/share them with other users, built in security, etc. Splat a GUI on top and you have something similar.
We need a little bit more but not much more: security by namespaces and/or rows so the same database is shared but you can restrict who change what: your "rows" are yours. I think something like OrbitDB but with namespaces will be cool.
> Modern web frameworks are getting pretty declarative.
Yes but my proposal was at a higher level. I don't want to know what a cookie is when I just want to create a website. I am not saying that you can create complex components with this idea but you can create common use cases.
Edit, since I remembered Go has this behaviour: I think Go's auto-interfaces I think are easily one of its biggest selling points.
https://erights.org/elib/capability/ode/ode-capabilities.htm... is a good start.
But you use capabilities all the time... operating system users work that way. As a user, you can't "just" execute some binary somewhere and thereby get access to parts of the system your user doesn't have rights to. (Forget setuid for a second, which is intended precisely to get around this, and let's just look at the underlying primitive.)
Capabilities in programming languages take the granularity further down. You might call some image manipulation code in a way that it doesn't have the capability to manipulate the file system in general, for example, or call a function to change a user's login name with capabilities that only allow changing that user, even if another user ID somehow gets in there.
It would be a fairly comprehensive answer to the software dependency issues that continue to bubble up; it would matter less if a bad actor took over "leftpad" if leftpad was actively constrained by the language to only be able to manipulate strings, so the worst an actor could do is make it manipulate strings the wrong way, rather than start running arbitrary code. Or put another way, if the result of the bad actor taking the package wasn't that people got hacked but users started getting
compile error in file.X:28: library "leftpad" tried to open a file without file system capabilities
compile error in file.X:30: library "leftpad" tried to open a socket without network capabilities
It's not a new idea, in that E already tried it, and bits and pieces of it are everywhere ("microkernels" is another place where you'll see this idea, but at the OS level and implemented in languages that have no native concept of the capabilities), but for the most part our programming languages do not reflect this.
Most operating systems don't have proper capabilities - they use things like ACLS, RBAC, MAC, etc for permissions.
The golden rule of capabilities is that you should not separate designation from authority. The capability itself represents the authority to access something, and designates what is being accessed.
It has this other link that explains more on it : https://en.m.wikipedia.org/wiki/Capability-based_security
I think I get it now. I honestly never had heard about this before and trying to Google search from the original post I was coming up empty.
There's some interest in the same thing, but at the programming language level. I'm only aware of it being implemented academically.
[1]: https://man7.org/linux/man-pages/man7/capabilities.7.html [2]: https://man.openbsd.org/pledge.2 [3]: https://man.openbsd.org/unveil.2
In a capabilities system, a program has permission to act on any object if it has a reference (aka a capability) to the object, there is no other access control. A program acquires a capability either by receiving it from is parent (or caller in the case of a function) or some other way like message passing. There is no other source of capabilities and they are unforgeable.
Unix file descriptors act in many ways as capabilities: they are inherited by processes from their parents and can be passed around via Unix sockets, and grant to the FD holder the same permissions to the referenced object as the creator of the file descriptor.
Of course as Unix has other ways from creating file descriptors other than inheritance and message passing is not truly a capabilities system.
What people often mean when they say this is a so-called pure capability system, where there are no ambient permissions at all. Such systems have terrible usability and indeed have never been made to work anywhere, not even in academia as far as I know.
So, no, the author claims it too.
Capabilities are a way to do access control where the client holds the key to access something, instead of the server holds a list of what is allowed based on the clients identities.
But when people use that word, they are usually talking about fine-grained access control. On a language level, that would mean not granting access for example for a library to do network connections, even though your program as a whole has that kind of access.
For example, consider a simple function to copy files. We could implement it like this:
def copy(fs: Filesystem, in: Path, out: Path) {
inH: HandleRead = fs.openRead(in);
outH: HandleWrite = fs.openWrite("/tmp/TEST_OUTPUT");
finished: Boolean = false;
while (!finished) {
match (inH.read()) {
case None: finished = true;
case Some(data) = outH.write(data);
}
}
inH.close();
outH.close();
}
In contrast, consider this alternative implementation:
def copy(inH: HandleRead, outH: HandleWrite) {
finished: Boolean = false;
while (!finished) {
match (inH.read()) {
case None: finished = true;
case Some(data) = outH.write(data);
}
}
inH.close();
outH.close();
}
In fact, we can see the same thing on the CLI:
- The first version is like `cp oldPath newPath`. Here, the `cp` command needs access to the filesystem, it needs permission to open files, and we have to trust that it won't open the wrong files.
- The second version is like `cat < oldPath > newPath`. The `cat` command doesn't need any filesystem access or permissions, it just dumps data from stdin to stdout; and there's no way it can get them mixed up.
The fundamental idea is that trying to choose whether an action should be allowed or not (e.g. based on permissions) is too late. It's better if those who shouldn't be allowed to do an action, aren't even able to express it at all.
You're right that this can often involve "keys", but that's quite artificial: it's like adding extra arguments to each function, and limiting which code is scoped to see the values that need to be passed as those arguments (e.g. `fs.openRead(inPath, keyThatAllowsAccess)`), when we could have instead scoped our code to limit access to the functions themselves (though for HTTP APIs, everything is a URL; so "unguessable function endpoint URL" is essentially the same as "URL with secret key in it")
---
To give an example of where this has been used in a programming language, Kernel[1] uses a capability model for mutating environments. Every function (or operative) has an environment which holds all of its local variables, the environment is encapsulated and internally holds a reference to the parent environment (the surrounding scope). The rule is that we can only mutate the local variables of an environment to which we have a direct reference, but we cannot mutate variables in the parents. In order to mutate the variables in the parent, we must have a direct reference to the parent, but there is no mechanism in the language to extract the parent reference from the environment it is encapsulated in.
For example, consider the following trivial bit of code: We define some variable `x` with initial value "foo", we then mutate it to have the value "bar", then look up `x`.
($define! x "foo")
($set! (get-current-environment) x "bar")
x
Technically we could've just written `($define! x "bar")`, where the current environment is assumed, but I used `$set!` because we need it for the next example.
When we introduce a new scope, the story is different.
($define! x "foo")
($define! foo
($lambda ()
($set! (get-current-environment) x "bar")))
(foo)
x
So far basically the same static scoping rules you are used to, but environments in Kernel are first-class, so we can get a reference to the top-level environment which grants the child environment the authority to mutate the top level environment.
($define! x "foo")
($define! env (get-current-environment))
($define! foo
($lambda ()
($set! env x "bar")))
(foo)
x
However, by binding `env` in the top-level environment, all child scopes can now have the ability to mutate the top-level.
To avoid polluting the environment in such way, the better way to write this is with an operative (as opposed to $lambda), which implicitly receives the caller's environment as an argument, which it binds to a variable in its local environment.
($define! x "foo")
($define! foo
(wrap ($vau () caller-env
($set! caller-env x "bar"))))
(foo)
x
---
This is only a trivial example, but we can do much more clever things with environments in Kernel. We can construct new environments at runtime, and they can have multiple parents, ultimately forming a DAG, where environment lookup is a Depth-First-Search, but the references to the parent environments are encapsulated and cannot be accessed, so we cannot mutate parent scopes without a direct reference - we can only mutate the root node of the DAG for an environment to which we have a direct reference. The direct reference is a capability - it's both the means and the authority to mutate.
---
We can use these first-class environments in conjunction with things like `$remote-eval`, which evaluates some piece of code in an environment provided by the user, which may contain only the bindings they specify, and does not capture anything from the surrounding scope.
($define! calculator-environment
($bindings->environment (+ +) (- -) (* *) (/ /)))
($remote-eval (+ 1 2) calculator-environment)
($remote-eval (write (+ 1 2)) calculator-environment)
This combination of features basically let you create "mini sandboxes", or custom DSLs, with more limited capabilities than the context in which they're evaluated. Most languages only let you add new capabilities to the static environment, by defining new functions and types - but how many languages let you subtract capabilities, so that fewer features are available in a given context? Most languages do this purely at compile time via a module/import system, or with static access modifiers like `public` and `private`. Kernel lets you do this at runtime.
---
One thing missing from this example, which is required for true capabilities, is the ability to revoke the authority. The only way we could revoke the capability of a function to mutate an environment is to suspend the program.
Proper capabilities allow revocation at any time. If the creator of a capability revokes the authority, this should propagate to all duplicated, delegated, or derived capabilities with immediate effect. The capabilities that were held become "zombies", which no longer provide the means nor the authority - and this is why it is essential that we don't separate designation from authority, and why these should both be encapsulated in the capability.
This clearly makes it difficult to provide proper capabilities in programming languages, because we have to handle every possible error where we attempt to access a zombie capability. The use of such capabilities should be limited to where they really matter such as access to operating system resources, cryptographic keys, etc. where it's reasonable to implement robust error handling code. We don't want capabilities for every programming language feature because we would need to insert error checks on every expression to handle the potential zombie. Attempting to check if a capability is live before using it is no solution anyway, because you would have race conditions, so the correct approach to using them is to just try and catch the error if it occurs.
Another take-away from this is that if capabilities are provided in a language via a type system, it must be a dynamic type system. You cannot grant authority in a static type system at compile time if the capability may have already been revoked by the time the program is run. Capabilities are inherently dynamic by nature because they can be revoked at any time. This doesn't mean you can't use capabilities in conjunction with a static type system - only that the static type system can't really represent capabilities.
You can find out a lot more about them on the erights page that others have linked, and I would recommend looking into seL4 if you're interested in how they're applied to operating systems.
---
This is just blocking code and it’s beautiful.
"Value Database" - Mumps? lol
As they should be. Not every language needs to turn into C++, Rust, Java, C#, or Kotlin.
The only group I see lamenting about features these days are PL theorists, which is fine for research languages that HN loves but very few use outside the bubble.
Some of us like the constraints of C or Go.
https://www.youtube.com/watch?v=BSymxjrzdXc
I found it amusing most of the language is supposedly contextually polymorphic by definition. =3
I mean, there still may be other things that make this a Bad Idea(TM). But I think this could get around the specific issue mentioned in the article.
If it's about well-contained applications in a well designed (and user-centric) OS with a proper concept of "application" and "installation", with a usable enough mechanism, I don't see anything that would make it bad.
On Windows it's a disaster. To the point that dumping random text files around in Linux works better.
* In general, whenever I hear "compiler will optimize this", I die a little on the inside. Not even because it's delegating solution of the newly created problem to someone else, but because it creates a disconnect between what the language tells you is possible and what actually is possible. It encourages this kind of multi-layer lie that, in anger, you will have to untangle later, and will be cursing a lot, and will not like the language one bit.
* Capabilities. Back in the days when ActionScript 3 was relevant, there was a big problem of dynamic code sharing. Many services tried to implement module systems in AS3, but the security was not done well. To give you some examples: a gaming portal written in AS3 wants to load games written by programmers who aren't the portal programmers (and could be malicious, i.e. trying to steal data from other programs, or cause them to malfunction etc.) ActionScript (and by extension JavaScript 4) had a concept of namespaces borrowed from XML (so not like in C++), where availability of particular function was, beside other things, governed by whether the caller is allowed to access the namespace. There were some built-in namespaces, like "public", "private", "protected" and "internal" that functioned similar to Java's namesakes. But users were allowed to add any number of custom namespaces. These namespaces could be then shared through a function call in a public namespace. I.e. the caller would have to call the function and supply some kind of a password, and if password matched, the function would return the namespace object, and then the caller could use that namespace object to call the functions in that namespace. I tried to promote this concept in Flex Framework for dealing with module loading, but that never was seriously considered... Also, people universally hated XML namespaces (similar to how people seem to universally hate regular expressions). But, I still think that it could've worked...
* All this talk about "dynamic languages"... I really don't like it when someone creates a bogus category and then says something very general about it. That whole section has no real value.
* A Truly Relation Language -- You mean, like Prolog? I wish more relational databases exposed their content via Prolog(like) language in addition to SQL. I believe it's doable, but very few people seem to want it, and so it's not done.