Now that I'm out of the corporate tyranny and have my own company, I use lisp for everything. There's certain satisfaction in writing config files and persisting data directly in s-expressions. Any json requirements are triggered by exports to foreign systems.
Does it really matter? There's a point in every Lisper's life, a threshold after which the question becomes immaterial - you'd stop thinking about intricacies of whatever Lisp and focus on the platform specifics instead. Any given day I might program in three-four different Lisp dialects, e.g. Clojure/Clourescript, Fennel, Elisp, Janet, etc. and it practically feels like I'm using the same PL. While switching between TS and JS (same family) never feels even close - there's always some mental burden.
not philosophically, but certainly practically. To double down, if all lisps are roughly equivalent from a language POV, then you'd want to pick the one or two that will give you the most practical advantage (libraries, documentation, dev environment, etc.)
I would call these different dialects of Lisp. The data doesn’t have to be a function. It’s illustrative. The patterns of application still work. What’s the difference if delimiters are different or if you are calling JVM libraries? The high-level ideas are still right there. Consider JavaScript. It is definitely not a Lisp, but if you model it as Lisp in C’s clothes, then all of a sudden IIFEs make total sense. The point is that it’s a helpful mental model for languages other than Lisp.
Limitation is the wrong way to think about things when computational equivalence is in play. It's about mental foundation. Lisp at its core is like driving a Turing machine, Clojure is not.
If the difference didn't matter, we wouldn't have so many different lisps. Obviously the difference mattered enough to the people that created Common Lisp when Scheme already existed. Rich Hickey thought it mattered when he created a completely new Lisp instead of just porting Scheme to the JVM.
> If the difference didn't matter, we wouldn't have so many different lisps
Literally the opposite. We can make and use so many, because writing them is more or less the same. We can quickly throw together a new lisp for a new platform or such and use it without problem.
Technically when you write in the domain, you are effectively making your own Lisp and then using it. It’s one of the amazing things that macros can do.
It's the mustard seed that keeps growing! And without the fragmentation other PL's suffer from, which are reminiscent of the Tower of Babel, conversely.
One way I find traditional Lisp style more painful for functional code than Ruby is that fully functional-style Lisp pushes me to read and write code the opposite way from how I think about it. In the author's example:
the equivalent Lisp code would either be written in imperative style as multiple statements that each write to a temporary variable or (let) binding, or would look like this:
(reduce #'+
(map (lambda (o) (getf o 'total))
; this group_by replacement function
; might be written as hash-table code
(my-group-by 'customer-id
(remove-if-not
(lambda (o)
(>
(getf o 'placed-at)
(- (my-now) (* 60 60 24 7))))
orders))))
where I now have to read from bottom to top to understand the order of operations on the `orders` record set, even though when I wrote the code earlier, I "logically" thought from first operation to last when deciding which high-level operations to use in which order.
Other imperative languages that support functional code either make you do things imperatively to get the "logical" ordering of functional operations like I feel Lisp pushes you to do, or they do something like Ruby where things can be chained left to right in a "single" statement even for operations that were not thought of ahead of time by the creators of opaque data structures you later need to operate on. (Everything is a user-extensible object like Ruby, unified function call syntax in D, extension methods in C#, or pipelines of structured objects in PowerShell.)
But I prefer the typical Lisp code where I get the sums of the totals of the orders with the same customer ID which were placed in the past week, instead of the orders made the past week grouped by customer ID their totals summed together.
In contrast to, say, Java (I can't speak to the code above):
List<Things> things = thingIds.stream()
.map(model::findThing)
.filter(Objects::nonNull)
.toList();
These are streamed. This is pretty much a pipe structure, whereas the threading macros will create a lot of temporary copies of the data (I don't know if that's a universal truth). That is, if you're processing a 1000 items, say `gather` returns a 1000 items, that 1000 item list is passed to `uppercase-list` which return a new 1000 item list to feed to `sort` which returns another 1000 item list (assuming none of these are destructive).
I wish CL had something like the Java streams (maybe it does).
The version with a threading macro, will create a lazy-sequence for each step in the pipeline. It will not instantiate the entire list, so it's O(1) memory overhead in terms of peak memory, but it churns O(N) extra garbage.
(->> things
(map model/find-thing)
(filter some?))
And the version with transducers, which will not create any intermediate sequences:
It's common to write the thrush combinator as a lisp macro. Clojure ships ->, ->>, as->, some->, some->>, cond->, and cond->> out of the box. You can find similar macros for CL[0], Racket[1], and a scheme SRFI[2]. Writing them is a fun exercise in your lisp of choice if you don't have a library available.
Using a threading operator where there is no such consistency is painful. This is why I dislike CL’s or Python’s map function, taking the list to operate on as second argument, instead of first. A threading operator wouldn’t be as effective there.
The issue is that these functions in Lisps are variadic and can accept more arguments than one. `map`, and `zipwidth` in lisps are actually the same function.
I've never been more thoroughly convinced that I would like ruby more than from this article. I'm currently stuck reaching for python a lot of the time (absolutely love it tbc), but maybe it's worth changing things up and trying to give ruby a shot.
It was one of the first programming languages I was introduced to at 16 or so, but an older person that I looked up to told me it would get me stuck in "hobby coder land". He was wrong in so many ways, but even if he was right, I wanna have fun in my hobby code :)
I've had one job in my life, still at the same company. (8 years).
I applied cause the listing mentioned Python, and I was programming in Python at the time.
Once I started they were like yeah we put that there to reach a broader public but we use Ruby (on Rails).
So that's what I learned. I've just returned to Python via LLM's. I literally have not felt the need nor desire to use Python once I got used to writing Ruby.
I love Ruby, use it for most of my projects that don't require performance.
Nothing I would love more than a Ruby with a Common-Lisp like compiler and runtime. Unboxed types, native compilation, partial compilation, live image (Ruby has this but "faster Rubies" like Crystal don't), etc...
I have a (self-hosted, but buggy and wildly incomplete; don't try to use - jRuby or TruffleRuby are better - and far faster - options) Ruby compiler that was partly born out of wanting to figure out what this would take, and the answer is it is massively painful because Ruby has failed to take some basic steps that makes delineating read-time and run-time very hard (e.g. you have fun patterns like overriding "require", and iterating over directories to decide what to require) even though most Ruby programs do have clearly separate load and run phases. It's just hard to programmatically separate it.
I still believe you could do pretty well there with a few basic "tricks" that could still also remain real/valid Ruby, by recognising the most common patterns, documenting them, and providing a way of marking exceptions. Combine that with freezing system classes after startup as an enabler for various optimization, and a compiler could do a pretty good job. But it's a massive piece of work to get it right for Ruby.
Common Lisp, and even more so Racket, has reader macros. With a little help from LLMs you might be able to get a Ruby-like language that translates into Lisp.
As a last resort look at Racket's "Rhombus" language, it's basically an infix, Python-like syntax on top of Racket. You can use that or see how they pull it off and add Ruby constructs to it.
Elixir has forever ruined me for other languages. Every new PL I dip my toe into gets measured against it. Jose and the core team seem to always land on the right decisions, or at least very good ones.
Most of the points listed are hardly considered lispy anymore these days, Python also has most of these.
Where Ruby's lisp lineage really shows is the fact that it's got Kernel#callcc, aka call with current continuation. It doesn't get any lispier than that!
As I read this, all of my favorite things about ruby compared to something like python were influenced by lisp.
Ruby is a joy to program and it seems mostly due to design influence from lisp.
> Matz has said as much. He’s described Ruby’s design as starting from a simple Lisp, stripping out macros and s-expressions, then adding an object system, blocks, and Smalltalk-style methods. The features most Rubyists fall in love with aren’t the object-oriented ones. They’re the functional ones, dressed in friendlier clothes.
> For better or worse, parenthesis aren't that bad with the proper IDE tooling.
Hell, even without [0], you can at least count the parenthesis by hand in a pinch. I remember seeing lots of crazy-awesome stuff done in AutoLisp by 'non-programmers', versus 'structure as spacing' in Python which really sucks if the Editor was designed to use the system default (probably non-monospaced, cause other products in the industry had dialogs that broke if you switched to a monospaced) font. [1]
[0] - but real talk parenthesis matching in an editor is a lifesaver
[1] - oooooold version of a very popular GIS product.
Reminds me of my first exposure to Lisp. It was in a survey course in my final year. This was using punched card input and output on a line printer. You could turn on paren counting, but it was clumsy at best. This was in 1970
Totalle agree, I just googled it:
"Yukihiro 'Matz' Matsumoto heavily credits Smalltalk as the deepest structural inspiration behind Ruby’s object model. He combined Smalltalk’s beautiful object-oriented architecture and message-passing system with features from other languages to create a tool designed primarily for developer happiness."
Including the closures and collection operations.
No, its actual influence from Lisp-family languages (including Scheme). Yes, Lisp also influenced Perl and Smalltalk, but Matz was not ignorant of Lisp with the only influence om Ruby from Lisp being indirect through those other languages.
Matz directly credits Lisp (through Emacs Lisp) as influence in the design of Ruby and its runtime, with Smalltalk influence on the language itself, and IIRC Perl as "what was popular and we tried to replace"
You kind of don't need them in Ruby, because everything is a method or an object or a closure and you can dynamically create and alter those at runtime. That's why Ruby is really good for ad-hoc DSLs in ways that Rust and Swift really are not.
Crystal don't have the dynamicity but has macros to get the next best thing. Most meta magic in Ruby in good code are done at startup anyhow so you don't miss out on that much. YMMV.
C has macros too, but it's a second preprocessor language. They both accomplish metaprogramming, but it's questionable whether they're both the same lisplike "macros" we're talking about. Ruby source could be passed through the C preprocessor and get C macros that way. I've actually seen Java code that does just that.
C macros are definitely much weaker; they're not by themselves Turing-complete (except maybe with vendor-specific extensions? I'm not an expert here). Rust has both macros by example (precisely analogous to Scheme macros, and equal in power) and procedural macros (conceptually analogous to Common Lisp macros, allowing arbitrary code at macro evaluation time, but I don't know enough about Common Lisp to say whether there are differences in power).
How does it work internally? It would have to output the new source code as data somehow, and have the Rust compiler consume it. How does that happen?
The lispy "macros" I speak of are FEXPRs, just everyday normal functions that just happen to not evaluate their arguments, they receive the source code as lists instead. It's easy to manipulate those lists and evaluate the result.
Lisps themselves moved away from FEXPRs because they were "too powerful" and made the compiler's life hard. Common Lisp and Scheme macros are the more restricted versions that allow compilers to make more assumptions, thereby enabling more aggressive optimization.
Rust has two form of macros: “macros by example” and “procedural macros.”
The latter is basically a function from token streams to token streams, and macros by example are more traditional macros which were initially designed by Dave Herman, who was heavily involved in Racket.
Yes, a Rust procedural macro is a function that takes a Rust syntax tree as an argument and returns a Rust syntax tree. When you use it, the compiler compiles it (for the host architecture), dynamically loads it into the compiler process, calls it, and inserts the output into the code to be compiled. https://doc.rust-lang.org/book/ch20-05-macros.html#procedura...
I don't see why this would inhibit optimization, unless you mean it slows down compilation, in which case, yep, that's a real and rather notorious downside.
> the compiler compiles it (for the host architecture), dynamically loads it into the compiler process, calls it, and inserts the output into the code to be compiled
That's actually amazing. So the compiler's own data structures are visible in the language.
I also like how in Haskell:
Is an actually idiomatic way to deal with the lack of trailing commata.not philosophically, but certainly practically. To double down, if all lisps are roughly equivalent from a language POV, then you'd want to pick the one or two that will give you the most practical advantage (libraries, documentation, dev environment, etc.)
Literally the opposite. We can make and use so many, because writing them is more or less the same. We can quickly throw together a new lisp for a new platform or such and use it without problem.
I do love that I learnt Clojure once like 5-7 years ago and more and more dialects keep expanding the choice of runtime I can target
Other imperative languages that support functional code either make you do things imperatively to get the "logical" ordering of functional operations like I feel Lisp pushes you to do, or they do something like Ruby where things can be chained left to right in a "single" statement even for operations that were not thought of ahead of time by the creators of opaque data structures you later need to operate on. (Everything is a user-extensible object like Ruby, unified function call syntax in D, extension methods in C#, or pipelines of structured objects in PowerShell.)
https://docs.racket-lang.org/threading/introduction.html
The threading macros are (as I understand it) pure sugar.
Turning (-> (gather my-list) uppercase-list sort) into (sort (uppercase-list (gather my-list))).
In contrast to, say, Java (I can't speak to the code above):
These are streamed. This is pretty much a pipe structure, whereas the threading macros will create a lot of temporary copies of the data (I don't know if that's a universal truth). That is, if you're processing a 1000 items, say `gather` returns a 1000 items, that 1000 item list is passed to `uppercase-list` which return a new 1000 item list to feed to `sort` which returns another 1000 item list (assuming none of these are destructive).I wish CL had something like the Java streams (maybe it does).
The version with a threading macro, will create a lazy-sequence for each step in the pipeline. It will not instantiate the entire list, so it's O(1) memory overhead in terms of peak memory, but it churns O(N) extra garbage.
And the version with transducers, which will not create any intermediate sequences: It looks like there's a Common Lisp transducers library, but I have no idea how widely it's used.https://github.com/fosskers/transducers
edit SICP has examples on how to implement streaming (in Scheme).
https://docs.racket-lang.org/reference/streams.html
For a variadic language you'd need something more involved though. But some kind of syntax can probably be invented in some language.
[0] https://github.com/dtenny/clj-arrows
[1] https://docs.racket-lang.org/threading/index.html
[2] https://srfi.schemers.org/srfi-197/srfi-197.html
It was one of the first programming languages I was introduced to at 16 or so, but an older person that I looked up to told me it would get me stuck in "hobby coder land". He was wrong in so many ways, but even if he was right, I wanna have fun in my hobby code :)
I applied cause the listing mentioned Python, and I was programming in Python at the time.
Once I started they were like yeah we put that there to reach a broader public but we use Ruby (on Rails).
So that's what I learned. I've just returned to Python via LLM's. I literally have not felt the need nor desire to use Python once I got used to writing Ruby.
Nothing I would love more than a Ruby with a Common-Lisp like compiler and runtime. Unboxed types, native compilation, partial compilation, live image (Ruby has this but "faster Rubies" like Crystal don't), etc...
I still believe you could do pretty well there with a few basic "tricks" that could still also remain real/valid Ruby, by recognising the most common patterns, documenting them, and providing a way of marking exceptions. Combine that with freezing system classes after startup as an enabler for various optimization, and a compiler could do a pretty good job. But it's a massive piece of work to get it right for Ruby.
I'm Ruby or Lean 4.
Also I'm working on a DSL/Macros that give me more Ruby-esque quality of life things in Lisp.
As a last resort look at Racket's "Rhombus" language, it's basically an infix, Python-like syntax on top of Racket. You can use that or see how they pull it off and add Ruby constructs to it.
Heck at least in my brain MLs are closer to a Lisp than Ruby...
Reminds me of an email I wish I still had.
Circa 2000, I wrote that I was leaning towards moving to Scheme, for more rapid R&D work than I could do in Java.
Some nice-sounding person I didn't know emailed me from Japan, to mention a language I hadn't heard of, called Ruby.
I don't know whether the person was Yukihiro Matsumoto himself, but it's a small world.
Where Ruby's lisp lineage really shows is the fact that it's got Kernel#callcc, aka call with current continuation. It doesn't get any lispier than that!
https://www.youtube.com/watch?v=Qc7HmhrgTuQ
> Matz has said as much. He’s described Ruby’s design as starting from a simple Lisp, stripping out macros and s-expressions, then adding an object system, blocks, and Smalltalk-style methods. The features most Rubyists fall in love with aren’t the object-oriented ones. They’re the functional ones, dressed in friendlier clothes.
But yeah, macros and S-expressions make it easier to write your own DSLs.
For better or worse, parenthesis aren't that bad with the proper IDE tooling.
Hell, even without [0], you can at least count the parenthesis by hand in a pinch. I remember seeing lots of crazy-awesome stuff done in AutoLisp by 'non-programmers', versus 'structure as spacing' in Python which really sucks if the Editor was designed to use the system default (probably non-monospaced, cause other products in the industry had dialogs that broke if you switched to a monospaced) font. [1]
[0] - but real talk parenthesis matching in an editor is a lifesaver
[1] - oooooold version of a very popular GIS product.
(Matz speaking at the LL2 conference some 20+ years ago)
Even Ruby's trailing blocks syntax are an homage to Perl's block list subroutines:
Put the macros back! It would be so cool!
well, except for pattern matching. That is just syntax.
The lispy "macros" I speak of are FEXPRs, just everyday normal functions that just happen to not evaluate their arguments, they receive the source code as lists instead. It's easy to manipulate those lists and evaluate the result.
Lisps themselves moved away from FEXPRs because they were "too powerful" and made the compiler's life hard. Common Lisp and Scheme macros are the more restricted versions that allow compilers to make more assumptions, thereby enabling more aggressive optimization.
The latter is basically a function from token streams to token streams, and macros by example are more traditional macros which were initially designed by Dave Herman, who was heavily involved in Racket.
I don't see why this would inhibit optimization, unless you mean it slows down compilation, in which case, yep, that's a real and rather notorious downside.
That's actually amazing. So the compiler's own data structures are visible in the language.
I see how it works now. Thanks for explaining.