Lush: My favorite small programming language

What does 'small' really mean?

I would think of a language like Go as small (say, in comparison to Rust or Swift) - the language itself at least, if you discount the standard library.

I find the use of the word 'small' quite confusing.

do you already program with this language? what is your paradigm?

“Already”?

This is about a language abandoned 15 years ago!

It's buried in the article, but Lush is from 1987!

Previously:

https://news.ycombinator.com/item?id=34908067

https://news.ycombinator.com/item?id=9602430

https://news.ycombinator.com/item?id=2406325

Also this comment:

> "Lush" stands for "Lisp Universal Shell". It has not just S-expression syntax but recursion, setq, dynamic typing, quoting of S-expressions and thus lists and homoiconicity, cons, car, cdr, let*, cond, progn, runtime code evaluation, serialization (though bread/bwrite rather than read/print), and readmacros. Its object system is based on CLOS.

https://news.ycombinator.com/item?id=28728302

Makes me curious what state R was at the time, or whatever else could've been useful for deep learning, and the benefits of a new language vs adapting something that exists. Seems like it was a big investment

  SN(1987) neural network simulator for AmigaOS (Leon Bottou, Yann LeCun)
   |
  SN1(1988) ported to SunOS. added shared-weight neural nets and graphics (LeCun)
   |   \ 
   |   SN1.3(1989) commercial version for Unix (Neuristique)
   |   /
  SN2(1990) new lisp interpreter and graphic functions (Bottou)
     |   \ 
     |   SN2.2(1991) commercial version (Neuristique)
     |    |
     |   SN2.5(1991) ogre GUI toolkit (Neuristique)
     |   / \ 
      \ /  SN2.8(1993+) enhanced version (Neuristique)
       |     \ 
       |   TL3(1993+) lisp interpreter for Unix and Win32 (Neuristique)
       |      [GPL]
       |        \_______________________________________________
       |                                                        |
     SN27ATT(1991) custom AT&T version                          |
       |        (LeCun, Bottou, Simard, AT&T Labs)              |
       |                                                        |
     SN3(1992) IDX matrix engine, Lisp->C compiler/loader and   |
       |       gradient-based learning library                  |
       |       (Bottou, LeCun, AT&T)                            |
       |                                                        |
     SN3.1(1995) redesigned compiler, added OpenGL and SGI VL   |
       |         support (Bottou, LeCun, Simard, AT&T Labs)     |
       |                                                        |
     SN3.2(2000) hardened/cleanup SN3.x code,                   |
       |         added SDL support (LeCun)                      |
       | _______________________________________________________|
       |/
       |
     ATTLUSH(2001) merging of TL3 interpreter + SN3.2 compiler
     [GPL]         and libraries (Bottou, LeCun, AT&T Labs).
       |
     LUSH(2002) rewrote the compiler/loader (Bottou, NEC Research Institute)
     [GPL]
       |
     LUSH(2002) rewrote library, documentation, and interfaced packages
     [GPL]      (LeCun, Huang-Fu, NEC)

https://lush.sourceforge.net/credits.html

R and its ecosystem have some unbeatable features, but, generally speaking, the "old", base R is too arcane to be widely useful. Also, being "made by statisticians for statisticians" should be a big warning sign.

The author appears to be defining it in terms of the effort put in to the language, basically, person-hours.

Go may be a small language by some definitions (and as my phrasing implies, perhaps not by others), but it is certainly one that has had a lot of person-hours put into it.

The problem is that there's no universal definition of "small" when it comes to languages.

An article on the Brown PLT blog [1] suggests analyzing languages by defining a core language and a desugaring function. A small core simplifies reasoning and analysis but can lead to verbose desugaring if features expand into many constructs. The boundary between the core and sugared language is flexible, chosen by designers, and reflects a balance between expressiveness and surface simplicity.

Feature complexity can be evaluated by desugaring: concise mappings to the core suggest simplicity, while verbose or intricate desugarings indicate complexity.

So, a possible definition of a "small" language could be one with both a small core and a minimal desugaring function.

--

1: https://blog.brownplt.org/2016/01/08/slimming-languages.html

And that's related to someone liking a language how? Especially one that's dead for a lot time...

Not to mention; you seem to be religiously pushing react which is more of a dsl but still..

You mean what do i mean and what do you mean ? Thanks.

Where does Ralf Juengling's work on lush fit in to this picture?

Note that the diagram ends in 2002.

https://sourceforge.net/p/lush/mailman/message/20287123/

In my opinion R should thought of as an unbeatable graphical calculator, but an awful programming language.

The tinyverse collection of packages makes things a lot more sane, IMO:

  penguins <- read_csv("penguins.csv") |>
    na.omit() |>
    select(species, island, bill_length_mm, body_mass_g) |>
    group_by(species, island) |>
    summarize(
      mean_bill_length = mean(bill_length_mm),
      mean_mass = mean(body_mass_g),
      n = n()
    ) |>
    arrange(species, desc(mean_bill_length))
  
  penguins |>
    ggplot(aes(x = species, y = mean_bill_length, fill = island)) +
    geom_col(position = "dodge") +
    labs(
      title = "Mean Bill Length by Species and Island",
      y = "Mean Bill Length (mm)"
    ) +
    theme_minimal()

Fun fact: Lush was invented by Yann LeCun, of convnet and FAIR fame.

I love this diagram. Is there a tool that generates such things? Or is there a name for this style of diagram that I could search for?

My prime use would be generating diagrams of function call chains in large Python code bases.

i would compare base R to basically a shell. meant to be used interactively. okay for small scripts. you can write big programs but it will get weird.

True, but trying to wrap any of that into a function rather than simple scripts makes you delve into the ever-deprecated API for non-standard evaluation.

How about pycallgraph that can be exported to Graphviz?

FWIW it is called evolutionary or lineage (or hierarchical lineage) diagram I believe.

i am also interested in this.

i found vijual[1] and mermaid-ascii[2] are good starting projects.

[1]: http://www.lisperati.com/vijual/ [2]: https://github.com/AlexanderGrooff/mermaid-ascii

That's how I view it. I still use R for plotting and quick stats analyses but it is painful to do any real work.

I recommend the article "Evaluating the Design of the R Language" [1] - it reads like a horror story. The memory usage and performance is abysmal, the OO features are a mess, and the semantics are very weird ("best effort semantics" is about as predictable as it sounds!). The lexical scoping is based on Scheme but has so many weird edge cases. It's a dumpster fire of a language, but it somehow works for its intended purpose.

[1] http://janvitek.org/pubs/ecoop12.pdf

Despite being made by statisticians, I ironically find that munging R packages together for certain classes of analysis such a slog that it prevents me from doing the actual statistical thinking. Sometimes the plots fall behind commercial packages, sometimes the diagnostics, and sometimes you have to combine multiple incompatible packages to get what a commercial package can do.

(Survival analysis and multilevel modeling comes to mind.)

On the contrary, I find base R less arcane than the current de jour python libraries which copied it

This is so far from my experience. For me, R codes do tend to skimp on polish so it takes longer to get to the initial figure, but that is made up for by enabling me to see the data from a much richer perspective (to some extent because I had to think harder about what the output meant) such that I can find all the bugs in the data and in the underlying experimental plan: the stuff which makes it clear all the commercial reports are mostly useless anyway because Garbage in -> Garbage out

You must hate lisp/scheme then too, which has similar semantics as R. In that case books such as SICP would be lost on you.

But commits exists even this year: https://sourceforge.net/p/lush/activity/