The Racket Way | Hacker News Reader

The Racket Way

infoq.com

70 pointssutro44 comments

I put together a (silly) post to highlight some of the features of Racket. http://blog.fogus.me/2013/01/21/enfield-a-programming-langua... It's worth a look if you want the overview and don't have time for the video.

I've got my little "nights and weekend" project, with the web server part currently running on Google's app engine (http://h4labs.com). I was trying to use Scala for a while but now it's just HTML/Javascript.

Occasionally, I do get this urge to move to Amazon so I can "play" with Haskell, Go, Racket, Scala, etc. Has anyone created a "Frankenstein" website on their side-project site? It would be a great way to learn many of these "non-mainstream" languages.

I run my personal website on Racket. The documentation is great: http://docs.racket-lang.org/web-server/index.html

Ask pg - I think he's done something with Arc. Seems like anything he'd do'd be pretty frequently trafficked...

AFAIK, HN is done in Arc.

Arc is written in Racket.

The plt team should push that more in their advertising (though I guess they don't have an ad budget).

GAE supports Go and Java and, by extension, Scala, Clojure and all the other JVM languages.

Nice talk, I like how he is humble but enthusiast.

The records are destructive in the example at the end. Is working with immutable data structures in Racket as easy as in Clojure?

IIRC, data structures are immutable by default in Racket.

Lists are. Vectors and hashes are not.

There are both mutable lists and immutable vectors and hashes though.

Edit: and user-defined structs are immutable unless explicitly declared with #:mutable keyword.

I know Matthew personally, and he's like that all the time. Very laid-back, and also very go-getter. Difficult combination, but he pulls it off well.

Racket's libraries are designed to encourage doing everything with immutable data, though there are plenty of mutable types if you need them. (And of course you can always use boxes.) I don't know enough about Clojure to make a comparison, but I can say that working with immutable data structures in Racket is easy.

One of Clojure's main selling points is that its vectors, maps and sets are persistent data structures with near-constant time operations (log32N).

http://www.infoq.com/presentations/Value-Identity-State-Rich...

Off-topic Why racket is not more popular than other lisp dialects?

I am also interested to know how it compares to Clojure.

Because Racket has a run-time it competes directly with Clojure, which is obviously widely popular. You can bundle the Racket runtime in your binaries but startup is slow (much slower than Gambit, for example).

So for me at least, it's simple. If I don't care about startup time and just want the best ecosystem, Clojure is a no brainer. If I do care about startup, you have to pick one of the Lisps, probably either SBCL or Clozure.

Do you mean to pick one of the "Common Lisps"?

Yes, sorry.

Here's an older HN post comparing some of the pros and cons of the two languages: http://news.ycombinator.com/item?id=1594479

Don't read this post.

Racket since grew a full library of purely functional datastructures, cleaned up hashes, provides builtin streams and generators and as for the syntax - well, it's a lisp: write your own syntax and be done with it! ;) [edit: https://github.com/takikawa/racket-clojure/blob/master/lang/...]

Anyway, this post should be either updated or at least marked as not entirely relevant three years later.

That's good to know, thanks.

I'm taking a Coursera course that teaches Racket, so I just started learning it this week. First time using a Lisp/Scheme family language. So far I actually really like the use of parentheses and the concept of the code itself being a data structure. I'll probably give Clojure a try once I have a firm grasp on Racket, they seem quite similar.

The most glaring difference between the two would be that clojure runs on the JVM, while racket has its own custom VM that can better support trick features like continuations but is not operable with a big existing ecosystem like Java.

Clojure has some nice syntax and conventions for dealing with sequences, vectors, maps, and sets, such that it encourages working with those data structures until performance requires something more specialized. It also has some very well thought out mechanisms for concurrency.

I would think these things could be implemented in Racket. In fact, I would love to see a Racket implementation of Clojure (minus the JVM bits) that was nicely integrated into DrRacket.

...because people tend to google who else is using it "in production" before using a language? I looked at it and despite all cool language and cool tooling, I wouldn't choose it over CL or Clojure. But isn't Racket the most popular of the Scheme dialects?

And, as a joke, their name is really bad for business, though I happen to like it just because of this ;) : For "racket", in my part of the world, the first meaning of the word that pops to one's mind is the one related to violent organized crime. Even in the English dictionary there's the "2. A dishonest business or practice, especially one that obtains money through fraud or extortion." meaning (http://www.thefreedictionary.com/racket)

I think the name is a pun on Scheme: pyramid scheme, Ponzi scheme, ..., Racket. It's intentionally meant to evoke that image of fraud.

http://en.wikipedia.org/wiki/Pyramid_scheme http://en.wikipedia.org/wiki/Ponzi_scheme

GNU Guile.

And Larceny http://www.larcenists.org/

Scheme was/is seen as a teaching tool/research platform while CL was/is the language to actually make something in. Doubly so with PLTScheme being marketed as a teaching tool, just never got the momentum.

The re-brand wasn't a bad idea in that respect but they probably should have held off until they had some shiney new showcase app/library to (ugh) synergize with (see Pharo piggy backing with Seaside).

"Scheme was/is seen as a teaching tool/research platform while CL was/is the language to actually make something in."

Scheme is great for both, and it's very unfortunate that the impression most people have of it is that it's an educational/reasearch tool only.

I think it depends how you measure popularity. For example at my work Racket is the most popular Lisp: I use it and no one else ever used or is going to use any other Lisp... :)

Racket is a very young language. Until a few (3?) years ago it was a (PLT) Scheme; Scheme has it's uses but they are very specific, nobody would use a Scheme as a general purpose, systems programming language. Racket evolved and is now much more than simply a Scheme-with-extensions: it's a platform for language creation as well as robust runtime and has many very good tools.

It all makes sense if you look at Racket as a fresh language, that only started a few years ago - think about Python in it's infancy. It is very impressive and surprisingly mature in comparison, but it's still very young and has many parts missing - it's almost ready for production, but not quite there yet and I don't think this should be rushed.

Sticking with Racket now is a kind of gamble. I think that in a six to eight years time Racket will be competing with Clojure for a place of the most popular Lisp in existence and I think that in the future it will be the "new hot thing" for a while, just as Python or Ruby or JavaScript were. Most importantly Racket is a language that I'd really like to use in ten years. But still, it's uncertain and any number of things can go wrong and Racket can remain unpopular, niche product for eternity. That's the risk of using young, aspiring languages - I once won betting on Python and now I'm going to try again with Racket.

So, to summarize this rant: Racket is very young; languages need time to gain traction; if Racket evolves as fast as it does it will become insanely popular when the time comes.

I would like to see alternate lisps grow up, like racket, joxa... :)

Technically, Racket is a Scheme dialect not a LISP one.

a Scheme dialect not a LISP one

I feel a bit confused about the distinction you're making. Didn't Guy Steele refer to Scheme as a LISP dialect?

All scheme dialects are lisps, but Racket and the various scheme implementations are Lisp-1's, with the same namespace for functions and variables, while Common LISP is a lisp-2, which have separate function and variable namespaces.

See http://en.wikipedia.org/wiki/Lisp-1_vs._Lisp-2#The_function_...

I was using the word "Lisp" to mean "all languages with s-expr based syntax" and not as an abbreviation of Common Lisp. I even explicitly noted that: "Until a few (3?) years ago it was a ... Scheme".

But I don't know the Lisp/lisps history that well, did I use the term incorrectly?

Actually... Racket is a Scheme dialect which is a language in the Lisp family.

Much like Clojure, Kernel, and newLISP are languages in the Lisp family.

That's somewhat debatable - I mean whether Racket is a dialect of Scheme. There's a page here: http://racket-lang.org/new-name.html that states that Racket is a descendant of Scheme, but not that it's a dialect. Also here: http://stackoverflow.com/a/3358638/280995 someone says that Racket is not a Scheme (I think therefore not a dialect of Scheme either) because it's not backwards compatible and doesn't fully implement any of the Scheme standards. On the other hand Racket includes a scheme language; but it's merely one of many possibilities and with default installation comes also algol60 implementation, yet obviously we don't say that Racket is a dialect of Algol.

Anyway, I think that day-to-day programming in Racket is sufficiently different from Scheme's that it (Racket) deserves to stand on it's own. Saying it's a descendant of Scheme is obviously ok, but saying it's a dialect of Scheme is a bit too much, I think.

Is my mind wired completely different from the speaker's, or are all his analogies really that bad? I just couldn't bear it after 15 min, my mind hurt after all the magic wells / documents / princesses / programs - they pushed my intuitions in the wrong directions and then made me feel like I have to fight my own thoughts to turn my mind to what he was explaining. He gets to the idea of Racket as a platform for DSLs in such a convoluted way... I'm really sorry for his students, at least if they have my "kind of mind"!

On the side, I find Rich Hickey's and other Clojure videos very inspiring and atuned to my "intuitions" (that's how you sell a Lisp!) and same for the SICP lectures from the 80's with their "processes as spirits inside the computer" metaphors.

Can you think of a different three level analogy?

Well, my first shot while keeping it in the myth and magic realm:

level 0 (the "word processor" or manipulating data with not programmable tools): you are wizard's apprentice - and you clean your room by using the tools yourself (as real world brooms and scrubbers are not "programmable" :) )

level 1 (the "programming language" or "programmable tools" or code level): you are a wizard now (or an apprentice that stole the master's wand or magic hat, something) and you use your wand to instruct/program the tools to do the cleaning for you according to your rules/program

level 3 (either meta-programmer or dsl or language for building languages or the code-as-data level): you are now a master wizard that can control other people and even other wizards that can in turn control other wizards and so on... (you have a language for writing languages or, in fact, any language that can treat code as data because you can now use it to make other languages)

...if you're looking for some imagery for it, master Walt's cartoon http://www.youtube.com/watch?v=jSTWy25hRiI is pgood inspiration for level 1 (especially if you want to emphasize how "hard" programming is), but you'll have to look for yourself for level 0 and 3 images

Does that mean soegaard is a level 3 master wizard -- since he caused you to make this alternate analogy? ;)

...or basically anything else that transmits the ideas of processes, or things that happen and programs/instructions - not static things like wells and castles and dumb actors (princesses), that make me think more of OOP, or the totally "incomputable" for me idea of "granting a wish".

FYI, Matthew Flatt will be doing at keynote at Clojure/West on Racket in Portland, March 18-20th. http://clojurewest.org/schedule

Also worth a mentioning Racket's ability to go static vs dynamic vs hybrid typed. An example of Typed Racket: Iteratees http://goo.gl/9uT4R Compare and contract with Haskell and Scala impls of same.

Reply on news.ycombinator.com