10 ms·

I know Haskell pretty well, having spent a cozy two or three years with it as my main squeeze for hobbyist programming. Now I'm doing more Prolog. I like stuff

I know Haskell pretty well, having spent a cozy two or three years with it as my main squeeze for hobbyist programming. Now I'm doing more Prolog. I like stuff that makes me think differently. Anyhow...is there something I'm missing from Lisp? I've never Lisp'd, but people talk like it turns you into Neo in the matrix.

rtpg 10y ago
I think Lisp is kind of magical if you buy into the iterative nature of computing. Sure, it's functional in the sense that functions are cheap and easy to declare/use as values, but if you look at how things like macros end up working, it ends up being about having control of your runtime. If you're into Haskell/Prolog, you probably are much more on the mathematical side of things... honestly I can't think of much you can express easily in Lisp (CL for example) that you cannot transform into an almost equivalent solution in Haskell. It's good to try out a Lisp (and get up to macros or something), but my personal feeling is that Lisp is probably more of an advancement for people in the dynamic language universe (JS/Ruby) than it is for people in the ML-y languages.
ACow_Adonis 10y ago
The "Neo in the matrix" thing, i think, is the "lisp enlightenment experience" under another name. Trying not to be some smug lisp weenie about it, I'm going to try to put my own experience into normal english. It happened for me somewhere around the 6-12 month mark of daily lisp study/coding. You usually start out learning lisp like other languages: you're just trying to learn the syntax of various commands and the quirks of how everything fits together of the concepts (if you've learnt another language) that you probably already know. And a big part of this is laying the foundation of a deep understanding of linked lists, cons cells, cars and cdrs. And after that, you start learning about the identation style available to you as well. And right now, its just another language. Then one day, you wake up, and you look at code, and some switch you weren't even aware of has flicked in your brain, and it all looks different. Mature lispers read/structure code subconsciously using both symbols and the indentation style to convey real information about the structure of a program, which is why you can really mess with their heads if you mess yours up too much. But there's a difference between being told you can do it this way, and the practice that actually makes you do it, because its not a conscious thing. Anyway, the reason lispers "think they're neo in the matrix" is because the mental effect this has on you is not unlike that scene towards the end of the matrix where Neo looks down the hallway and his perception of the thing he's been interacting with this entire time just fundamentally changes. Its no longer writing code, that is to say, strings of text. Its interacting with and perceiving the very structure of the programs themselves. And it is a real psychological phenomenon. It makes you really happy, and its why lispers talk about it. It makes you want to go and shake your neighbour and be all like "Oh my god, I get it, do you see this!"...and of course they don't. They see scribbles on the screen. The funny thing is, they always say they're Neo in the matrix, which i understand is cooler, partly because that final scene captures the phenomenon of what it feels like, but its arguably summed up in a more lackadaisical way by the character Cypher in one of his quotes from the same movie: "You get used to it, though. Your brain does the translating. I don't even see the code. All I see is blonde, brunette, redhead. Hey uh, you want a drink?"
- MichaelGG 10y ago
  But what are the practical implications? What things can lisps do that nothing else can? Macros seem to be the biggest thing mentioned; pg says that was what let him write features so much faster than competitors. On the next hand he says he can't imagine not having different types of items in a list. And like Cypher, if you reach some higher enlightenment, then can't use it (practical constraints) it just makes you want to go back. E.g. a friend that recently picked up F# now laments so much about having to work in C#; it's now painful. Whereas a couple months ago, C# was pretty nice. 1: Though I'd like to see which features really made much of a difference.
  - ACow_Adonis 10y ago
    Warning: hairy high falutin' answer... The "what can lisp do that other languages can't" is, I think, a bit of a red herring...or rather, depends on what problems you face and in what context we're discussing the issue. The short answer is of course: nothing. And if most of the problems you face are defined by by working on a particular technology (database tables/javascript), in a particular social concept (say a corporate workplace), or a problem that isn't LISP based, then the answer is probably "REALLY NOTHING...and possibly less". But if your problem doesn't currently have a technology acceptably applied to it, or you think the current offerings fail in some way, or you're finding the load of trying to smoosh things together because in your head the expression is crystal clear, but your language can't express it easily... The longer answer: Apart from a genuinely "linguistic" analysis of the language itself, I think "a good computer language" is one that minimises, in the abstract, some distance function between the programmer's mental models, the machine's mechanical implementation, the problem's domain, and the aesthetic simplification of the source code. Now, no language can minimise that equation in the absolute for all possible problems, if not because several of those things are often fixed absolutely or practically. And if you're working in a domain that you feel has already gotten it pretty damn good/close. Some languages trade-offs of maximizing certain aspects to the detriment of others, but it doesn't matter because it fits their "biological niche" as it were. So what is LISP (i'm talking Common Lisp here in my own personal experience) biological niche, or what does it offer that others don't? I think my answer is that its a fantastic flexible generalist that allows you to minimise the aesthetics/complexity of your code across multiple domains. - It lets you get pretty low level like C, but gets you a cheap move to a high level language like python to trade off convenience against efficiency. You get high level, like python/perl, but can drop down to far more performant code while minimising the cognitive load by staying in the same language. - SBCL, at least, mixes some of the benefits of strongly typed and highly dynamic languages (static compiler checks and fully dynamic code in the same place) - Macros allow computation and expression on the language itself more easily than in other languages, allowing you to move your language closer to your problem scope and to the programmers mind and aesthetic expression of source code. - Deals quite nicely with functions, objects and imperative programming all in the one place, allowing you to move the language closer to the programmer's mind/problem scope/aesthetic expression of source code. - And there's the whole dynamic interaction with your program, rather than an explicit code/compile/run cycle. So in short, use it where you need to care about those things seriously, which in my land is really: - Cross domain problems where the clash between the two tools/paradigms are too great - Problems where you disagree with some of the trade-offs made with the tools/languages that have come to dominate a specific area - Problems where there isn't a dominant/well-fitting tool currently (of course, for those problems, community/social/library issues isn't really an issue because there isn't one/aren't any...) - Problems where you honestly don't know what the solution is going to look like yet and no one, to your knowledge, has gotten there before you. I generally say that LISP is the second-best solution to a lot of solved problems, but perhaps the best solution to multiple-solved problems and unsolved problems for now. And that's what I think LISP gets you...the shortest distance between disparate optimisations. Now...I personally accept that 95% of programming these days ISN'T that...
lostcolony 10y ago
I've only played with Lisps, so can't quite say I've had the 'enlightenment' experience, but, realizing that "I know how to manipulate lists...and my program is just another list (of lists of lists of lists)" was a bit eye opening. The regularity of the language makes it so that taking advantage of this is far more reasonable (both a logical thing to do, and something that can be reasoned about) than any other language I've seen (where the closest equivalent would be the eye gougingly bad decision to read in your code as a string and perform string operations to munge it and then 'eval' it. Or, arguably, since this is obviously metaprogramming, reflection, but there it still ends up being something 'special', something very different than just writing a program, and as such is still frowned upon. Lisp is homoiconic, and realizing the ramifications of that made metaprogramming go from a "avoid if at all possible" to just another tool, one that sometimes provides the most elegant solutions, and is reasonable to choose in those situations)
sago 10y ago
Metaprogramming in Lisp is easier than in other languages I've used (though I tend to use Scheme, not CL). Grokking apply/eval and the code-is-data nature of the lambda calculus is the key enlightenment. I stopped thinking of metaprogramming as some weird esoteric edge-case and started using it to solve problems. I mean I write tools to help me write code in all languages, to some extent, but in Scheme it is ubiquitous. I went to Haskell from Lisps, and found quite a lot of similarity there, so if you're fluent Haskell, you might not find anything particularly revelatory. Though Haskell's more strict nature makes it much harder, in my experience, so I do it less often. Eval/apply is a different model of code than imperative languages, which can be 'mindblowing' if you've only thought of languages as higher level assembly. I do love Scheme, but I struggle to justify using it in practice. Which is telling I think.
sklogic 10y ago
And in a lisp you'd be able to mix Haskell and Prolog together, with dozens of other languages as well, seamlessly and efficiently (unlike the non-destructive Prologs people tend to embed in Haskell).

rtpg 10y ago

I think Lisp is kind of magical if you buy into the iterative nature of computing. Sure, it's functional in the sense that functions are cheap and easy to declare/use as values, but if you look at how things like macros end up working, it ends up being about having control of your runtime. If you're into Haskell/Prolog, you probably are much more on the mathematical side of things... honestly I can't think of much you can express easily in Lisp (CL for example) that you cannot transform into an almost equivalent solution in Haskell. It's good to try out a Lisp (and get up to macros or something), but my personal feeling is that Lisp is probably more of an advancement for people in the dynamic language universe (JS/Ruby) than it is for people in the ML-y languages.

ACow_Adonis 10y ago

The "Neo in the matrix" thing, i think, is the "lisp enlightenment experience" under another name. Trying not to be some smug lisp weenie about it, I'm going to try to put my own experience into normal english. It happened for me somewhere around the 6-12 month mark of daily lisp study/coding. You usually start out learning lisp like other languages: you're just trying to learn the syntax of various commands and the quirks of how everything fits together of the concepts (if you've learnt another language) that you probably already know. And a big part of this is laying the foundation of a deep understanding of linked lists, cons cells, cars and cdrs. And after that, you start learning about the identation style available to you as well. And right now, its just another language. Then one day, you wake up, and you look at code, and some switch you weren't even aware of has flicked in your brain, and it all looks different. Mature lispers read/structure code subconsciously using both symbols and the indentation style to convey real information about the structure of a program, which is why you can really mess with their heads if you mess yours up too much. But there's a difference between being told you can do it this way, and the practice that actually makes you do it, because its not a conscious thing. Anyway, the reason lispers "think they're neo in the matrix" is because the mental effect this has on you is not unlike that scene towards the end of the matrix where Neo looks down the hallway and his perception of the thing he's been interacting with this entire time just fundamentally changes. Its no longer writing code, that is to say, strings of text. Its interacting with and perceiving the very structure of the programs themselves. And it is a real psychological phenomenon. It makes you really happy, and its why lispers talk about it. It makes you want to go and shake your neighbour and be all like "Oh my god, I get it, do you see this!"...and of course they don't. They see scribbles on the screen. The funny thing is, they always say they're Neo in the matrix, which i understand is cooler, partly because that final scene captures the phenomenon of what it feels like, but its arguably summed up in a more lackadaisical way by the character Cypher in one of his quotes from the same movie: "You get used to it, though. Your brain does the translating. I don't even see the code. All I see is blonde, brunette, redhead. Hey uh, you want a drink?"

MichaelGG 10y ago
But what are the practical implications? What things can lisps do that nothing else can? Macros seem to be the biggest thing mentioned; pg says that was what let him write features so much faster than competitors. On the next hand he says he can't imagine not having different types of items in a list. And like Cypher, if you reach some higher enlightenment, then can't use it (practical constraints) it just makes you want to go back. E.g. a friend that recently picked up F# now laments so much about having to work in C#; it's now painful. Whereas a couple months ago, C# was pretty nice. 1: Though I'd like to see which features really made much of a difference.
- ACow_Adonis 10y ago
  Warning: hairy high falutin' answer... The "what can lisp do that other languages can't" is, I think, a bit of a red herring...or rather, depends on what problems you face and in what context we're discussing the issue. The short answer is of course: nothing. And if most of the problems you face are defined by by working on a particular technology (database tables/javascript), in a particular social concept (say a corporate workplace), or a problem that isn't LISP based, then the answer is probably "REALLY NOTHING...and possibly less". But if your problem doesn't currently have a technology acceptably applied to it, or you think the current offerings fail in some way, or you're finding the load of trying to smoosh things together because in your head the expression is crystal clear, but your language can't express it easily... The longer answer: Apart from a genuinely "linguistic" analysis of the language itself, I think "a good computer language" is one that minimises, in the abstract, some distance function between the programmer's mental models, the machine's mechanical implementation, the problem's domain, and the aesthetic simplification of the source code. Now, no language can minimise that equation in the absolute for all possible problems, if not because several of those things are often fixed absolutely or practically. And if you're working in a domain that you feel has already gotten it pretty damn good/close. Some languages trade-offs of maximizing certain aspects to the detriment of others, but it doesn't matter because it fits their "biological niche" as it were. So what is LISP (i'm talking Common Lisp here in my own personal experience) biological niche, or what does it offer that others don't? I think my answer is that its a fantastic flexible generalist that allows you to minimise the aesthetics/complexity of your code across multiple domains. - It lets you get pretty low level like C, but gets you a cheap move to a high level language like python to trade off convenience against efficiency. You get high level, like python/perl, but can drop down to far more performant code while minimising the cognitive load by staying in the same language. - SBCL, at least, mixes some of the benefits of strongly typed and highly dynamic languages (static compiler checks and fully dynamic code in the same place) - Macros allow computation and expression on the language itself more easily than in other languages, allowing you to move your language closer to your problem scope and to the programmers mind and aesthetic expression of source code. - Deals quite nicely with functions, objects and imperative programming all in the one place, allowing you to move the language closer to the programmer's mind/problem scope/aesthetic expression of source code. - And there's the whole dynamic interaction with your program, rather than an explicit code/compile/run cycle. So in short, use it where you need to care about those things seriously, which in my land is really: - Cross domain problems where the clash between the two tools/paradigms are too great - Problems where you disagree with some of the trade-offs made with the tools/languages that have come to dominate a specific area - Problems where there isn't a dominant/well-fitting tool currently (of course, for those problems, community/social/library issues isn't really an issue because there isn't one/aren't any...) - Problems where you honestly don't know what the solution is going to look like yet and no one, to your knowledge, has gotten there before you. I generally say that LISP is the second-best solution to a lot of solved problems, but perhaps the best solution to multiple-solved problems and unsolved problems for now. And that's what I think LISP gets you...the shortest distance between disparate optimisations. Now...I personally accept that 95% of programming these days ISN'T that...

lostcolony 10y ago

I've only played with Lisps, so can't quite say I've had the 'enlightenment' experience, but, realizing that "I know how to manipulate lists...and my program is just another list (of lists of lists of lists)" was a bit eye opening. The regularity of the language makes it so that taking advantage of this is far more reasonable (both a logical thing to do, and something that can be reasoned about) than any other language I've seen (where the closest equivalent would be the eye gougingly bad decision to read in your code as a string and perform string operations to munge it and then 'eval' it. Or, arguably, since this is obviously metaprogramming, reflection, but there it still ends up being something 'special', something very different than just writing a program, and as such is still frowned upon. Lisp is homoiconic, and realizing the ramifications of that made metaprogramming go from a "avoid if at all possible" to just another tool, one that sometimes provides the most elegant solutions, and is reasonable to choose in those situations)

sago 10y ago

Metaprogramming in Lisp is easier than in other languages I've used (though I tend to use Scheme, not CL). Grokking apply/eval and the code-is-data nature of the lambda calculus is the key enlightenment. I stopped thinking of metaprogramming as some weird esoteric edge-case and started using it to solve problems. I mean I write tools to help me write code in all languages, to some extent, but in Scheme it is ubiquitous. I went to Haskell from Lisps, and found quite a lot of similarity there, so if you're fluent Haskell, you might not find anything particularly revelatory. Though Haskell's more strict nature makes it much harder, in my experience, so I do it less often. Eval/apply is a different model of code than imperative languages, which can be 'mindblowing' if you've only thought of languages as higher level assembly. I do love Scheme, but I struggle to justify using it in practice. Which is telling I think.

sklogic 10y ago

And in a lisp you'd be able to mix Haskell and Prolog together, with dozens of other languages as well, seamlessly and efficiently (unlike the non-destructive Prologs people tend to embed in Haskell).