(This is a snapshot of my old weblog. New posts and selected republished essays can be found at raganwald.com.)
Wednesday, July 19, 2006
car and cdr have had their fifteen minutes of fame
In response to Six Things I Dislike About Ruby (and Four I Don't), someone quipped:
car and cdr are historical anachronisms. They are like secret handshakes: their only remaining value is to help a small cadre of enthusiasts bond around a shared collection of obscure trivia.
I wouldn't be suprised if Lispers have droppped these from production systems in favour of the logo-isms first and rest. But then again, I wouldn't be surprised if some Lispers hang onto them either. After all, if every textbook uses them as examples, they become familiar through repetition, even if their names no longer make any sense whatsoever.
The last time I checked nobody was running Lisp on IBM 704s any more. These names really deserve a quiet retirement in a shady home where they can share war stories with puts while they listen to goto complain about what happened when call/cc made him obsolete.
All that being said, car and cdr were rather prolific parents: Lisp implementations usually support a special syntactic sugar called composition where you can make up a function with the following form: (/c[ad][ad]*r/ ...) and this is executed as if you types a series of nested cars and cdrs. For example, (caddr foo) has the same effect as (car (cdr (cdr foo))).
I really like composition. We can sort of get that with dot notation and messages in languages like Ruby. But I secretly admire the way the German language lets you compose entirely new nouns by sticking smaller nouns together.
I think car and cdr's composition has a spiritual descendant in Ruby's metaprogramming style. For example, dynamic finders in Rails. Person.find_all_by_username_and_password really feels like caddaddr. Uh, maybe not.
But it's nice to see that some of the good ideas from Lisp thrive in new homes.
In response to Six Things I Dislike About Ruby (and Four I Don't), someone quipped:
while we are stamping out those silly ungrammatical C++-isms like "puts", can we also get rid of those silly lispisms of 'car' and 'cdr'?Yeah, sister (or brother, as the case may be)!
car and cdr are historical anachronisms. They are like secret handshakes: their only remaining value is to help a small cadre of enthusiasts bond around a shared collection of obscure trivia.
I wouldn't be suprised if Lispers have droppped these from production systems in favour of the logo-isms first and rest. But then again, I wouldn't be surprised if some Lispers hang onto them either. After all, if every textbook uses them as examples, they become familiar through repetition, even if their names no longer make any sense whatsoever.
The last time I checked nobody was running Lisp on IBM 704s any more. These names really deserve a quiet retirement in a shady home where they can share war stories with puts while they listen to goto complain about what happened when call/cc made him obsolete.
All that being said, car and cdr were rather prolific parents: Lisp implementations usually support a special syntactic sugar called composition where you can make up a function with the following form: (/c[ad][ad]*r/ ...) and this is executed as if you types a series of nested cars and cdrs. For example, (caddr foo) has the same effect as (car (cdr (cdr foo))).
I really like composition. We can sort of get that with dot notation and messages in languages like Ruby. But I secretly admire the way the German language lets you compose entirely new nouns by sticking smaller nouns together.
I think car and cdr's composition has a spiritual descendant in Ruby's metaprogramming style. For example, dynamic finders in Rails. Person.find_all_by_username_and_password really feels like caddaddr. Uh, maybe not.
But it's nice to see that some of the good ideas from Lisp thrive in new homes.
Labels: ruby
Comments on “car and cdr have had their fifteen minutes of fame”:
<< Home
that picture is totally goofed
Shrug.
I use Firefox on OS X and I just tried it on Windows XP.
It is hosted on flickr.com... Perhaps there is an issue there when you tried to view it?
How many programmers does it take to change a lightbulb?
The light works in development...
Shrug.
I use Firefox on OS X and I just tried it on Windows XP.
It is hosted on flickr.com... Perhaps there is an issue there when you tried to view it?
How many programmers does it take to change a lightbulb?
The light works in development...
I'll go along with getting rid of car and cdr in Lisp if we can agree to get rid of C++ and Perl at the same time.
# posted by 
: 9:04 PM
: 9:04 PM
first and rest make perfect sense when you are working with lists, but cons cells can be used to build many other data structures. They are basic building blocks and as such they need an abstract interface, one that does not restrict your thinking to any particular application. In that context car and cdr work much better that first and rest.
# posted by : 7:01 PM
: 7:01 PM
cons cells can be used to build many other data structures
Sure. But look at any non-Lisp discussion of data structures. You won't see the words car or cdr, would you?
What's wrong with left and right? For that matter, why is a cell containing two pointers magic? Why isn't a cons cell simply a special case of an array?
I think cons cells are only magical on architectures where you can squeeze them both into a single machine word, or on the 704 where they were in separate registers but there was special efficiency loading them both at the same time. And even there, why let the implementation leak into the abstraction?
Given how clever programmers are, why not build a language where everything is an array, but the implementation special cases arrays with two elements into cons cells?
Isn't that what compilers are supposed to do?
Sure. But look at any non-Lisp discussion of data structures. You won't see the words car or cdr, would you?
What's wrong with left and right? For that matter, why is a cell containing two pointers magic? Why isn't a cons cell simply a special case of an array?
I think cons cells are only magical on architectures where you can squeeze them both into a single machine word, or on the 704 where they were in separate registers but there was special efficiency loading them both at the same time. And even there, why let the implementation leak into the abstraction?
Given how clever programmers are, why not build a language where everything is an array, but the implementation special cases arrays with two elements into cons cells?
Isn't that what compilers are supposed to do?
What's wrong with left and right?
They differ in length, which may preclude elegant formatting of code in some cases. And they are longer, and for something that elementary shorter is better.
And one couldn't write caadr.
For fundamental language contructs, being convenient to use is more important than being easy to remember. In natural languages verbs like to be or to have tend to be irregular.
lt and rt could be great, one could even write llrt, but I'm not sure if it would be such a big win.
Given how clever programmers are, why not build a language where everything is an array, but the implementation special cases arrays with two elements into cons cells?
Of course! I wholeheartedly agree. But that would hardly be Lisp anymore, would it?
They differ in length, which may preclude elegant formatting of code in some cases. And they are longer, and for something that elementary shorter is better.
And one couldn't write caadr.
For fundamental language contructs, being convenient to use is more important than being easy to remember. In natural languages verbs like to be or to have tend to be irregular.
lt and rt could be great, one could even write llrt, but I'm not sure if it would be such a big win.
Given how clever programmers are, why not build a language where everything is an array, but the implementation special cases arrays with two elements into cons cells?
Of course! I wholeheartedly agree. But that would hardly be Lisp anymore, would it?
# posted by : 4:05 AM
: 4:05 AM
I like caadr, that's a nice feature.
The next time I'm myself writing some tree navigating methods, I promise to over-reduce things a follows:
.l and .r will be aliases for left and right. Furthermore, any message consisting entirely of ls and rs (such as llr) shall be treated recursively by applying the first letter and passing the rest of the letters as a message to the result.
The next time I'm myself writing some tree navigating methods, I promise to over-reduce things a follows:
.l and .r will be aliases for left and right. Furthermore, any message consisting entirely of ls and rs (such as llr) shall be treated recursively by applying the first letter and passing the rest of the letters as a message to the result.
Aah. Here's why car and cdr are better than first and rest:
You can do something like caaaddr and everyone know what that means. Try doing that with first and rest!
You can do something like caaaddr and everyone know what that means. Try doing that with first and rest!
<< Home
