Comment by CalChris
8 years ago
The LISP people should have changed those names way back when. Not 79, 59. McCarthy himself said then that this representation is independent of the IBM 704 computer, or of any other electronic computer [1]. So why stick with the names of 704 registers? Was there a purpose to this?
[1] https://courses.engr.illinois.edu/cs421/sp2012/project/mccar...
Lisp has both CAR/CDR and FIRST/REST.
It's usual programming style to use CAR/CDR when working with CONS trees/graphs and FIRST/REST when working with lists.
OK, but isn't CalChris's point still valid? Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959.
Do I think it's a big problem? Not really. Are the names less than optimal? Yes.
These names were very well thought out. They were part of a larger vocabulary in fact; most of that vocabulary died out and only car and cdr remained. There were things like cwr, ctr and others. MacCarthy and gang were well aware (of all people!) that the language is machine-independent and not something that will be forever stuck to the particular IBM machine; they used these names anyway because there was nothing wrong with them.
By the way, the C and R are not machine specific; any machine can have a "register" and any register can have "contents". On any machine where we implement cons cells, we can just call the two fields registers, understood as being A and D.
These, in turn, can be "ante" and "de", if someone desperately needs mnemonics:
https://en.wiktionary.org/wiki/ante
https://www.etymonline.com/word/de-
"active word-forming element in English and in many words inherited from French and Latin, from Latin de "down, down from, from, off; concerning."
2 replies →
> Something other than CAR/CDR would have been a better choice for working with CONS cells, even back in 1959.
> Are the names less than optimal? Yes.
I don't think that's self-evident. The only better alternatives I've ever heard suggested are something like left and right, which I think are probably slightly clearer names for what the functions do, but not significantly so (I don't think "a cons is a thing with a left and right side" is that much more intuitive than "a cons is a thing with two parts called a car and a cdr").
Further, I think in 1959 when most everyone using Lisp was hacking on the implementation itself, names that are mnemonic for what the machine they were actually using is actually doing make sense; I don't think they're clearly sub-optimal now, but I think they were even less so in 1959.
2 replies →
> So why stick with the names of 704 registers? Was there a purpose to this?
Are there other, much better names for the two parts of a pair? First and second aren't great because they're not clear that it's a pair and not a larger sequence (and the difference is important in Lisp; since lists are built on conses, it would be odd to get the tail of a list by calling 'second'). The only other names I can think of would be something like left and right, but I don't think those would be substantially easier to understand for beginners than car and cdr.
When working with lists, many people prefer to use the functions first and rest, which behave identically to car and cdr but are more meaningfully named for list applications. They would be terrible names for the general cons-handling functions, though.
Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing.
I don’t mind car and cdr though. I kind of wish they were in Clojure tbh. It feels nice to pay tribute to a half-century of computing history.
> Left and Right are really commonly used in this situation in ML-descended languages. It’s not the worst thing.
Are you talking about Either types? OCaml and Haskell at least both call car and cdr fst and snd respectively, but they're rarely used.
As for left and right as names, I don't think they're any worse than car and cdr, I just don't think they're substantially better.
Would it be silly to use 'head' and 'tail' then?
I learned car/cdr in school so that's what I use.
I use FST and RST (for FIRST and REST). I like them because they are short, mnemonic, and can be composed like CAR and CDR, e.g. CADADDR = FRFRRST (though I also believe that if you find yourself doing this you're almost certainly doing something wrong).
In C++, I often use LHS and RHS (left hand side / right hand side). I know other people who like fst, eat (first, rest). Both seem better to me, as they have some meaning other than random made up strings (for people who don't use 704).
> Both seem better to me, as they have some meaning other than random made up strings (for people who don't use 704).
Calling the lefthand child of a node in a binary tree the "first" and the righthand child the "rest" seems like a markedly worse naming system than random made-up names, to me, because the use of English names seems to be implying meaning that it doesn't really offer.
The idea of a thing with two values, one on the left side and one on the right side is probably a bit more intuitive, but I really doubt it's that much moreso than just saying it has two parts called the car and the cdr. People who've never touched a 704 have been learning about conses as a data structure with two parts, a car and a cdr, for almost 60 years; to most of them, it's just vocabulary you learn now, the same way that numerator and denominator are just arbitrary names to people learning fractions.
I also think it's interesting how much people complain about car and cdr yet it's very rare that people complain about cons, which is just as badly named. It should probably have been called make-pair or something, which I think is a much bigger improvement than car vs lhs. But really, it's 3 words of vocabulary, whose definitions can be completely understood by
That's not very much to ask someone to wrap their head around, it was meaningful to the original implementors, and for people now there's about 60 years of precedent for using that terminology.
5 replies →