[ANTSANTS]

My question is, given the cons pair is not a meaningful primitive datatype to modern CPUs (can't fit in a register unless you're using 32 bit atoms on a 64 bit CPU, the individual cells cannot be meaningfully manipulated while packed into a single register even in that case...), is there a good reason to use an unrestricted pair/2-tuple/2 element vector for the sexp representation? Would it be "wrong" to remove dotted pair notation and make sexps a restricted type whose cdr could only be another pair or the empty list? You'd have to give up alists, but those would probably be better served by doing as Clojure does and introducing a hash table syntax.

Other than alists, simplifying the implementation[1], and tradition, I can't think of a good reason to keep improper lists and dotted pair notation around. They aren't really that useful, confuse newcomers, and encourage the use of inefficient data structures. I haven't written much Lisp, but when I see a dotted pair (usually in the context of an alist or some hideous approximation of a struct[2]), it comes across as a "code smell" to me, like "this person is taking SICP too literally."

[1] You pay for extra reader syntax, but gain uniformity in your treatment of the car and cdr cells. On the other hand, maybe you could optimize the cdr cell a bit if it only needed to address pairs and not arbitrary atoms.

[2]

  ;; barfage
  (define (make-thing a b c) (cons a (cons b c))) ; barf
  (define (thing-field-a x) (car x))
  (define (thing-field-b x) (cadr x))
  (define (thing-field-c x) (cddr x)) ; more barf

[brudgers]

Going a little Abelson&Sussman, the utility of cons persists because it is a powerful abstraction. Fitting into registers was, as a scholastic might say, an as accidental feature as the input line voltage of a toaster [110 @ 60Hz makes things easier in this part of the world but isn't intrinsic to making crisp English muffins].

cons is useful for dealing with linked memory structures, including singlely linked lists and particularly singly linked lists where the stored value in each list element is a pointer because it abstracts over all the pointers. Writing code to process and manipulate and pass pointers is idiomatic in C. Using pointers is idiomatic Lisp.

Your barfing code shows why cdr and car persist. They are also powerful abstractions due to their rules of combination and probably because their rules of combination aren't constrained by a bias toward English language. There ain't no synonyms for "caddr" and "cdddr" in English or Python.

The idea that lists should be everywhere replaced by structs is assembly in any language thinking (also known as C). Lists provide a standard interface. Each struct definition creates a new data type.

It is better to have 100 functions operate on one data structure than 10 functions on 10 data structures. Alan J. Perlis, Epigram 9.

[chc]

I think the question is more along the lines of "Should the core data type be the cons cell rather than the list?" Clojure takes the approach the GP was talking about. It has lists with heads and tails, but it does not have dotted pairs. The tail of a list is always a list.

[munificent]

> There ain't no synonyms for "caddr" and "cdddr" in English or Python.

In Python:

    x[1][1][0] # caddr
    x[1][1][1] # cdddr

[anaphor]

It has pedagogical value to define things in terms of a very small kernel language. You will find things like alists in fairly advanced books on lisp (e.g. Lisp In Small Pieces) simply because it reduces the number of extra concepts you need to explain. It's assumed that the reader can figure out how to make the code cleaner and more idiomatic as long as they understand the ideas being presented.

Also I have made use of improper lists in my code, usually when I need to do something with multiple values and I don't want to mess with multiple return values, or if I just want to create a list of tuples for some reason and don't feel like defining an actual struct (can't think of a good example right now...)

[ANTSANTS]

>Also I have made use of improper lists in my code, usually when I need to do something with multiple values and I don't want to mess with multiple return values, or if I just want to create a list of tuples for some reason and don't feel like defining an actual struct (can't think of a good example right now...)

I get what you're saying, but isn't this just a shortcoming of the language? It's pretty common to want to write something like

  (define (string-lookup-that-might-fail)
    (if (random) <"string you were trying to look up" nil> <nil "It wasn't there.">))
  (define <some-string error-msg> (string-lookup-that-might-fail))

without wanting to destructure a list or use the call-values hack. That's a bad example, just pretend there were multiple reasons the lookup could fail.

I totally agree about there being pedagogical value in having a simple core. I'm torn between appreciating minimalism and believing that it's ok to introduce complexity to a language that people will use to get things done every day, so long as it's useful complexity and not just historical accident.

[anaphor]

Okay yeah I agree with you there. The pain of destructuring cons cells is somewhat mitigated by the various forms that allow pattern matching, but it's still basically the same.

So I think that is actually a good argument for introducing option types (like Haskell's Either a b) into the language. I think Typed Racket has something like this, but in either case you'd want some nice syntax to handle it as well.

Hash tables make this sort of thing fairly easy though (at least in Racket)

(define h #hash{["a" . 3] ["b" . 5]})

(hash-ref h k "failed")

I think most lisps support this right? I remember seeing something similar in pg's bayesian spam filter code.

[ANTSANTS]

> So I think that is actually a good argument for introducing option types (like Haskell's Either a b) into the language. I think Typed Racket has something like this, but in either case you'd want some nice syntax to handle it as well.

I need to look into Typed Racket more. Ever since I learned the tiny bit of Haskell that I know, I can't help but think that I need ADTs in my parenthesis.

Right at the top of the docs, though, it says that they designed the Typed Racket type system to provide static typing for existing untyped Racket programs. That makes me wonder what a statically typed lisp that wasn't trying to be backwards compatible would look like. I've heard of Shen and Qi but at first glance they (especially Qi) seemed to be drifting too far from Lispy prefix notation goodness for my taste.

> Hash tables make this sort of thing fairly easy though (at least in Racket)

It doesn't really help if you're trying to look up a string, though, because then you can't differentiate between a "good" string and an error string based on their types. That means it won't help for similar string fetching operations either, like making a network request and expecting either a response or a string indicating that the request timed out, the connection couldn't be made, etc.

According to the docs, you can also pass an error continuation to hash-ref instead of a failure string. That's interesting, though I have to admit I have trouble thinking in those terms.

[brudgers]

   #lang typed/racket

Isn't bound by backwards compatibility in the sense [I think] you are implying. What the docs are saying is that modules which enjoy static typing may be called transparently from modules which don't - though as is common with functions in the Math library there may be a performance penalty. The backwards compatibility is that the implementation of static typing does not impede dynamically typed code. Statically typed code works just fine with dynamic type checking.

+ regex-match-event from racket/port might be an alternative to handle network connections than hash table lookup. http://docs.racket-lang.org/reference/port-lib.html#%28def._...

[belovedeagle]

> It's pretty common to want to write something like [. . .] without wanting to destructure a list or use the call-values hack.

This makes no sense. You want feature X (the ability to return multiple values) without having to either (a) pick apart a simple structure containing those values or (b) assign names to those values, letting the runtime system pick them apart (or never stick them together, its choice) for you. So what more do you want?!

You want an option type, right? Well what's an option type? Let's consider Haskell:

data Maybe a = Nothing | Just a

No matter what kind of cleverness you do in memory or whatever, in the end this type is a tuple of two values: (constructor, data). The only special thing going on is that the type system makes sure that (a) constructor == Maybe or constructor == Just, and (b) if constructor == Maybe the data field is meaningless (has no type, can't be read, does not effect equality, whatever).

But it's still a tuple. The cons cell is the exact same thing in the absence of a strict compile-time type system, except for your code example we have it flipped: (data, constructor) (well, actually you seem to have a flipped 'Either String a' going on there, but let's stick with option for now) where data is actual data or nil when meaningless, and constructor is like Just when nil and like Nothing when true (btw, lisp convention is the other way around; second value is true like Just and false/nil like Nothing).

Just like the original linked article, what your complaint boils down to is "non-compile-time-typed systems allow you to do things which would break compile-time-typed systems"—and while I love compile-time typing as much as the next Haskell aficionado, I don't demand compile-time typing in a language [family] whose very essence is its absence.

EDIT: if you could reply and explain what would resolve this "shortcoming of the language", maybe I could understand better what you're trying to say...

[ANTSANTS]

> This makes no sense. You want feature X (the ability to return multiple values) without having to either (a) pick apart a simple structure containing those values or (b) assign names to those values, letting the runtime system pick them apart (or never stick them together, its choice) for you.

I was actually aiming for (b) here, just incorporated directly into the normal function call and return syntax. Why shouldn't you be able to return multiple values in registers or on a results stack without "tupling them up"? It's clearly supported by the hardware. By "call-values" (sp) hack, I meant that needing to use "call-with-values" and "values" is the hack. Lua, for instance, lets you (with very clean syntax) return multiple values from a function, and assign the results to separate variables, during which at no point is an intermediate table constructed:

  thing, err = getthing()
  if thing == nil then print(err) return nil end
  do_stuff(thing)

I don't think I got it quite right with my angular brackets, but I was trying to think of a nice, clear syntax for multiple return values that wouldn't look out of place in a Lisp.

Maybe I was being misleading by using an example that resembled option types. I wasn't trying to bring static typing into this, though I am interested in static type systems and curious about static Lisps, it's just that emulating simple option types for error handling purposes is a very common usage for multiple return values in dynamically typed languages (and, cough, static ones like Go that have lame type systems).

> data Maybe a = Nothing | Just a

> No matter what kind of cleverness you do in memory or whatever, in the end this type is a tuple of two values: (constructor, data).

Is that true? "Maybe" seems like the perfect candidate for a nullable pointer representation, or at least some kind of small tag. I'd be surprised if GHC represents Maybe as a full blown tuple.

[belovedeagle]

I still don't see how you're advocating a solution of anything other than call-with-values under a different name.

I don't know Racket itself, so let me switch to common lisp which has the same "problem"... except it comes with a super-simple macro to do exactly what you want, apparently:

(multiple-value-bind (quotient remainder) (floor x y) ;; some code referring to quotient and remainder )

which (excepting perhaps edge-cases I'm not considering, and definitely excepting some error-checking) is just

(defmacro mvb (vars form . body) `(multiple-value-call (lambda ,vars ,body) ,form)

The same thing should work in Racket, modulo syntax and names.

Asking a lisp programmer to do everything with special forms/primitive language elements is like asking a Haskell programmer to do everything in the IO monad. Sure, it's possible, it will work just fine; but... it's just wrong.

By the way, the existence of the #'values function by no means implies there's some data structure being created; indeed, that's the whole point of the #'values construct. Anything you could do with #'values you could do in a list which you then destructured; #'values is to be used when the common-case is to throw away the "structure" and just use each value individually or not at all, like with the return values of #'floor. So in fact #'values is exactly what you want; I'm not sure why it's a "hack" when lisp programmers write it under one name and "the solution" when you write it under a different name.

Ah, it occurs to me that maybe racket doesn't work like this, but in common lisp at least, the transiency of values can be seen in that when you say, e.g., (+ (floor a b) c), the remainder of a/b is thrown away; just the first value is used. That's another huge benefit over structures-to-be-destructured, but obviously doesn't work if the extra values are always important.

The point of my discussion with Maybe is that /even if/ it's just a nullable pointer, conceptually it's essentially a [restricted] tuple. (pointer-is-null,data). There's still two pieces of data there, even if the representation of one of them is cleverly folded into generic object representation or what-have-you. By definition it can't be done with a single piece of data: (Maybe a) is not the same type as (a), for all a. Even if (Maybe Word16), e.g., were represented in a single value at a single place in memory, that value would have to hold more than 16 bits. Then it's viewable as a... tuple, again; one element in 16 bits and the other in the remaining bits.

[Sean-Der]

I come from a background using Common Lisp for system and web development so I may see things differently than people who were introduced to it academically. Your code looks great, mine just needed to compile :)

The cons pair is a a really powerful primitive data structure, it is the linked list of Lisp. You can build a lot of powerful structures given this great base. I really don't it matters what architecture your running on.

Do they really confuse people new to Lisp? car+cdr is a pretty simple concept, when talking to new Lispers they usually don't complain about this.

Hash Tables are part of the Common Lisp Hyperspec so that is a non issue, they have been first class citizens since before Clojure existed (and are nicely integrated into things like loop)

[anaphor]

I think the parent was saying car and cdr are confusing because they don't actually map to registers (i.e. address register and decrement register). On the other hand you don't have to think too hard about the composed versions of them (caddr, cadar, etc...) but they can easily make your code seem obfuscated. Racket offers both that and first/rest anyway so you can choose which style you prefer.

[ANTSANTS]

I don't mind car, cdr, and their composed versions at all. `caddr` is much easier for me to read than `(head (tail (tail ...`

I think their usage is sometimes a code smell (the ad-hoc structs I mentioned), but they're really useful when you want to deal with an "unlispy" list of tokens (say, a parser for a language that isn't Lisp).

[ANTSANTS]

I mean, it's neat that you can build everything out of cons, in the sense that it's neat that Turing machines and the lambda calculus can perform any computation, but that doesn't mean it's a good idea to so in your code, any more than encoding numbers with Church numerals. Why use alists when you can use hash tables, why use... weird SICP-style struct list things when you could just use structs/vectors, etc.

>Hash Tables are part of the Common Lisp Hyperspec so that is a non issue, they have been first class citizens since before Clojure existed (and are nicely integrated into things like loop)

I'm not that familiar with Common Lisp, but a quick search didn't reveal a syntax for hash table literals, which I would consider to be a prerequisite for calling them first class citizens. There are other funny uses of lists (not improper ones, but still) in Common Lisp, like named procedure arguments[1]. I guess my point is, why encourage these weird constructs in the first place? Sexps are good at representing recursive structure, tables are good at mapping names to values, why not just have both and let them do what they're good at? Obviously Common Lisp is very old and can't be changed after the fact, I'm just trying to imagine the "ideal Lisp," whatever that is.

>Do they really confuse people new to Lisp? car+cdr is a pretty simple concept, when talking to new Lispers they usually don't complain about this.

I wasn't, but I've seen it confuse other people. Either way, it's a wart on the syntax that seems out of place in Scheme (not so in Common Lisp, which has lots of other warts ;))

>I come from a background using Common Lisp for system and web development so I may see things differently than people who were introduced to it academically.

Well, I was introduced to Scheme through SICP, but my motivation for picking it up was more practical than theoretical; I'd heard all the message board talk, Paul Graham essays, etc. proclaiming Lisp to be the most powerful, productive family of programming languages that will turn you into an expert programmer and so on, and wanted in on that ;) Maybe it's more practical to be Getting Shit Done™ instead of worrying about syntactic warts, but I think choosing to not have dotted pairs runs a bit deeper, by forcing the language to promote at least tables to a first class syntactic citizen.

[1] For the unfamiliar, it's something like:

  (message "Text" :style bold :color red)

Reads pretty well, but it still seems wrong to me (ymmv), like it's another spot in the language begging for first class hash tables. In Lua, you'd emulate named arguments by passing a table to the function. I wonder what Clojure programmers do here. This?

  (message "Text" {style: bold, color: red})

Which doesn't make a big deal for readability, but I bet if you want to use a macro or whatever accepts named arguments, it'd be much easier to deal with the one with the real hash table.

[ICWiener]

> but a quick search didn't reveal a syntax for hash table literals, which I would consider to be a prerequisite for calling them first class citizens

http://en.wikipedia.org/wiki/First-class_data_type

Hash-tables are part of the standard, and available in conforming implementations out-of-the-box. (make-hash-table) produces an instance of a table which can be passed as argument to function, be assigned to variables, ... I don't think syntax defines what feature is or isn't first-class, even though it has an impact on it's usage.

Adding support for litteral hash-tables is possible, as it was already said: define a reader macro; that macro could even simply use the utility function "plist-hash-table", from Alexandria, which is used like this:

     (alexandria:plist-hash-table '(:a 1 :b 2))

So that you could write #H{:a 1 :b 2} and get what you want.

On the first hand, people say that CL is bloated, but on the other hand, they want to have everything available by default. Do you expect Python or C++ implementations to come with regular expression built-in, or is it okay if you need to add an 'import' statement or link with Boost?

CL allows you to define libraries that can change even the basic surface syntax, for your convenience. Considering that there is more than enough defined in the specification, what benefits would be in having an "official" hash litteral syntax? consistency accross different projects? many people use the "cl-ppcre" regex library without problem.

Yes, regular expressions come in a library even though they could easily be labelled as a fundamental feature of a language, and hence be expected to be "first-class", or "built-in"; but again, that is not the case in CL, which is neither Perl nor Lua.

And about keyword parameters, this is just the surface syntax, which is on purpose based on a simple, basic representation of the syntax tree as lists of expressions.

Then, the compiler will work with the argument list in your function definition and take one or another approach to represent them in the object code. And I doubt that they are stored in a hash-table, which is unlikely an appropriate way to handle a couple of keyword arguments.

Similarly, when you define a class, everything is declared by a simple list of slots, where slots are lists of parameters: but in fact, the class might store the actual instance slots in a hash-table, an array, a list, a database or a memory mapped file.

The fact that you are mainly manipulating lists does not mean that everything is eventually represented as supposedly inefficient linked-list at runtime.

[ANTSANTS]

> http://en.wikipedia.org/wiki/First-class_data_type

Fair enough, pretend I said "first class syntactic object."

> CL allows you to define libraries that can change even the basic surface syntax, for your convenience. Considering that there is more than enough defined in the specification, what benefits would be in having an "official" hash litteral syntax?

Reader macros are powerful, no doubt. The benefit of building this into a language would mainly be syntactic regularity. CL as it stands has keyword args, alists, property lists, and real hash tables (probably amongst other things I don't know about), that all fill basically the same purpose of mapping names to values. If the creators of Common Lisp and its parent lisps had started with hash tables with standard reader syntax, they probably wouldn't have felt the need to introduce so many subtly different but conceptually identical concepts. Lua, for instance, uses its single table data structure both as syntax and as an internal representation for pretty much any case where it needs to map some names to some values; from keyword arguments to environments to "metatables" that provide fairly powerful metaprogramming support, they're all just tables that you can write literally and use the same functions to manipulate. You could say that alists fill this purpose and are even more syntactically regular since they're still sexps, but then why doesn't CL use alists for keyword args, and why don't Scheme's keyword arg extensions use alist syntax? Historical accident, or are alists just ugly? If you can admit that they're a little too ugly and verbose to include in function calls, then maybe you can see why I don't like writing alist literals.

> The fact that you are mainly manipulating lists does not mean that everything is eventually represented as supposedly inefficient linked-list at runtime.

I'm well aware that a compiler can easily expand keyword arguments into regular ones. I strongly doubt your compiler will transform all the literal alists in your program into hash tables, and replace all the alist-ref functions and so on operating on them with the equivalent hash table functions. As long as alists are more syntactically blessed than hash tables, people will use them where another data structure would be more appropriate.

[lispm]

Assoc lists make little sense as argument lists. Named arguments are basically like property lists. The point of argument lists for functions in Common Lisp is that they enable a contract. This requires special built-in support in the language. I want to see during compile time if an arg is missing or additional.

Example: the function WRITE takes several keyword arguments, but not :BAR. The compiler complains about wrong use.

    * (defun test () (write "foo" :bar 10))
    ; in: DEFUN TEST
    ;     (WRITE "foo" :BAR 10)
    ; 
    ; caught WARNING:
    ;   :BAR is not a known argument keyword.

Now if you allow a hashtable or some other data structure to be passed, then it would also be great, if one could specify at definition time which keys it takes, their default values, etc. We also may want to find out which values were default values, and which were actually passed.

The keyword argument facility for functions in Common Lisp provides all of that.

This allows you compile-time checks for code and makes interfaces easier to use.

Common Lisp also allows access to arguments as lists. This is simpler and more Lispy than using hash tables. For most purposes lists are useful enough and hash-tables would just add overhead.

[ICWiener]

> I strongly doubt your compiler will transform all the literal alists in your program into hash tables,

Time for experiment. SBCL doesn't, as far as I know. But you seem to imply that it is always better to use hash-tables, and I don't think so (even though hash-table can be implemented in a smart way, like in Lua).

Under roughly 10 elements, alist result in shorter code.

     (defun my-fun (x)
       (declare (optimize (speed 3) (debug 3) (safety 0))
                (type (member a b c) x))
       (let ((a '((a . 10) (b . 21) (c . 31))))
         (cdr (assoc  x a :test #'eq))))

     (disassemble #'my-fun)
     ; disassembly for MY-FUN
     ; Size: 53 bytes
     ; 04A7FF2F:       483B1592FFFFFF   CMP RDX, [RIP-110]         ; 'A
                                                                   ; no-arg-parsing entry point
     ;       36:       7511             JNE L1
     ;       38:       488B0D91FFFFFF   MOV RCX, [RIP-111]         ; '(A . 10)
     ;       3F: L0:   488B5101         MOV RDX, [RCX+1]
     ;       43:       488BE5           MOV RSP, RBP
     ;       46:       F8               CLC
     ;       47:       5D               POP RBP
     ;       48:       C3               RET
     ;       49: L1:   488B1D98FFFFFF   MOV RBX, [RIP-104]         ; '(B . 21)
     ;       50:       488B0D89FFFFFF   MOV RCX, [RIP-119]         ; '(C . 31)
     ;       57:       483B157AFFFFFF   CMP RDX, [RIP-134]         ; 'B
     ;       5E:       480F44CB         CMOVEQ RCX, RBX
     ;       62:       EBDB             JMP L0

This is roughly equivalent to a "case" construct (not shown here). Then, this is what I have with a hash-table:

     (let ((table (make-hash-table :test #'eq)))
       (declare (optimize (speed 3) (debug 3) (safety 0)))
       (setf (gethash 'a table) 10)
       (setf (gethash 'b table) 21)
       (setf (gethash 'c table) 31)
       
       (defun my-fun-hash (x)
         (declare (optimize (speed 3) (debug 3) (safety 0))
                  (type (member a b c) x))
         (gethash x table)))

     (disassemble #'my-fun-hash)
     ; disassembly for MY-FUN-HASH
     ; Size: 115 bytes
     ; 05467EB0:       .ENTRY MY-FUN-HASH(X)                       ; (FUNCTION (#)
                                                                   ;  (VALUES T # ..))
     ;      EE8:       8F4508           POP QWORD PTR [RBP+8]
     ;      EEB:       488D65E8         LEA RSP, [RBP-24]
     ;      EEF:       488B7805         MOV RDI, [RAX+5]
     ;      EF3:       4C8BC7           MOV R8, RDI
     ;      EF6:       498BF8           MOV RDI, R8
     ;      EF9:       BE17001020       MOV ESI, 537919511
     ;      EFE:       488B05FBFDFFFF   MOV RAX, [RIP-517]         ; #<FDEFINITION object for SB-IMPL::GETHASH3>
     ;      F05:       B906000000       MOV ECX, 6
     ;      F0A:       FF7508           PUSH QWORD PTR [RBP+8]
     ;      F0D:       FF6009           JMP QWORD PTR [RAX+9]
     ;      F10:       6A20             PUSH 32
     ;      F12:       B9604F4200       MOV ECX, 4345696           ; alloc_tramp
     ;      F17:       FFD1             CALL RCX
     ;      F19:       59               POP RCX
     ;      F1A:       488D490B         LEA RCX, [RCX+11]
     ;      F1E:       E917FFFFFF       JMP #x1005467E3A

The function with an hash-table has a constant size relatively to the number of elements in the table, which is fully known at compile-time. The size of the code with a case/alist grows linearly with the number of elements.

Now, let's compare speed.

In the following versions, I have added more elements in the alist, just to be sure the code with an hash-table is the shortest (from 'a to 'l).

     (time
      (dotimes (i 10000000)
        (dolist (x '(a b c d e f g h i j k l))
          (my-fun-hash x))))

     (time
      (dotimes (i 10000000)
        (dolist (x '(a b c d e f g h i j k l))
          (my-fun-case x))))

     (time
      (dotimes (i 10000000)
        (dolist (x '(a b c d e f g h i j k l))
          (my-fun x))))

Results:

   HASH:
    2.491 seconds of real time
    33,072 bytes consed
  
   ALIST:
    0.852 seconds of real time
    0 bytes consed
  
   CASE:
    0.778 seconds of real time
    0 bytes consed  
 

Even though in terms of footprint, the code tend to be larger with alist quite rapidly (10 elements), the resulting code is faster and does not allocate memory.

Also, just to clarify, alist and property lists have a different behavior than hash-tables, namely that the sequential access allows you to shadow values from another list: if you write (cons (cons 'a b) older-alist), you have a new list where the value for key 'a is b, and where values for other keys are those found in older-list (even though older-list also contains key 'a).

I don't quite remember what is my point anyway ;-) but it was fun to test the different behaviors.

[ANTSANTS]

Microbenchmarks are fun but silly. I tried to write a similar one for Lua, but LuaJIT ultimately recognized the program was useless and boiled it down to a 3 instruction loop:

  loop:
    add ebp, 1
    cmp ebp, 10000000
    jle loop

Anyway, in regards to your results, I could be wrong, but I think I read somewhere that LuaJIT uses linear search for tiny tables for this reason. Dunno what the threshold is, if there is one. The performance then would be similar to an alist, but saving a few bytes and cycles by not needing to deal with the extra list pointers and indirection.

> Also, just to clarify, alist and property lists have a different behavior than hash-tables, namely that the sequential access allows you to shadow values from another list: if you write (cons (cons 'a b) older-alist), you have a new list where the value for key 'a is b, and where values for other keys are those found in older-list (even though older-list also contains key 'a).

Oh yes, doesn't emacs make good use of this trick for its configuration variables? You can get the same effect with Lua's metatables, however, with a little elbow grease:

  parent = { a = "beep", b = "boop" }
  print(parent.a) --> beep
  print(parent.b) --> boop
  child = { a = "poing" }
  print(child.a) --> poing
  print(child.b) --> nil
  mt = { __index = parent }
  setmetatable(child, mt)
  print(child.a) --> poing
  print(child.b) --> boop

Not quite as easy as "cons", but very flexible; __index can be another table to searched if the lookup on the child fails (which in turn can have its own parent and so on), but it can also be a "metamethod" that is called whenever a lookup fails and can then do arbitrary things. A neat example off the top of my head: OpenGL has the peculiarity that you do not know the address of any of its functions until runtime, requiring a program to call a lookup function for each function to get a usable pointer. Declaring FFI function prototypes for many OpenGL functions is not such a big deal, but looking up hundreds of functions that you will never use can add significantly to startup time. So someone (possibly an HN user?) wrote an OpenGL FFI library that uses the __index metamethod in a clever way; the first time a function like "GL.CreateShader" is called, the lookup fails, and the __index metamethod mangles the index name a bit and in turn calls (on Windows) the C function wglGetProcAddress to look up its address, which it then stores in the original table. Using this library, you can write code that uses GL functions willy-nilly, and their addresses will automatically be looked up at runtime the first time they are used.

Sure, you can do the same thing in any language with macros or a preprocessor, but is that cool or what?

[jamii]

> I wonder what Clojure programmers do here.

About the same.

    (defn message [text & {:keys [style color]}]
      ...)

    (message "Text" :style :bold :color :red)

Because typing those extra two brackets is apparently too much trouble.

[cgag]

I would just accept the hash, I think a lot of us prefer it to the pseudo keyword args.

    (defn message [text {:keys [style color]}] ...)
    (message "Text" {:style :bold :color :red})

To the person asking, that {:keys [style color]} is destructuring, it's not required, but it's nice, it's equivalent to doing:

    (defn message [text opts]
       (let [style (:style opts)
             color (:color opts)]
          ...))

[ANTSANTS]

I can't say I expected it to look like that.

[aidenn0]

I don't understand the argument for using a hash-table for kw-args. Hash tables only win against plists when you don't need to read every single element; when you always need to read every single element, then it's a losing proposition to use a hash-table instead of a plist.

Yes I agree with your later comment about syntactic blessing and all, but I think you picked a terrible example by choosing the single place in lisp where plists make more sense than hash tables.

[ANTSANTS]

Maybe it was a bad example, but in either case, I would expect a "sufficiently smart compiler" to transform

  (defun fun (a &key (b 0) (c 0)) ...)
  (fun 5 :b 3)

into something like

  (defun fun (a b c) ...)
  (fun 5 3 0)

Whether it does so as a regular macro or as a special form of the compiler doesn't matter, there's no need for there to be any difference in efficiency between representing keyword args with tables or plists.

As I later mentioned in another comment, it was a mistake to talk about hash tables when I was really more interested in the abstract map datatype.

[lispm]

Named arguments are there to be interfaces to procedures. With hash-tables you know nothing about them. With named arguments, we can ask for argument lists, check for missing arguments, complete arguments, prompt for arguments, ... Much of that can be done in the IDE or at compile time.

Named arguments had been introduced to Lisp with MDL (a Lisp dialect, brought to Lisp Machine Lisp and then to Common Lisp).

Using hash-tables for it is a step back from the view of development support.

[ANTSANTS]

  ; CL keyword arg syntax, taken from Practical Common Lisp
  (defun foo (&key (a 0) (b 0)) (+ a b))
  (foo :a 1) -> 1
  (foo :a 1 :b 2) -> 3

  ; hypothetical table keyword arg syntax
  ; clojure defines commas to be whitespace, you can pretend they aren't there if you want
  (defun foo ({a: 0, b: 0}) (+ a b))
  (foo {a: 1}) -> 1
  (foo {a: 1, b: 2}) -> 3

I fail to see why the table syntax would be any less usable or introspectable by compilers or editors. They'd have the advantage (which would be shared with alists, if Common Lisp had used them for keyword args) of sharing the syntax for keyword args with a syntax for optional structure properties (think XML's attributes)

[lispm]

Basically the syntax for specifying keyword arglists is based on assoc lists. The calling syntax is based on property lists.

There is nothing to be gained by hashtables.

The Common Lisp keyword syntax actually is a bit more powerful then what your PCL example shows:

(defun foo (&key ((var keyword) init var-arg-supplied-p) ...)

Generally I think it is a slight mistake to use specific data types in arglists - basically mixing syntax and data types.

[ANTSANTS]

> Generally I think it is a slight mistake to use specific data types in arglists - basically mixing syntax and data types.

That's kinda what alists and plists do! They mix up the abstract "map" or "table" data structure (something that maps keys to values) with some concrete implementation of it. '((key . value) (key . value)) and '(key value key value) are literal representations of two particular map-like data structures, whereas {key: value} or {key = value} could have any concrete representation that the language implementer desires.

Maybe I should have dropped the "hash" from "hash table" in my above posts, so that it was clear that I was interested in the syntactic benefit of using one syntax for mapping names to values, and not some imagined efficiency gains from having (read keyword-function-call) include a hash table instead of an alist.

[arethuza]

In Common Lisp you can create your own syntax for hash table literals fairly easily with a reader macro:

http://frank.kank.net/essays/hash.html

Mind you - reader macros should be used with a degree of care otherwise you basically create completely new language. I remember working on a project where I got a drop of some code from one of the other organizations in the project (by tape, this was a long time ago) opening the file and wondering what language the code was written in and taking a while to realize that it was indeed Lisp - but one that had used reader macros to do strange and terrible things.

Mind you, once I learned about reader macros I went on to do my own crazy stuff with them - they are almost too powerful a feature (almost!).