02:40 (quit) jonrafkind: Ping timeout: 252 seconds
03:51 (join) masm
04:06 (join) spidermario
08:02 (join) tony_
08:03 (join) mceier
08:03 (quit) tony_: Client Quit
08:21 (quit) clklein: Ping timeout: 252 seconds
09:07 (join) hanDerPeder
09:17 (quit) hanDerPeder: Quit: hanDerPeder
09:21 (join) hanDerPeder
09:28 tonyg: Hi all
09:29 tonyg: why does (expand '+) => #<syntax +>, but (expand 'other) => #<syntax (#%top . other)> ? Shouldn't they both be #%top?
09:43 Lajla: tonyg, I guess because + is bound to and other is not
09:43 mao_: because  '+ is builtin?
09:43 mao_: or what Lajla said
09:43 Lajla: tonyg, try it with (define other +) and see what happens
09:47 (join) anRch
09:51 tonyg: Well, after (define other +), the results are still the same.
09:51 mao_: Welcome to MzScheme v4.2.1 [3m], Copyright (c) 2004-2009 PLT Scheme Inc.
09:51 mao_: > (expand '+)
09:51 mao_: #<syntax>
09:51 mao_: > (expand 'crap)
09:51 mao_: #<syntax>
09:51 mao_: >
09:51 mao_: maybe that's because I have plt
09:51 tonyg: But after (define + 3), (expand '+) => #<syntax (#%top . +)>
09:51 tonyg: mao_: try syntax->datum
09:52 tonyg: so it's looking like it's because + is builtin somehow. Maybe because it's bound to a primitive or something. Interesting.
09:52 mao_: reference to undefined identifier: synax->datum
09:52 mao_: i have old version prob
09:52 mao_: w/e
09:53 tonyg: "synax->datum" -> "syntax->datum" ? :-)
09:54 tonyg: Hm, it's interesting: it's like, when +'s binding is original, expand treats it like a *local* variable!
10:06 mao_: tru
10:08 Lajla: Using syntax-case for me is kind of like 'Okay, it works, but I still don't really know what I'm exactly doing'
10:08 Lajla: And the specs are soooo vague and hard to get what they understand
10:08 tonyg: yes, agreed. I found reading the original papers helped -- which suggests, for me, that the only real way to grok the system is to understand the whole mechanism of it.
10:09 tonyg: It seems that surface understanding is insufficient, for me, in that it leads to occasional wtf moments
10:09 tonyg: and I have to remind myself of the full algorithm to straighten out what's happening
10:10 Lajla: tonyg, you understand syntax-case?
10:10 tonyg: heh
10:10 tonyg: occasionally
10:10 Lajla: I understand continuations and monads though
10:11 tonyg: yeah - something about syntax-case feels non-local.
10:11 tonyg: Not sure.
10:11 Lajla: Let's see what I can boast with ehh, references, pointers, continuations, monads, and last but not least.... the entire mechanism and rules behind what sigil you should use in perl 4. \o/
10:11 tonyg: Whereas continuations and monads are amenable to local reasoning
10:12 tonyg: hahaha
10:12 tonyg: yeah :)
10:12 Lajla: Those sigils have no logic behind them whatsoever
10:12 tonyg: have you looked at the original syntax-case papers?
10:13 Lajla: tonyg, what is the original?
10:13 tonyg: # R. Kent Dybvig. "Writing Hygenic Macros in Scheme with Syntax-Case". Computer Science Department, Indiana University. TR-356. June 1992. Available online: ps.
10:13 Lajla: The one pdf that is circling around?
10:13 tonyg: # Robert Hieb, R. Kent Dybvig and Carl Bruggeman. "Syntactic Abstraction in Scheme". Computer Science Department, Indiana University. TR-355. June 1992. Available online: ps.
10:13 Lajla: Yeah, that one
10:13 tonyg: ftp://ftp.cs.indiana.edu/pub/scheme-repository/doc/pubs/iucstr356.ps.gz
10:13 tonyg: ftp://ftp.cs.indiana.edu/pub/scheme-repository/doc/pubs/iucstr355.ps.gz
10:13 tonyg: those two
10:13 Lajla: Yeah, I read them.
10:13 tonyg: ok
10:14 Lajla: Also, I think that this supposed 'syntax object' goes against the idea of homo-iconicity.
10:14 tonyg: yeah, i find myself at times laboriously reminding myself of how my syntax is transformed all the way down to lambdas and back, in order to get where the binders come from
10:14 Lajla: Surely syntax should just be manipulated by the same functions that manipulate any list?
10:14 tonyg: hmm
10:14 tonyg: no, i disagree
10:14 tonyg: code is a kind of data, but it's not as simple as a list
10:15 Lajla: Hmm, well, I'd praefer it that it is.
10:15 tonyg: you recover homoiconicity via syntax-quote and syntax-quasiquote (whatever they're called)
10:15 Lajla: You mean #'
10:15 Lajla: ?
10:15 tonyg: you then have quote/quasiquote for lists, and the syntax-* variants for syntax
10:15 (quit) spidermario: Read error: Connection reset by peer
10:15 tonyg: yeah
10:15 Lajla: That's just synrax
10:15 tonyg: well, so is everything ;-)
10:16 Lajla: #'bla <-> (syntax bla)
10:16 tonyg: Yes, right
10:16 Lajla: No, I mean that
10:16 tonyg: sorry when I meant syntax-quote before I should have said "syntax"
10:16 tonyg: as you say
10:16 tonyg: It does the same job
10:16 tonyg: it's notionally a syntax-quote kind of thing :)
10:16 Lajla: Well, I'd find it more elegant if syntax truly just were lists like any other.
10:17 Lajla: Basically, your compiler then becomes a function that takes in a scheme datum and writes assembly to a port as a side effect or something
10:17 Lajla: Like (compile datum port)
10:17 tonyg: Mmm, but that's what is going on already. Just replace ' with #'
10:17 Lajla: Well, except that it doesn't take a normal datum as you enter it, but 'syntax' objects.
10:17 tonyg: Your compiler would be typed compile : Syntax Port -> Void
10:17 tonyg: that's data like any other
10:18 tonyg: cons cells aren't magic
10:18 tonyg: frinst I'd balk at representing numbers using them
10:18 tonyg: in real programs
10:18 Lajla: Like, I want (compile (list 'if #t (list 'display "hello world"))) to work, basically
10:18 tonyg: (compile #`(if #t (display "hello world")) my-port)
10:18 Lajla: Yes, and that is my problem.
10:19 Lajla: THat compile does not accept ordinary lists
10:19 Lajla: But 'syntax objects'
10:19 tonyg: (compile (datum->syntax-object '(if #t (display "hello world"))) my-port)
10:19 Lajla: Yap, and thatis my problem
10:19 Lajla: this supposed 'syntax object'
10:19 tonyg: the reason is that ordinary lists are not sufficiently expressive to cover everything that needs representing about syntax
10:19 Lajla: Exactly
10:19 tonyg: namely, the structure of the bindings involved
10:20 Lajla: Well, you can actually solve this.
10:21 Lajla: Say you have a function (macro-transform datum macros ...)
10:21 Lajla: Where you have 'macro objects'
10:21 Lajla: Which are pieces of data which describe what keywords they match, how, and how they treat bindings.
10:21 Lajla: You just have to transform it all at once.
10:21 Lajla: You can't just do it recursively.
10:21 Lajla: You have to feed it the top level datum
10:22 (quit) hanDerPeder: Quit: hanDerPeder
10:22 tonyg: That sounds awfully similar to what expand does, except that it uses the current global environment instead of an explicitly-given environment.
10:23 Lajla: Yap
10:23 Lajla: But then you can circumvent the use of 'syntax objects' and can use macro transformers like functions on any list.
10:24 tonyg: You still need syntax-objects *or similar* internally, to represent bindings. For example, define-syntax needs to hold the transformer in such a way that bindings are preserved hygienically.
10:24 tonyg: The API can take lists, sure
10:24 tonyg: but the macro transformers themselves can't.
10:24 Lajla: tonyg, sure, but this is done internally in the transform function
10:24 tonyg: After all, the reader does just read sexprs to begin with
10:25 tonyg: I'm not sure what you're asking for, then: did you want an API in terms of sexprs (in which case there already is one), or an implementation in terms of sexprs (in which case you can't have one ;-) )?
10:25 (join) hanDerPeder
10:25 Lajla: What basically bugs me is in racket is this. I can actually use (define foo :random-syntax-case-expression) and then do (define-syntax keyword foo)
10:26 Lajla: But I can't then use (foo :arbitrary-list:) to achieve the same transform at runtime
10:26 Lajla: It then outputs a syntax object, not a list
10:27 tonyg: which is what datum->syntax and syntax->datum are for, right? the transformer needs syntax-objects to maintain hygiene, so you can transform sexps into syntax at the API layer
10:27 tonyg: if you want to drop hygiene, you can write simple sexp transformations using match and quasiquote pretty easily
10:28 tonyg: it's just not syntax anymore is all
10:28 tonyg: it only *looks* like syntax ;-)
10:57 (quit) anRch: Quit: anRch
11:41 (join) aaSSaa
11:42 aaSSaa: Hi
11:45 (quit) aaSSaa: Client Quit
12:18 (quit) adlaius: Quit: Leaving
12:40 Lajla: tonyg, but it's stil not required if you transform the entire source code at once.
12:40 Lajla: It just means that the expand function cannot be recursive and call itself.
12:40 Lajla: Or how about this:
12:40 Lajla: (expand datum some-datum-carrying-all-info-of-bindings transformer-information ...)
12:40 Lajla: That could work
12:41 Lajla: It could call itself recursively with an updated version of that binding info
14:05 (join) Madars
14:07 Madars: I have a simple program '#lang racket\n(require some-lib)\n(define (somefunc) (...))'. however if somefunc is already defined in some-lib, then I get "identifier is already imported" error. is there a way I could redefine somefunc?
14:10 (join) b-man_
14:20 Lajla: Madars, you can use set!
14:20 Lajla: (set! somefunc (lambda () ...))
14:24 Madars: oh, that could do
14:25 Madars: thanks!
14:46 Lajla: Madars, you could also use (define Lajla 'awesome) and hope that the lib does not already bind a symbol 'lajla'.
14:50 Madars: Lajla, I didn't specify it before, but I want to redefine a function from some-lib, so that other functions from some-lib would call my redefined function. therefore I am actually hoping that lib does bind a symbol with the same name :-)
14:50 Lajla: Madars, I guess then your best shot is to define the lib
14:51 Lajla: If you for instance define cdr, the the length function will not fail for instance
14:51 Lajla: re-define cdr
15:54 (join) offby1`
15:56 (quit) offby1: Ping timeout: 272 seconds
16:33 (nick) offby1` -> offby1
16:33 (quit) offby1: Changing host
16:33 (join) offby1
16:46 (quit) eli: Remote host closed the connection
16:46 (join) eli
17:37 (join) jao
18:01 (join) Karl
18:02 (quit) Karl: Client Quit
18:39 (quit) masm: Quit: Leaving.
18:47 (quit) b-man_: Remote host closed the connection