#clojure log - Sep 04 2010

The Joy of Clojure
Main Clojure site
Google Group
IRC
List of all logged dates

4:34 Raynes: chouser: Whenever you hop back on: I replied to the group post with some snapshots. <3

5:04 triyo: .quit

5:05 oops

6:19 ryoko-sha: I'm trying to construct a loop in which an index integer variable is parsed from a string via (java.lang.Integer/parseInt "1'), but am getting an exception java.lang.RuntimeException: java.lang.IllegalArgumentException: recur arg for primitive local: i must be matching primitive (test.clj:10)

6:19 ...mystified

6:21 since (type 1) => java.lang.Integer and also (type (java.lang.Integer/parseInt "1")) => java.lang.Integer

6:21 mrBliss: try wrapping it in (int ..)

6:21 Integer != int

6:22 ryoko-sha: yep same deal !

6:23 mrBliss: what are the other arguments to the recur call?

6:24 ryoko-sha: (recur (+ i 1))...

6:24 couldn't be simpler!

6:24 LaPingvino: (unchecked-add i 1) ?

6:24 mrBliss: try wrapping that in (int ..) or replace it by (inc ..) (I'm also guessing a bit ;-)

6:25 LaPingvino: + is not simple :P

6:25 + is a function on (add)

6:28 ryoko-sha: [LaPingvino] - bingo! thanks (unchecked-add)

6:28 is (unchecked-add) very naughty in some way?

6:28 LaPingvino: (+) does the boxing

6:29 which destroys the primitives

6:29 basically :)

6:29 ryoko-sha: thanks - its all al learning experience!

6:29 LaPingvino: ryoko-sha: unchecked-add can cause overflows

6:29 because you use primitives

6:29 mrBliss: it's in the name:

6:30 ,(unchecked-add Integer/MAX_VALUE 1)

6:30 clojurebot: -2147483648

6:30 LaPingvino: and it will not automatically change the type to a higher capacity

6:30 there you go :)

6:30 ryoko-sha: super

7:26 raek: ,(doc num)

7:26 clojurebot: "([x]); Coerce to Number"

7:26 raek: http://gist.github.com/442269 <-- primitives and recur, with comment from rhickey

7:27 (the bug that the example illustrated has been fixed)

7:30 ryoko-sha: I had to insert a call to num in the loop form: (defn fact [n] (loop [n n, (num 1)] (if (zero? n) acc (recur (dec n) (* acc n)))))

7:30 *sigh* the loop binding vector should of course have been [n n, acc (num 1)]

7:32 I don't think you should have to resort to unchecked-add

7:48 ryoko-sha: raek: yep - that worked for me - thanks

7:49 raek: I think that will throw an exception on overflow, rather than silently wrap around

7:50 ryoko-sha: better - although my values for i are very small - defensive programming etc...

8:11 fliebel: *puzzled* what is the long way to make a map/dict/whatever? {:i "mean" :those "things"}

8:12 mrBliss: ,(hash-map :a 1 :b 2)

8:12 clojurebot: {:a 1, :b 2}

8:12 fliebel: ah… I was thinking in the map direction, but map is something else :P

8:12 mrBliss: ,(zipmap [:a :b] [1 2])

8:12 clojurebot: {:b 2, :a 1}

8:13 mrBliss: looks more like the other map :-)

8:13 raek: ,(into {} [[:a 1] [:b 2]])

8:13 clojurebot: {:a 1, :b 2}

8:24 fliebel: Can anyone give em an example of max-key? the docs are chinese to me, but the fn name sounds like what I need. I want to get the key from a map whose number is the heigest.

8:25 raek: http://clojuredocs.org/v/1506

8:25 the key function should return a number

8:26 fliebel: ah...

8:26 raek: the element in the sequence for which the key function returns the highest number is the one returned by max-key

8:27 fliebel: raek: so Can I apply that to a map?

8:28 ,(apply max-key {:a 1 :b 3})

8:28 clojurebot: [:b 3]

8:28 fliebel: yay!

8:29 uuhm, nope

8:33 okay, this seems to work :)

8:33 ,(apply (partial max-key last) {:a 1 :b 3 :c 2})

8:33 clojurebot: [:b 3]

8:38 raek: (apply max-key key {1 :a, 10 :b, 100 :c})

8:38 ,(apply max-key key {1 :a, 10 :b, 100 :c})

8:38 clojurebot: [100 :c]

8:39 fliebel: ty

8:39 raek: the "key" function that the docs refer to could be called the "scale" function

8:39 as in "pick the element that has the highest value on this scale"

8:40 ,(apply max-key val {:a 1, :b 10, :c 100})

8:40 clojurebot: [:c 100]

9:07 fliebel: What is the proper way to do this?

9:07 ,(map reverse {:a 1 :b 2})

9:07 clojurebot: ((1 :a) (2 :b))

9:08 mfex: hey fliebel, the proper way to do that is to do it that way :)

9:08 fliebel: mfex: I mean, I want a map witht he keys and values switched

9:09 Raynes: -> (into {} (map reverse {:a 1 :b 2}))

9:09 sexpbot: java.lang.ClassCastException: java.lang.Integer cannot be cast to java.util.Map$Entry

9:09 Raynes: wut

9:09 fliebel: yea...

9:09 Raynes: -> (apply hash-map (map reverse {:a 1 :b 2}))

9:09 sexpbot: => {(2 :b) (1 :a)}

9:09 * Raynes is too tired for this stuff.

9:09 fliebel: tried all of that...

9:10 _mst: ,(let [m {:a 1 :b 2 :c 3}] (zipmap (vals m) (keys m)))

9:10 clojurebot: {3 :c, 2 :b, 1 :a}

9:11 _mst: (it always takes me about three minutes to remember that it's called "zipmap" :)

9:11 fliebel: _mst: Yea, I never get there at all, I end up browsing the sipper package.

9:11 *zipper

9:13 mfex: ,(into {} (map (juxt val key) {:a 1 :b 2}))

9:13 clojurebot: {1 :a, 2 :b}

9:13 Raynes: ->(into {} (map (comp vec reverse) {:a 1 :b 2}))

9:13 sexpbot: => {2 :b, 1 :a}

9:14 Raynes: I like mine better.

9:14 :D

9:14 hoeck: ,(clojure.set/map-invert {:a 1 :b 2 :c 3})

9:14 clojurebot: {3 :c, 2 :b, 1 :a}

9:14 Raynes: I like his the best.

9:15 >_>

9:15 fliebel: hoeck: Yea, that's the better way to do things :)

9:15 Raynes: Why are there map functions in the set namespace...?

9:15 They trick me out.

9:16 hoeck: maybe because of their relational nature

9:17 ssideris: what's wrong with (apply hash-map (map reverse {:a 1 :b 2}))

9:17 it looks like the easiest to read

9:17 apart from the map-invert of course

9:17 hoeck: core.set has also rename-keys, index and join functions

9:17 fliebel: ssideris: Try it...

9:17 Raynes: ssideris: The fact that it doesn't work is a pretty bad deal breaker.

9:18 ssideris: oops, yeah sorry

9:18 Raynes: :)

9:18 ssideris: but the zipmap solution is ok, right?

9:19 fliebel: This is how map-invert does it: (reduce (fn [m [k v]] (assoc m v k)) {} m))

9:21 jcromartie: I'm playing with designing a system that uses the SICP notion of "streams" for representing the state of an object

9:21 bec

9:22 Raynes: ssideris: I think any of the examples above are reasonable.

9:22 jcromartie: because the sequence of states of certain things over time is really important for the type of system I'm building

9:22 but I've hit a couple of spots where I'm not really sure how to represent things

9:23 fliebel: Is there any difference between using a map as a function with a key as argument, or the reverse? Both seem to work.

9:23 jcromartie: mostly because you can't serialize functions

9:23 fliebel: only one works when the map is nil

9:25 (and that's the keyword-as-fn)

9:28 mfex: fliebel: :keyword as the function works when the map is nil and with records as well, but for maps with keys that are not lookup functions there is no choice

9:44 fliebel: Is there something like rand-nth that lets you define the odds of getting certain items? The easy way out is letting certain items occur more often in the seq, but that seems stupid.

9:48 raek: fliebel: look at the "wrand" function in rhickey's ant.clj example

9:48 http://raek.se/ants.clj

9:50 fliebel: raek: That looks like what I need :

9:50 :)

10:24 Bahman: Hi all!

10:25 ssideris: fliebel: what you want to do is called "roulette selection"

10:26 it's used in genetic programming among other things

11:02 fliebel: Aaah, the dreaded "Can only recur from tail position"… what is considered the tail position? I'm sure this has been asked before, I'm even sure *I* asked it before. I'm recurring from the if clause of an if-let, with and else clause following.

11:04 fullets: ,((fn [x] (if-let [_ (pos? x)] (recur (dec x)) :done)) 3)

11:04 clojurebot: :done

11:05 fullets: Could you pastebin your code?

11:06 fliebel: fullets: http://gist.github.com/564552

11:06 (near the end)

11:06 line 43

11:09 fullets: That let is the final statement in the if

11:09 That is, there is no else clause as written

11:10 fliebel: you're right, I think I got my parens mixed up…

11:13 yea, works :)

11:14 See here, my new and improved rock-paper-scissors bot: http://gist.github.com/564552

11:18 It has zero randomness and will hopefully improve over time. By knowing how it works, you can win.

11:19 lpetit: hello

11:19 fliebel: lpetit: hey

11:21 lpetit: do some people have feedback on ccw version 62 to give ? I'm considering releasing a corrective version, so if more must be included ...

11:21 ... now's the time for it :)

11:44 kumarshantanu: lpetit: I have a small correction request for CCW

11:45 lpetit: when one wants to create a clojure file with dash in the file name, it should say "Consider using underscore instead of dash"

11:46 lpetit: instead of simply saying "invalid filename"

12:03 zmyrgel: hi, how can I cast java class in clojure?

12:08 for example, findComponentAt returns Component which I want to cast to JLabel

12:41 kumarshantanu: zmyrgel: casting is not required in clojure

12:42 zmyrgel: because clojure is a dynamically types language

12:46 zmyrgel: kumarshantanu: so in the above scenario it automatically finds the that the Component is actually JPanel and I can call JPanels method for it?

12:47 kumarshantanu: zmyrgel: yes, it depends on the object type, not on the reference type (because there is no reference type in clojure)

12:52 zmyrgel: kumarshantanu: ok, thanks

13:03 ApeShot: So I've compiled the latest clojure.jar and clojure-contrib.jar, and put them in the lib folder of my clojure project, and started up (successfully) slime from within emacs, but I can't (use 'clojure.contrib.string) or (use 'clojure-contrib.string), even though there does seem to be such a lib in the clojure-contrib source I compiled from. Any idea what gives?

13:16 I have been away from clojure for a few months, at least - has something changed in the way that libraries work?

13:16 I can (use 'clojure.contrib.str-utils) but, though it executes without error, this does not seem to actually import any of the functions in that library

14:01 fliebel: What functions are there to return a lazy cumulative map/seq? So that every item is itself plus every previous item. [1 2 3] -> [1 3 6] and then with a hash-map. *browsing the core docs*

14:03 somnium: ,(take 5 (reductions + (iterate inc 0)))

14:03 clojurebot: (0 1 3 6 10)

14:03 fliebel: somnium: Nice :) thank you

14:04 somnium: oh, I guess we can do (range) nowadays instead (iterate ...)

14:04 np

14:04 arohner: map entries can't hold metadata?

14:04 fliebel: somnium: With the slight difference that I'm not doing a list of 1 2 3, but rather 1 5 8, and then want to have 1 6 something

14:05 arohner: I wish everything, including ints and strings, could hold metatdata

14:05 fliebel: arohner: Anything that's java can't

14:05 arohner: fliebel: I know. It's just getting annoying

14:07 kumarshantanu: what's the best way to transform [:a 10 :b 20 :c 30] --> {:a 10 :b 20 :c 30} ? the two ways I know are (reduce #(assoc %1 (first %2) (last %2)) {} (partition 2 [:a 10 :b 20 :c 30])) and (apply assoc {} [:a 10 :b 20 :c 30])

14:08 tomoj: arohner: how would you get a map entry with some metadata into a map?

14:08 arohner: tomoj: using into

14:08 (into {} seq-of-map-entries

14:09 chouser: ,(apply array-map [:a 10 :b 20 :c 30])

14:09 clojurebot: {:a 10, :b 20, :c 30}

14:10 tomoj: ,(let [x (first {:a 2})] (identical? x (first (into {} [x]))))

14:10 clojurebot: false

14:10 kumarshantanu: arohner: (with-meta {:a 10 :b 20} {:motd "Hello World"})

14:10 tomoj: I suppose it would copy the metadata in if it were there

14:13 somnium: has anyone done an ADT library with deftype/defrecord?

14:16 integers aren't valid names for fields, right?

14:16 arohner: somnium: right

14:17 somnium: but something like {:ADT_TAG t, T0 a ... TN z}, and then a pattern matching macro could add the right field names

14:17 then you could use ADT pattern matching internally, but still plug the class into protocols

14:17 kumarshantanu: chouser: thanks, that's the first kind of use of array-map I noticed now

14:18 somnium: hmm, maybe I'll give it a go

14:18 kumarshantanu: added an example @ array-map on clojuredocs.org too

14:19 somnium: this red-black-tree implementation convinced me of their potential utility: http://gist.github.com/561648

14:20 tomoj: somnium: https://gist.github.com/cf84a59543e31f42d2a2

14:20 re ints as field names

14:20 not that that's a good thing to do!

14:21 somnium: tomoj: yeah, just some ordered naming convention for (TAG a ... n) matching like haskell

14:21 tomoj: ,(class (symbol "1"))

14:21 clojurebot: clojure.lang.Symbol

14:22 somnium: ?

14:23 tomoj: my point was that while (defrecord Foo [1 2 3]) fails, with a macro we can do it

14:23 somnium: right

15:02 fliebel: I give up… How can I get a map sorted by its values? There are all sorts of ways to do it, but none actually works for me :(

15:04 tomoj: you can't, really

15:05 arohner: clojure's sorted maps are sorted by key

15:05 tomoj: ,(sort-by val {:a 3 :b 2 :c 1})

15:05 clojurebot: ([:c 1] [:b 2] [:a 3])

15:05 tomoj: but once you've done that you don't have a map anymore

15:05 arohner: if your map is 1:1, you can swap the keys and values

15:08 fliebel: I can live with tomoj's thing… I'm just doing ugly things I guess

15:09 arohner: That could also work

15:11 LauJensen: Map lookups are usually by keys tomoj, so sort-by val makes a lot of sense to me

15:12 n2n3: ,

15:12 clojurebot: EOF while reading

15:12 n2n3: ,(System/exit 0)

15:12 clojurebot: java.security.AccessControlException: access denied (java.lang.RuntimePermission exitVM.0)

15:13 n2n3: ,{}

15:13 clojurebot: {}

15:31 jkkramer: fliebel: clojure.contrib.priority-map was recently added -- http://github.com/clojure/clojure-contrib/blob/master/modules/priority-map/src/main/clojure/clojure/contrib/priority_map.clj

15:31 it sorts by value

15:32 fliebel: jkkramer: Nice, but it's solved already. Good for next time :)

15:33 LauJensen: jkkramer: Whats the use case for something like that?

15:33 jkkramer: LauJensen: priority queue, for one

15:33 LauJensen: http://clj-me.cgrand.net/2010/09/04/a-in-clojure/

15:35 LauJensen: ok, will take some time to grok

15:36 hmm, no I think I get it

15:37 jkkramer: i found out about this right after jerry-rigging a sorted-set priority queue for dijkstra's algorithm

15:48 LauJensen: jkkramer: for fun or work ?

15:49 jkkramer: LauJensen: currently for fun. writing a general-purpose graph lib

15:49 LauJensen: cool

15:51 jkkramer: planning to make a mailing list post about it to get some feedback

15:51 seems there's no fleshed-out graph lib for clojure yet

15:57 cmihai: What was that tool that gave command history to any repl called?

15:57 fliebel: jline?

15:57 rlwrap?

15:57 cmihai: That was the one, thanks :-)

15:58 LauJensen: cmihai: slime

15:59 cmihai: Nah, I wanted it for db2 :P

15:59 emh: jkkramer: cool. I'm also working on a graph library. more special purpose than general though. for my master thesis, due in 1 week

15:59 cmihai: I remembered using it in conjunction with clojure, just forgot the name

15:59 LauJensen: emh: ouch, work fast :)

16:00 emh: yeah :)

16:02 jkkramer: emh: awesome. i'm trying to write a protocol- & record-based API that would be easy for anyone to write algorithms for

16:04 emh: are you doing anything fancy like multigraphs? or just simple/weighted/directed graphs?

16:05 emh: jkkramer: I'm also using protocols heavily. if you want to peek at my very work-in-progress-thesis, look here: http://rasmus.uib.no/~ehv022/thesis.pdf

16:05 jkkramer: just simple undirected graphs for now. will probably extend it later

16:05 after thesis

16:06 depends on if I get a ph.d. after

16:07 jkkramer: since I'm pressed for time I can't make it as elegant as I would wish. just have to crank out code as fast as possible

16:09 jkkramer: emh: very interesting!

16:10 emh: the concept of neighborhoods is new to me

16:10 emh: jkkramer: thx. we had all of it implemented in Java, but it's a huge ugly mess. I figured I'd have to rewrite in pseudo-code for presentation in my thesis, but clojure is sufficiently close to pseudocode so..

16:11 jkkramer: neighborhoods are very important in the decompositions, so I devote more data structures and protocols to them than would be normal in a general graph lib

16:12 jkkramer: emh: http://gist.github.com/565453 is the basic graph API so far. there are also some basic algorithms to go with it

16:13 emh: would be cool to be able to mix in capabilities with protocols. e.g., labeled graphs, neighborhoods

16:14 emh: jkkramer: yeah

16:14 jkkramer: looks pretty good

16:14 jkkramer: one thing I've thought about is whether it's worth distinguishing between read / write interfaces for graphs

16:15 jkkramer: since read-only graphs might be implemented more efficiently

16:16 jkkramer: emh: interesting. with protocols you could define ReadonlyGraph record(s), and it could still work with all the polymorphic methods, and be able to convert easily to a writeable graph

16:17 emh: hmm, yeah, true, easy to do conversions

16:19 I need to get some nice infrastructure for mixins and defaults later. I looked at the methods a la carte lib, looks like a nice start at least

16:21 http://onclojure.com/2010/08/26/reusable-method-implementations-for-deftypes/

16:23 I'm using some quick hacks right now, sometimes implementing protocols on top of the Java interfaces for other protocols. for example implementing tex serialization protocol on top of interfaces for Graph protocol etc

16:26 jkkramer: seems there's still a lot of sussing out to be done when it comes to best practices for protocols, types, and records

16:26 emh: agreed

16:59 LauJensen: Does anyone have some pointers on how to make a design resizeable (grow/shrink) using contrib.miglayout

16:59 ?

16:59 pdk: my favorite thing about freenode's ircd is that it makes you leave half the channels you're on just to change your name if you dc

17:07 LauJensen: bozhidar: ^^

17:07 bozhidar: LauJensen: hi

17:07 LauJensen: bozhidar: hi - got a clue about my q above?

17:09 bozhidar: LauJensen: I haven't gotten to using contrib.miglayout yet, but miglayout supports this on two levels

17:09 first you can declare growing columns/rows

17:09 and min/preferred/max sizes

17:09 and then you can use the same technique on the components themselves

17:10 when you add them to a container using miglayout

17:10 LauJensen: I might be in trouble with the wrapper then, I think it only supports components restraints. My setup is jframe -> jpanel -< jsplitpane -> 2 components. Where in that do you see the columns?

17:11 bozhidar: in the MigLayout object constructor

17:11 it takes a string representation

17:11 of the layout

17:11 with settings for every row and column

17:11 actually two strings

17:11 LauJensen: oh

17:11 bozhidar: one for the rows

17:12 and one for the columns

17:12 new MigLayout("wrap 1", "[grow]", "[grow]") - layout with one row and one column that grow

17:13 you should add additional [] for every other row/column you have

17:14 LauJensen: I don't think the wrapper lib supports constructor args, can they be passed in after construction ?

17:14 hmm, no actually it might

17:14 bozhidar: Thanks for the tips, I'll fiddle

17:15 bozhidar: there are setters for them

17:15 but oddly they take different arguments

17:15 LauJensen: oddly?

17:15 bozhidar: I personally always use the constructors

17:15 LauJensen: I have Java so bad - We mock PHP for being inconsistent, but Java is right up there with PHP

17:15 s/have/hate

17:16 bozhidar: hmm

17:16 I looked at the sources

17:16 even though the setters take Object

17:16 LauJensen: I iterate through some components, calling setBackground on them. Then I notice 2 of them arent affected. Google the first one and the answer comes up that under certain LookAndFeels its not guaranteed the calls to setBackground would be respected..?!

17:16 bozhidar: instead of String

17:16 they are internally cast to String

17:16 and parsed the same way

17:17 but this way you can pass something of type AC(which I've never seen so far)

17:18 this seems like a rather poor design decision - an overloaded method would have been easier to grasp

17:19 well, Swing is a pretty nasty story

17:19 LauJensen: bozhidar: And when the entire panel doesn't react to resizes at all, thats miglayout I have to blame?

17:19 bozhidar: there is simply no way to do everything the same way on totally different platforms

17:19 I don't if you're aware of the performance issues when you're using GTK+ L&F

17:20 when you start resizing the frame long enough

17:20 the JVM dies

17:21 if the panel is in another container - you'll have to configure it's layout manager as well

17:21 LauJensen: the panel is just in a jframe

17:21 bozhidar: if it's a top level panel(one contained directly in a JFrame) only setResizable(false) on the frame

17:21 might be causing something like that

17:22 jkkramer: is there a way to use 1.3 contrib complete with lein? [org.clojure.contrib/complete "1.3.0-SNAPSHOT"] doesn't seem to work

17:22 individual modules work

17:25 bozhidar: LauJensen: I think that JFrame used by default BoxLayout which would make the panel grow/shrink, but you can explicitly set some layout just to be on the safe side

17:25 I personally don't set manually the layout for my frames

17:25 LauJensen: yea, I'll have to import miglayout myself it seems

17:25 bozhidar: I just add one panel to each frame and set the panel as the frame's content pane

17:26 LauJensen: so you dont add it ?

17:27 bozhidar: I use code like this setContentPane(topPanel);

17:27 LauJensen: yea me too, no change

17:28 bozhidar: maybe you've set some size restrictions on the frame?

17:28 LauJensen: Not that I know

17:30 of

17:31 bozhidar: hmm, I can't think of anything else

17:31 maybe something is wrong with the wrapper

17:32 LauJensen: Either Im not getting it, or its half baked

17:39 bozhidar: both the panel and JFrame have (.setLayout "fill" "[grow]" "[grow]") and still do not react to resizes

17:40 (and the jtabbedpane)

17:41 bozhidar: LauJensen: AFAIK JFrame's setLayout method expects a layout object

17:41 LauJensen: I get no errors

17:43 bozhidar: it should be (doto (JFrame.) (.setLayout (MigLayout. "wrap 1" "[grow]" "[grow]")))

17:43 LauJensen: why?

17:43 clojurebot: http://clojure.org/rationale

17:43 LauJensen: hehe

17:44 bozhidar: because according to the JFrame javadoc the method's signature is public void setLayout(LayoutManager manager)

17:45 it puzzles me that your invocation is not causing an error

17:45 LauJensen: (MigLayout. "wrap 1" ...) returns a manager right?

17:46 oh

17:46 bozhidar: yes

17:46 LauJensen: I wrote it wrong here

17:46 sorry

17:46 (.setLayout (MigLayout. "wrap 1" "[grow]" "[grow]"))

17:46 I wonder if constructing a JPanel with a Layout as its argument is similar to setting the layout later

17:47 bozhidar: it's the same

17:47 just a matter of taste most of the time

17:47 LauJensen: it is - but I cant be sure with all of these other inconsistencies

18:16 Raynes: chouser: Pingzors.

21:03 danlarkin: interesting tidbit, which I suppose makes sense when you take the time to think about it: the datastructure returned by (promise) mutates when delivered upon

21:03 so putting it inside a reference type is meaningless

21:04 hiredman: how so?

21:04 a promise is a promise delivered or not, dereferencing blocks until it has been delivered to

21:04 danlarkin: ,(let [a (ref {:foo (promise)})] (deliver (:foo @a) :ok) a)

21:04 clojurebot: #<Ref@1ffa176: {:foo #<core$promise$reify__5534@1805bdb: :ok>}>

21:05 danlarkin: yeah I don't mean derefing the promise, but that it gets mutated inside a reference type

21:05 see I don't have a dosync there

21:06 hiredman: mutated how? if you ever look inside the promise you will always find the same value

21:07 danlarkin: it's mutated in the sense that first it was outstanding and then it was delivered

21:08 and since it's contained inside a ref in my example, it just exhibits different behavior than when persistent datastructures are in a ref

21:08 its behavior is obviously correct, it just tripped me up for a minute

21:08 hiredman: if from an observer's perspective something cannot be observered to changed, is mutable?

21:09 danlarkin: I don't get what you're driving at

21:09 hiredman: danlarkin: you wouldn't happen to know how to write applescript that takes input from quicksilver would you?

21:10 danlarkin: I do not. I tried to write a QS plugin once and found it much easier to just make it a shell script

21:11 hiredman: ok

21:14 danlarkin: so as a corollary to my discovery, I have a question. Is there a way, other than promise/deliver, to block on some arbitrary, programmer defined thing (ie not I/O)

21:14 (while @foo) or something is not true blocking, since it will just poll it every nanosecond or something

21:20 hiredman: there are a few blocking queues in j.u.concurrent

21:20 and there are always Locks

21:23 danlarkin: locks to the rescue! who'd've thought

21:24 ataggart: does awaiting on an agent qualify?

21:25 danlarkin: interesting thought, but await only blocks on actions sent from the thread on which you await

21:25 ataggart: ah right

21:30 can anyone think of an environmental issue that would cause a binding to not be applied? There are 3 tests in contrib which are failing, but only do so on the build machine.

Logging service provided by n01se.net