#clojure log - Sep 04 2008

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

0:00 Myoma: by the way I don't intend to hack it so maybe this is slightly off topic ..

0:01 arohner: one thing that helped me was to read the stack trace

0:02 do something that will cause a stack trace, and then read the stack trace

0:02 i.e. (foo 1)

0:02 and then start at something that looks interesting, like Repl.main or eval()

0:03 and then go over to Repl.java and follow the code

10:40 Chouser: ah! I just now got the point about a partial refer of 'clojure and then a later full default refer.

10:52 rhickey: Chouser: ok, so now what about defns?

10:53 Chouser: makes sense to me

10:53 rhickey: I can work with your patch as-is, but I think I'm locked out of checkins on SF today

10:54 Chouser: gah

10:54 I didn't hear a whole lot of objection to moving the repo to google code.

10:54 rhickey: this kind of thing has been pretty rare on SF, I wonder how Google Code compares?

10:55 Chouser: that is one of the benefits of a distributed vcs (git or hg instead of svn) -- the "central" server isn't all that important anymore.

10:55 rhickey: http://code.google.com/p/clojure/ is in the wings

10:56 Chouser: I'm not sure that's true - where do I put my stuff so you can get it?

10:56 Chouser: Since I'm using git-svn, I have a complete history of the clojure repo locally. I could look into importing that to google if you want.

10:56 rhickey: where's the canonic version, etc

10:57 Chouser: sure, there's still value in a generally available canonical server. But routing around damage is much easier.

10:57 rhickey: Chouser: I'll have to move to EPL before I put it up

10:57 Chouser: oh, that's right. ok.

10:58 rhickey: Chouser: I think DVCS is neat, but maybe not a good fit for the benevolent dictator model :)

11:02 Chouser: heh, well, I don't think that matters much. We're all using dvcs anyway, even if you aren't.

11:02 rhickey: right

11:02 Chouser: ok maybe not "all".

11:04 but with dvcs if you're in a situation like this where the central server's down, you can continue to do checkins, and all your data and meta-data can be synced up later -- either with the same central server or a new one (in the case of the old one going away forever).

11:04 of course the tool integration is not as deep yet.

11:05 You'll be using dvcs eventually, so I don't really feel I need to convince you right now. :-)

11:06 * Chouser mutters something that sounds like "revision control, you're doing it wrong"

11:06 rhickey: Chouser: I understand, but not having that forces a 'working head' mentality that multiple local working copies doesn't, plus far less merging (i.e. none for me)

11:07 I've used Mercurial for personal things, much more of a hassle than SVN built into IntelliJ

11:07 cemerick: I guess I'll have to get over my google code aversion issues ;-)

11:08 rhickey: I routinely work on Clojure on 3 different machines in different locations, and when I move I need to check in - the result is lots of working checkins

11:09 cemerick: what are your issues with google code?

11:10 cemerick: rhickey: nothing substantive, really. I think we had this discussion before, though. I don't trust google on a variety of levels, but that's my problem more than anything else.

11:11 rhickey: I'd love to hear from someone re: uptime etc

11:11 cemerick: I don't actually think google code provides much more than svn + trac (if you were to host such things on your own, although I'm sure there are trac hosts floating about). Given their resources, I was pretty disappointed in it.

11:12 Chouser: I'm happy to know that there are various dvcs copies of the clojure repo around -- my own, github, hg mirrors, etc. If google (or sf.net for that matter) loses all the data, we'll be able to reconstitute it with very little fuss.

11:13 rhickey: Right now Clojure doesn't even have trac, and I'm not interested in hosting anything. So the issue tracking is something I'm looking for from google code

11:13 Chouser: really? You're giving in on issue tracking?

11:13 rhickey: Chouser: I fully acknowledge the superiority of DVCS in the large

11:13 Chouser: I didn't say anyone else could enter issues

11:13 Chouser: hehe

11:14 rhickey: Well, I'm still trying for 'no known bugs'

11:14 * Chouser can hear the stress bearing down like a freight train.

11:15 cemerick: rhickey: personally, I'd much prefer something like assembla, but this is your baby :-)

11:15 rhickey: But I keep a todo list that's not web based and is private - keeping it online will let people know where things are going

11:20 albino: rhickey: the reason you don't want an issue tracker is for the perception of no known bugs?

11:21 rhickey: albino: no, I fix bugs as they are reported, rather than put them on a list

11:22 cemerick: assembla looks neat, have you used them?

11:23 cemerick: rhickey: only in passing -- we host all of our stuff in-house (our svn and project mgmt stuff reside on the only server we don't lease or have colo'd somewhere). Friends of mine have used it heavily, and have very good things to say.

11:24 I presume you'd be able to get them to set you up with a full-featured account, given that it's an open source project.

11:29 ozzilee: Does anyone know if it would be possible for multiple versions of a java library to be loaded at the same time, cordoned off in different namespaces perhaps?

11:30 rhickey: cemerick: I didn't see that offered, but did see this: http://blog.assembla.com/assemblablog/tabid/12618/bid/5430/An-explosion-ate-our-servers-how-cool-is-that.aspx

11:30 cemerick: ozzilee: you would need to use different classloaders -- same as in regular java

11:31 ozzilee: cemerick: Hmm. Ok thanks.

11:32 cemerick: rhickey: hrm, unapologetically in startup mode, I suppose

11:32 rhickey: cemerick: right

12:17 jgracin: default target jvm version on Java 6 is 1.4, right?

12:17 in javac, I mean

12:19 nope.

15:17 rhickey: the documentation on reader forms claims that symbols beginning or ending with : are reserved by Clojure. Surely, you didn't mean keyword literals, did you? Am I understanding something incorrectly?

15:18 rhickey: if it begins with : it's not a symbol. "reserved by Clojure" means its interpretation is reserved, i.e. it can't be used generally as a symbol

15:19 jgracin: oh, ok.

15:20 what is the equivalent of keywordp in CL? how do I check whether something is a keyword?

15:21 rhickey: keyword?

15:21 Chouser: (keyword? :foo) ==> true

15:21 jgracin: aieee! :-) how stupid can I get...

15:21 thanks

15:22 rhickey: find-doc is your friend: (find-doc "keyword")

15:22 jgracin: cool! I didn't know about that one.

15:29 wlr: will it always be the case that recompiling anything (not just macro(s)) in a :use'd ns requires recompiling a :use'ing ns for the :use'ing ns to see the new effects?

15:31 rhickey: wlr: depends on what you are trying to 'see', new definitions of existing fns are seen without any recompiling

15:31 getting refers to new names requires re-using (probably by just re-evaluating the defns form)

15:33 wlr: rhickey: that's what i thought but i swear i had to do a recompile after having changed a function in a a:use'd ns.

15:34 rhickey: wlr: shouldn't be

16:22 wlr: rhickey: just for the record, i mistakenly compared old output with (duh!) old output leading to the faulty conclusion that the change in the :use'd ns wasn't visible.

16:22 embarassing!

16:22 rhickey: wlr: np, glad it works

18:17 rapido: ping

20:04 joubert: Hi,

20:04 rhickey: hi

20:06 joubert: I'm trying to determine whether the parameter to a function is a map.

20:07 There is (vector?) but I am unable to find an equivalent for maps

20:07 (according to the API docs anyway)

20:07 not sure whether it is in the code base?

20:08 I guess I should look, thinking about it now

20:08 hmmm, nevermind

20:08 rhickey: (map? m)

20:09 joubert: thanks, Rich. what about a predicate for lists?

20:11 rhickey: (instance? clojure.lang.IPersistentList x)

20:11 joubert: ok, great.

20:19 btw, I see that (map?) is on the main api doc page, but not on the page for data structures where the other map forms are documented.

20:20 rhickey: not every function will appear in the descriptive docs

20:21 joubert: oh ok; but will in api main?

20:21 rhickey: there's also find-doc at the repl to help you find things that are only in the SVN and not in the API reference

20:21 API docs document the release

22:10 arohner: local variables are not vars, right?

22:21 DrewR: You mean those defined by let?

22:21 arohner: yes

22:21 DrewR: I think they're Vars.

22:21 They're immutable.

22:22 I'm not sure how to tell though.

22:22 user> (let [x 3] (class x))

22:22 java.lang.Integer

22:23 arohner: well, (let [foo 5] (var foo)) doesn't complain

22:30 arbscht: I don't think they're Vars

22:30 local bindings are not variable, and Vars are mutable

22:33 arohner: that form should complain. maybe you have defined foo globally previously?

22:34 arohner: arbscht: ah, you're right

22:34 I had previously defined it

22:34 after unmapping, Caused by: java.lang.Exception: Unable to resolve var: foo in this context

22:34 DrewR: Vars are mutable?

22:35 How so?

22:35 arohner: using set!

22:36 arbscht: Vars exist for the purpose of thread-local mutability :)

22:37 DrewR: Ah, you're right. I just never use them that way.

23:11 mbeau: just wondering where I should look in the source code to understand how the '.' operator works?

23:19 arohner: mbeau: this is kind of throwing you off the deep end, but clojure/src/jvm/clojure/lang/Compiler.java

23:19 search for DOT

23:19 a lot of Compiler.java has to do with parsing the source, and it's kind of hard to escape that

23:21 mbeau: arohner: thanks, that's where I was looking around, but as you warn it is pretty tough to see what is going on

23:23 ok, actually I see that HostExpr is a good place to start

23:28 really, the thing I'm curious about is how the (.foo ...) syntax is implemented

23:35 yikes

23:38 arohner: mbeau: ?

Logging service provided by n01se.net