[00:00] creationix has joined the channel
[00:00] isaacs: so, problem with that is that we all break ModuleIdentifiers#2
[00:01] kriskowal: yeh, that needs to be fixed in the next rev of the spec
[00:01] kriskowal: let's assume we've fixed that.
[00:01] isaacs: but the test could be something like this:
[00:02] isaacs: a.js: exports.moduleId = module.id;   b.js: var a = require("./a");  assert(a === require(a.moduleId));
[00:02] isaacs: at least, the test for module.id
[00:03] kriskowal: that's a start
[00:03] isaacs: again, i don't exactly see why it matters what module.id looks like.
[00:04] isaacs: is there an actual use case where having a module identifier starting with a / is a bad idea? or where module.id needs to be something that doesn't start with a /?  because i've got a few use cases where having module.id NOT start with / causes all kinds of problems.
[00:04] kriskowal: yeah, i know what you're talking about
[00:06] kriskowal: the big deal is that we're both trying to prevent the same module from being loaded twice.
[00:06] isaacs: right... and this does that...
[00:06] bpot has joined the channel
[00:06] kriskowal: by different aliases.  i consider the top-level identifier canonical, and you the file name.  we're both right depending on how you look at it.
[00:07] isaacs: well, since require.paths is changeable, that's not a safe assumption.
[00:07] isaacs: and your suggestion would be to lock it down, i'm sure
[00:07] kriskowal: my solution prevents two implementations of the same specification from being loaded even if they exist in multiple modules, in the face of the assumption that require.paths is consistent
[00:07] kriskowal: multiple paths, i mean
[00:07] kriskowal: which is important for narwhal; not as much in node since it's not designed for multiple engines.
[00:08] isaacs: oh, you see, that's exactly the use case I need to *enable*
[00:09] isaacs: i can't trust that require("foo") alwasy refers to the same thing, and if there are multiple foo.js files, and hte path has changed, then i need to ensure that you'll get a new foo module, since you might actually depend on a different version that the last guy who pulled it in.
[00:09] kriskowal: fiddling with require.paths after you've loaded any significant modules can only lead to pain
[00:09] isaacs: kriskowal: not if you do what i did.
[00:09] isaacs: kriskowal: check it out.  it works.  it's stable.
[00:09] isaacs: better yet, it's simple
[00:09] kriskowal: i'm dead certain that with use of relative identifiers, i would be able to load a module that's not supposed to be.
[00:10] isaacs: i will buy you a beer if you can do that.
[00:10] kriskowal: by subverting the version that would have been discovered by searching require.paths with a top-level
[00:10] isaacs: kriskowal: no, srsly.
[00:10] isaacs: do that.
[00:10] isaacs: show me the code, and i'll buy you a beer.  we've got 135 witnesses ;)
[00:10] kriskowal: explain how you prevent that.
[00:11] kriskowal: do you scan require.paths to verify equivalence?
[00:11] jasonw22 has joined the channel
[00:12] isaacs: http://github.com/isaacs/node/blob/module-refactor/lib/module.js#L70-119
[00:12] kriskowal: if you're doing something like that, you could easily normalize the path to an id and have the same guarantee
[00:12] isaacs: kriskowal: yeah
[00:12] kriskowal: while internally using the path as the cache key
[00:12] kriskowal: but providing the id as top-level
[00:12] kriskowal: no compromise
[00:12] jed has joined the channel
[00:13] isaacs: i use both the cache and the from-the-root id as cache keys, but i only return the stem cache-hit if the paths list is just a single location (that is, in the case of / or . requires)
[00:13] isaacs: er, both the *filename
[00:13] isaacs: oh, no, i don't scan require.paths to see if it's changed.
[00:14] isaacs: i just only cache based on the filename if there's mor ethan one search target.
[00:19] kriskowal: at this point, i'm convinced that it is possible to guarantee correctness regardless of the shape of module.id
[00:19] isaacs: um... yay?
[00:19] kriskowal: yeah, not necessarily.
[00:20] kriskowal: good enough for you; i'll withdraw from the argument since we've broken it down to a subjective decision for the purpose of module searching.
[00:20] isaacs: what's un-yay about that?
[00:20] kriskowal: i would *like* a stronger guarantee about the value of module.id across engines, but in the context of node, i have not got a compelling argument
[00:21] isaacs: i think the spec should specifically not guarantee the shape of module.id, only its behavior.
[00:21] isaacs: which it does.
[00:21] kriskowal: it does if you read it closely, but there are other problems if you read it closely, so it's irrelevant
[00:22] kriskowal: it's ambiguous at worst
[00:22] isaacs: in any event, this was a funner bit of code analysis fisticuffs than i've had in a while.  thanks, kris
[00:22] kriskowal: i have gained some insights as well, thanks.
[00:23] mjr_: _ry: sometimes when I do make test, simple/test-http exhibits the same error I get on my http proxy, but not all the time
[00:24] mjr_: https://gist.github.com/d3fdc389075bea59da61
[00:24] mjr_: There's some sort of timing issue in there.
[00:25] CodeOfficer has joined the channel
[00:26] mjr_: _ry: also, the fix for node-repl that you pointed me to earlier (missing default prompt) is here: http://github.com/mranney/node/commit/c13aede7f2c8ecdf2962edeb957c74eae03d7534
[00:28] mrjjwright has joined the channel
[00:29] creationix has joined the channel
[00:31] sveisvei has joined the channel
[00:38] fizx has joined the channel
[00:45] mravaux_ has joined the channel
[00:59] tlrobinson_ has joined the channel
[01:06] justinlilly has joined the channel
[01:10] ditesh|cassini has joined the channel
[01:16] tlrobinson_ has joined the channel
[01:16] derferman has joined the channel
[01:28] ditesh|cassini has joined the channel
[01:34] charlesjolley_ has joined the channel
[01:53] tk has joined the channel
[01:59] joshbuddy has joined the channel
[01:59] joshbuddy has joined the channel
[02:05] mcarter has joined the channel
[02:42] nsm has joined the channel
[02:44] mrjjwright: kriskowal: trying to upgrade restler to latest node and getting an error at the line require("uri") which pulls in the narwhal uri lib.  
[02:44] deanlandolt: mrjjwright: IIRC narwhal's uri was deprecated
[02:44] mrjjwright: oh ok
[02:44] mrjjwright: what should be used instead
[02:45] mrjjwright: anybody know
[02:45] deanlandolt: what fork are you using?
[02:45] mrjjwright: the buitin url lib from node?
[02:45] deanlandolt: in kriskowal's master he's been moving things into the narwhal/ namespace
[02:46] mrjjwright: i think I just use the built-in url lib from node, this probably wasn't out or stable when dan wrote restler
[02:46] mfeiri_ has joined the channel
[02:46] isaacs: mrjjwright: what's it using uri for?
[02:46] deanlandolt: err, wait, i think i'm confused...i don't think it's deprecated, but it's certainly not in kriskowal's master where it used to be
[02:46] isaacs: mrjjwright: i mean, what is it doing with it?
[02:46] mrjjwright: i am not using narwhal, just updating restler to node 0.1.90
[02:46] isaacs: deanlandolt: he means the lib/uri.js that used to be in node
[02:47] mrjjwright: to parse urls
[02:47] isaacs: deanlandolt: and was a port of narwhal's lib uri
[02:47] mrjjwright: oh right, ok
[02:47] deanlandolt: mrjjwright: hmm...he said he's pulling in the narwhal uri lib
[02:47] deanlandolt: err ** isaacs
[02:47] mrjjwright: no not the narwhal  uri lib, isaacs is right, i was confused
[02:48] deanlandolt: okay, nvm :)
[02:48] isaacs: deanlandolt: right, it was "the narwhal uri lib" because it was pulled from narwhal into nodejs
[02:48] isaacs: 
ACTION ported narwhal's uri lib, got frustrated with its  bugs, and rewrote it as lib/url.js

[02:48] deanlandolt: heh, well that makes sense
[02:48] mrjjwright: it's the old node port and it's not being used for much except to parse the url, I think everything should be in node's "url" now.
[02:48] deanlandolt: cool
[02:48] mrjjwright: do you think isaacs?
[02:48] isaacs: mrjjwright: "parse" is such a... vague word.
[02:49] isaacs: mrjjwright: if you just s/uri/url/g you'll probably have to also rename whichever fields it's using
[02:49] isaacs: url models teh parsed data object after the location object inthe browser.
[02:49] isaacs: arguably not a great convention, but that's why we have setTimeout and setInterval, etc.
[02:50] isaacs: it's *A* convention, and node uses it in a bunch of other places.
[02:50] mrjjwright: k
[02:51] isaacs: mrjjwright: so, like, if it's using parsed.authority, you'll change that to (i think?) parsed.host
[02:51] isaacs: etc.
[02:52] mrjjwright: isaacs: is there any parsing domain within node you don't know something about?
[02:52] isaacs: hehe
[02:53] isaacs: parsers are fun :)
[02:53] JimBastard_ has joined the channel
[02:53] JimBastard_: is there a good js markdown module for node yet?
[02:53] isaacs: and it seems like the internet is just FULL of random crazy ways to encode data.
[02:53] JimBastard_: im looking at http://github.com/visionmedia/markdown.js but that seems to be in C? im confused
[02:54] deanlandolt: JimBastard_: showdown's packaged as a commonjs module...
[02:54] JimBastard_: i went a little crazy this weekend and wrote a book, need to parse that shit into html
[02:55] dandean has joined the channel
[02:55] JimBastard_: http://github.com/kriskowal/markdown ?
[02:55] deanlandolt: http://github.com/gmosx/markdown/blob/master/lib/markdown.js
[02:55] deanlandolt: err, the kriskowal would work too
[02:56] deanlandolt: but yeah, i don't think they should be calling it markdown...gruber's a dbag w/ his license and explicitly says you can't do that
[02:56] kriskowal: yeh, JimBastard_ let me know if it doesn't work out for you.
[02:56] JimBastard_: looks like the same code
[02:56] JimBastard_: awesome thanks kriskowal 
[02:56] isaacs: JimBastard: it's not so much that it's in C as that it doesn't actually include what it uses...
[02:56] JimBastard_: im adding a markdown protocol for hook.io
[02:56] JimBastard_: trying now
[02:56] kriskowal: there's a bin/markdown in that package, which isn't applicable to node, alas
[02:57] kriskowal: it might if you massage it.
[02:57] isaacs: you know, if you added a version to the package.json, then i'll bet npm could install that thing and even make the bin work.
[02:58] isaacs: oh, nope, nvm
[02:58] isaacs: system.stdin.read()
[02:58] JimBastard_: im using gitsubmodules for my package management right now
[02:59] JimBastard_: then the hook.io protocols import from the /lib/ folder
[02:59] JimBastard_: so its not the most elegant, but it works
[03:00] kriskowal: yeh.  i have not yet achieved harmony with that.  i have started to put together a narwhal compatibility layer though.
[03:01] kriskowal: pretty exciting to converge, isaacs.  my last gist was on raw node, using parts shared with the narwhal+node stuff
[03:02] isaacs: neat
[03:02] JimBastard_: you guys gonna be at jsconf?
[03:02] kriskowal: i might be able to bridge the gap with a .node thingie
[03:02] kriskowal: i will be.
[03:03] JimBastard_: huzaah! ill be at the scurvyconf party
[03:03] deanlandolt: kriskowal: was the faustian bargain you mentioned the commonjs panel discussion?
[03:03] kriskowal: part of it
[03:03] kriskowal: i also offered my firstborn's soul
[03:03] deanlandolt: cool...i saw that and made sure my track b pintura talk was right after it :)
[03:04] kriskowal: looking forward to it
[03:07] isaacs has joined the channel
[03:08] fizx has joined the channel
[03:23] tobyvs has joined the channel
[03:25] JimBastard_: aight kriskowal its all hooked up as a hook.io protocol now. im gonna pipe in my MD files and see how they render
[03:30] tmpvar has joined the channel
[03:40] JimBastard_: man node-persist is really starting to act up as i add more stuff 
[03:47] tmpvar: lol
[03:47] tmpvar: i thought you were doing docs
[03:50] kenneth_reitz has joined the channel
[03:51] PyroPete1 has joined the channel
[03:53] JimBastard_: whats the fs command for appending to a file?
[03:53] JimBastard_: i added two more protocols tmpvar 
[03:54] JimBastard_: running the doc generator on startup causes node-persitence to bug on race con
[03:54] JimBastard_: waiting for felix to release dirty >.<
[03:54] tmpvar: ahh
[03:54] tmpvar: hehe
[03:54] JimBastard_: do you know how to append to a file>
[03:59] tmpvar: JimBastard, not offhand
[04:07] joshbuddy has joined the channel
[04:08] JimBastard_: anyone else using http://github.com/creationix/step ?
[04:08] tlrobinson_ has joined the channel
[04:15] kenneth_reitz has joined the channel
[04:26] aho has joined the channel
[04:26] malkomalko has joined the channel
[04:31] hassox has joined the channel
[04:41] binary42 has joined the channel
[04:44] rtomayko has joined the channel
[04:50] fizx has joined the channel
[04:51] okito has joined the channel
[04:57] JimBastard_: ugh whats the best approach for reading files in a loop
[04:57] JimBastard_: im having some control flow issues
[04:58] gfxmonk has joined the channel
[05:12] isaacs has joined the channel
[05:20] brapse has joined the channel
[05:33] keeto has joined the channel
[05:36] xla has joined the channel
[05:52] charlesjolley has joined the channel
[05:56] dgathright has joined the channel
[06:18] derferman has joined the channel
[06:20] nsm has joined the channel
[06:28] JimBastard_ has joined the channel
[06:34] micheil: _ry: are the man docs for node read much, or do you think most people reference the website api docs?
[06:35] kriskowal has joined the channel
[06:35] micheil: (eg, would it make sense to move the documentation towards being mainly web based, and then make the manpage just describe the bin/node commandline options?)
[06:36] micheil: meaning that we wouldn't need to worry about how we compile both
[06:36] micheil: it'd also allow as to take the web docs further, as for as breaking it into say, multiple pages for the different modules and such
[06:36] fizx has joined the channel
[06:37] fizx has joined the channel
[06:40] piranha has joined the channel
[06:50] tisba has joined the channel
[06:52] keeto_ has joined the channel
[06:52] markwubben has joined the channel
[06:53] alexandere has joined the channel
[06:59] felixge has joined the channel
[07:12] kriskowal: isaacs, i presume that not searching require.paths for relative ids is deliberate.
[07:17] zimbatm has joined the channel
[07:21] JimBastard_: kriskowal: your markdown lib seems to be behaving
[07:21] kriskowal: oh, good.
[07:21] JimBastard_: :p
[07:21] kriskowal: i wouldn't know what to do if it didn't, honestly
[07:21] kriskowal: it's a port
[07:21] JimBastard_: do you have any idea how to generate anchors in markdown?
[07:21] kriskowal: hm.  you prolly can use html
[07:21] JimBastard_: inline html then?
[07:22] kriskowal: i presume there's nothing special
[07:22] kriskowal: if the language i wrote in perl 10 years ago had taken off, it would have been J>name :P
[07:23] kriskowal: i should prolly write a swil engine in js :/
[07:23] JimBastard_: yeah that works awesome
[07:24] JimBastard_: continuous documentation for the win
[07:24] kriskowal: yeh.  i'm also working on a doc system
[07:24] kriskowal: as is ashb.  seems like it's what everyone's up to these days.
[07:25] micheil: docs system, where?
[07:25] kriskowal: http://github.com/kriskowal/whatsupdoc
[07:26] kriskowal: it's alpha.  generates a JSON tree from jsdoc/markdownish for all installed packages.
[07:26] kriskowal: working on a formatter
[07:27] kriskowal: when all is said and done, it'll work on narwhal, node, and narwhal+node
[07:27] teemow has joined the channel
[07:32] kjeldahl has joined the channel
[07:34] brapse has joined the channel
[07:34] CIA-77: node: 03Matt Ranney 07master * rb744104 10/ (lib/repl.js test/simple/test-repl.js): 
[07:34] CIA-77: node: REPL can be run from multiple different streams.
[07:34] CIA-77: node: e.g. from UNIX sockets with socat. - http://bit.ly/9xkwOi
[07:35] mape: Hmm I read somewhere node.js was faster then java.. Anyone know what article/text that was? (trying to get the boss to look at using node)
[07:35] _ry: micheil: i want to keep the man page 
[07:37] _ry: mjr_: you should write some docs for the repl :)
[07:44] micheil: _ry: as the full documentation?
[07:49] kriskowal: isaacs: clone this http://gist.github.com/363360 and run it
[07:51] kriskowal: one nice attribute of narwhal's loader is that we can overlay each directory in require.paths onto a single directory for browsers or performance and be guaranteed equivalent behavior.
[07:51] kriskowal: it works because we resolve relative identifiers on the top-id of the requiring module, not the path of the requiring module.
[07:53] kriskowal: constructing a unified overlay of modules from each of require.paths is one technique that could also be used in node to bundle modules for couchdb views; in which case you'd need to resolve on top-ids too to guarantee consistent behavior
[07:53] kriskowal: same for module transports.
[08:01] mfeiri_ has joined the channel
[08:02] mravaux has joined the channel
[08:03] micheil: _ry: the Freelist thing in net.js is freaking epic.
[08:03] romainhuet has joined the channel
[08:03] micheil: _ry: that should really be made into a separate module
[08:03] felixge has joined the channel
[08:04] felixge has joined the channel
[08:04] kjeldahl has joined the channel
[08:04] micheil: so that we can use it in the apps that we build, because it is really useful
[08:06] micheil: _ry: I think it'd be better if the modules within the net module were broken up a little, eg, have a FreeList.js, Stream.js etc.
[08:09] B1scuitsLap has joined the channel
[08:29] micheil: woah. wtf did sys.log come from?
[08:31] BiscuitsLap has joined the channel
[08:34] javajunky has joined the channel
[08:35] N` has joined the channel
[08:38] xla has joined the channel
[08:41] mape: Been there a while?
[08:53] tbassetto has joined the channel
[08:56] teemow has left the channel
[08:57] hellp has joined the channel
[09:00] botanicus has joined the channel
[09:01] micheil: hmm.. is 8 bytes == eight utf-8 encoded characters?
[09:01] Gruni has joined the channel
[09:02] piranha: micheil: not always
[09:02] micheil: :/
[09:03] piranha: utf-8 is encoding with variable character length
[09:03] piranha: so if it's latin alphabet, then yep, 8 chars will take 8 byte
[09:03] piranha: *bytes
[09:03] piranha: but if it's cyrillic, then 8 chars will take 16 bytes
[09:05] javajunky: might take 16 bytes ;) 
[09:06] javajunky: micheil: if you're interested, I ported a chunk of Javascript code an age ago for encoding and decode utf8 streams: http://ciaranj.blogspot.com/2007/11/utf8-characters-encoding-in-javascript.html
[09:07] TomY has joined the channel
[09:07] micheil: javajunky: node has that.
[09:07] micheil: it's in the C++ api
[09:07] javajunky: micheil: yah I know, but I figured you might be a Js guy and that might help your understanding ;)
[09:07] piranha: javajunky: I'd say with cyrillic you can be sure it will take 16 bytes :)
[09:08] javajunky: piranha: yeah on reflection I'd say you're right and I'm wrong ;)
[09:08] javajunky: (Stupid reflexes)
[09:08] piranha: :)
[09:10] micheil: actually, just received information that it'll be 8 bytes in UTF8 from the ascii character range
[09:12] micheil: anyone know about getting a byte array in node?
[09:18] statim has left the channel
[09:35] charlesjolley has joined the channel
[09:40] javajunky: so bascially its an ascii stream not a utf-8 stream ? 
[09:42] javajunky: (but to answer your question, this file http://github.com/ciaranj/flickrnode/blob/master/lib/md5.js  does that sort of manipulation, so might be some usage there) 
[09:44] micheil: javajunky: it's a stream of utf8 characters, although, it can only be the ascii characters in it
[09:44] derbumi has joined the channel
[09:45] javajunky: micheil: apologise for my stupidity, but that surely means its also a stream of 'ascii characters' .. the fact its utf8 is largely irrelevant (unless there's a byte order marker at the start (which is optional/pointless in utf8 anyway) )
[09:46] micheil: javajunky: it's as per the WebSocket protocol.
[09:46] javajunky: ah, k, so there will be ffff and 0000 in there (which also aren't part of utf8 just to mess with you) 
[09:46] micheil: so, it's a UTF8 stream, however, for this particular segment, it's only the ascii characters.
[09:46] micheil: but it doesn't matter, I used a temp buffer
[09:52] derbumi_ has joined the channel
[10:01] derbumi_ has joined the channel
[10:06] psynaptic has joined the channel
[10:15] kjeldahl has joined the channel
[10:22] chakrit has joined the channel
[10:22] chakrit has left the channel
[10:23] sveisvei has joined the channel
[10:24] hellp has joined the channel
[10:34] derbumi has joined the channel
[10:35] alexandere has joined the channel
[10:42] keeto_ has joined the channel
[10:52] xla has joined the channel
[10:56] maritz has joined the channel
[11:02] ssteinerX has joined the channel
[11:16] mrjjwright has joined the channel
[11:23] derferman: I am currently using JSON.parse to parse data sent via my client. However, if the data isn't valid JSON, JSON.parse throws an error and brings down node
[11:23] derferman: how can I get around that?
[11:23] derferman: Does the JSON parser have error checking?
[11:23] jbrantly: try/catch?
[11:24] ashb: s/\?//
[11:25] derferman: d'oh, forgot that javascript has try / catch
[11:30] derbumi has joined the channel
[11:37] andrewhampton has joined the channel
[12:06] javajunky: derferman, jSON.parse is strict (according to EC5 ) 
[12:15] confounds has joined the channel
[12:26] trochala has joined the channel
[12:30] trochala has joined the channel
[12:41] toabi has joined the channel
[12:43] javajunky: anyone know if there's a node.js base6 decode call available to me in a lib ? .. (I've got base64 decode functions, but wondering if there's something provided by node/v8 ?
[12:51] jbrantly: there's atob(), not sure if its supported by v8
[12:58] teemow has joined the channel
[13:02] micheil: is there a way to retreive the localhost and localport that a server is listening on?
[13:05] dnolen has joined the channel
[13:06] javajunky: yeah pretty sure you can get that fromt he socket object (sys.p it to see whats available) ?
[13:07] micheil: nup
[13:07] micheil: net.Server.prototype.address
[13:16] jherdman has joined the channel
[13:22] micheil: _ry: you about?
[13:22] micheil: _ry: I'm getting errors in two tests running on HEAD
[13:23] cainus_ has joined the channel
[13:23] micheil: _ry: OS X/10.6, errors: test-http-head-request, test-mkdir-rmdir
[13:23] micheil: _ry: and I'll have a patch for you soon on the net.js file
[13:23] micheil: _ry: turns out the callbacks are called back within current scope
[13:34] micheil: actually. doesn't do what I wanted.
[13:34] micheil: nvm
[13:34] micheil: errors still present though
[13:40] keeto has joined the channel
[13:41] davidsklar has joined the channel
[13:51] gf3 has joined the channel
[13:58] JAAulde has joined the channel
[14:07] micheil: damn.. _ry ever tried making a net.Stream and then telnetting it and pressing ctrl+c (killing)
[14:07] micheil: it creates a stack overflow
[14:12] alex-desktop has joined the channel
[14:18] gwoo has joined the channel
[14:20] softdrink has joined the channel
[14:21] quirkey has joined the channel
[14:24] cloudhead has joined the channel
[14:34] creationix has joined the channel
[14:50] TheEnd2012 has joined the channel
[14:52] piranha_ has joined the channel
[14:55] alexiskander has joined the channel
[14:58] _ry: micheil: ping
[14:58] micheil: pong
[14:58] KungFuHamster: SYN
[14:59] micheil: never mind that ctrl+c one.
[14:59] _ry: I can't repeat your ^c
[14:59] _ry: why did you think that?
[14:59] _ry: someone else reported that too
[14:59] micheil: yeah, turned out I'd accidentally called the method I was invoking the same as the net.Stream.ondata method I was overriding
[14:59] Yuffster has joined the channel
[14:59] micheil: it was clearly my bad, too much recursion style error
[15:00] micheil: although, it did report that it was a stack overflow.
[15:00] _ry: hm
[15:00] _ry: maybe i should make ondata more hidden
[15:00] micheil: 
ACTION is deliberately overriding it.

[15:00] _ry: ;)
[15:00] creationix: _ry, I see you patched the dns module to not break with stuff in Object.prototype
[15:00] _ry: creationix: yeah, thanks for figuring that out
[15:00] creationix: while debugging the issue saturday I noticed several un-protected for..in loops as well
[15:01] micheil: _ry: I suppose I could also use addListener("data") because that does the exact same, but oh well
[15:01] creationix: those will break as will if not patched
[15:01] _ry: creationix: oh, where at?
[15:01] charlesjolley has joined the channel
[15:01] micheil: I did see someone having an Object.forEach, which used the patch style
[15:01] creationix: mixed throughout, I think about 6 different places
[15:02] creationix: the problem is that a protected for..in loop is slow
[15:02] creationix: even slower than a hand implemented Object.forEach
[15:02] creationix: the fastest of course if an inlned Object.keys and then a regular c-style for loop over that
[15:02] creationix: but that's really messy
[15:06] creationix: hmm, can't copy paste the textmate search results
[15:06] creationix: http://static.creationix.com/forin.png
[15:06] creationix: here's an image
[15:06] _ry: creationix: well - I wouldn't mind patching those with hasOwnProperty()
[15:06] micheil: http://github.com/creationix/conductor/blob/master/lib/conductor.js#L29-39
[15:06] micheil: was what I was thinking
[15:07] creationix: yep, but _ry already said no to including that ;)
[15:07] micheil: although, that could be refactored
[15:07] micheil: creationix: oh, yeah, you're doing websockets wrong.
[15:07] _ry: micheil: i know that test-http-head-request is broken
[15:07] micheil: okay
[15:08] micheil: the other one that I had broken is fixed after a fresh checkout
[15:08] _ry: it's kind of a big problem and i'm just leaving that there to remind me
[15:08] creationix: micheil, I'm not surprized, I implemented it in about an hour, what's wrong with it
[15:08] micheil: creationix: lots.
[15:09] micheil: eg, incorrectly parsing / handling the initial headers
[15:09] creationix: but it works
[15:09] micheil: it'll work for some clients, but it's not spec compatible
[15:09] micheil: 
ACTION is writing a full websocket implementation to the spec right now

[15:09] creationix: ahh, go for it
[15:09] creationix: I like my minimal one that works for my cases
[15:10] steadicat has joined the channel
[15:11] creationix: _ry, since you don't want to include Object.prototype.forEach with node, what about a local definition that's used to replace all these slow for..in loops
[15:12] creationix: it will make the code cleaner and faster than it's currently
[15:12] _ry: creationix: i don't think you cna make it faster
[15:12] creationix: then for..in, sure I can
[15:13] micheil: yeah, for ... in is slow.
[15:13] creationix: I benhmarked it, I was really surprized how slow for..in is especially once you add the hasOwnProperty check
[15:13] micheil: about 200ms difference / 10000
[15:13] micheil: i think
[15:13] nsm has joined the channel
[15:14] blackdog` has joined the channel
[15:14] creationix: it's because for..in has to look up the prototype chain for all enumerable properties
[15:14] creationix: Object.keys only checks the local object
[15:14] blackdog`: hi there, how do i do multipart in 0.1.90?
[15:15] creationix: hmm, I wonder how Object.keys(obj).forEach performs
[15:15] creationix: didn't bench that one
[15:15] N`: blackdog`: http://github.com/isaacs/multipart-js
[15:15] blackdog`: thanks
[15:16] creationix: 
ACTION does more benchmarking

[15:16] sveisvei has joined the channel
[15:17] blackdog`: N`: will this be the official way of doing it?
[15:18] micheil: probably
[15:18] blackdog`: or rather, is Ryan not providing multipart in the base dist?
[15:18] micheil: _ry: ^^
[15:19] felixge has joined the channel
[15:19] felixge has joined the channel
[15:19] creationix: https://gist.github.com/bdd159b4f8f7c900d939
[15:19] creationix: I think isaacs' code is based on the one pulled from node
[15:19] N`: blackdog`: yes, multipart is no longer in node.js and isaacs has taken the code that was in node.js and made a lib from it
[15:20] blackdog`: ok, thanks
[15:20] creationix: is that the one that felixge is using too (or will use when he updates transload.it)
[15:20] creationix: _ry, see the gist for bench results
[15:20] felixge: creationix: I tried doing that earlier today, still working on it
[15:21] felixge: ryan unfortunately nuked the callbacks for fileStream.write() which makes things more diffcult
[15:21] felixge: plus, the new multipart module does not handle request objects directly anymore
[15:21] felixge: 
ACTION pain

[15:22] creationix: felixge, ouch, I'm sorry to hear that
[15:24] _ry: creationix: interesting
[15:24] _ry: felixge: well, they're not nuked yet
[15:25] creationix: so doing a forEach on the keys isn't that much slower than the raw inlined for over the keys
[15:25] creationix: I think that's a good compromise for node
[15:26] _ry: felixge: it should be pump(part, file)
[15:26] creationix: don't need an external helper function
[15:26] _ry: felixge: which, i admit, is probably rife with errors. but you should aim for that interface.
[15:27] creationix: do we have an official pump function yet?
[15:27] _ry: no
[15:27] creationix: or at least a defacto working example
[15:27] _ry: there is one here
[15:27] _ry: http://wiki.github.com/ry/node/streams
[15:27] creationix: oh yeah, how did I miss that one
[15:28] _ry: one problem is that i'm just not sure where to put it :)
[15:28] felixge: _ry: I don't think part complies to the stream api
[15:28] _ry: felixge: not yet, maybe, but it ought to
[15:28] felixge: 
ACTION hopes he won't end up writing his own multipart lib again

[15:28] felixge: _ry: I'll talk to isaacs
[15:28] joshbuddy has joined the channel
[15:28] felixge: 
ACTION can't wait for 0.2.0 :)

[15:29] _ry: 0.2.0 is going to look very much like what it is now
[15:29] _ry: it's basically bug fixes from here on out
[15:29] weepy has joined the channel
[15:29] weepy: hi im trying to convert some websocket code to 0.1.9
[15:30] creationix: ok, _ry, if it's ok I'll replace the for..in loops with a forEach over the keys and submit a patch
[15:30] creationix: I looked at implementing process.stdout as a sync stream, but it was a little over my head
[15:30] _ry: creationix: there are some places where i use cntinue
[15:31] weepy: I'm thinking socket.write('\u0000' + data + '\uffff', times); isn't working because of the move to Buffer 
[15:31] _ry: creationix: stderr you mean
[15:31] weepy: but I'm not really sure
[15:31] creationix: _ry, yes, sorry
[15:31] creationix: ok, then I'll make inline for loops over the keys then
[15:31] felixge: _ry: right, I mostly mean the time where node will finally have a stable ecosystem of libraries again and I don't have to write everything myself anymore :)
[15:31] creationix: that will keep the current semantics and be the fastest possible too
[15:31] creationix: just a little more code
[15:32] _ry: creationix: okay
[15:32] _ry: creationix: run 'make test-all' to test it
[15:32] creationix: of course, thanks
[15:32] _ry: creationix: no, thank you :)
[15:33] micheil: weepy: when do you need websockets by?
[15:33] creationix: weepy, it's a simple fix, just send the framing bytes as binary with seperate calls to write, or build a buffer and write using that
[15:34] micheil: weepy: I'm busy implementing a spec compliant version against HEAD, so, if you don't need it for a while, yeah
[15:35] micheil: it doesn't do much yet, but will: http://github.com/miksago/node-websocket-server
[15:36] creationix: weepy, also If you want something quick, mine is working today, it's just a little rough on the edges.  I'll fix issues as they arise in real code
[15:40] brapse has joined the channel
[15:45] fictorial has joined the channel
[15:46] mape: Is there any reason to use delete varName; or does the gc do it automagic after a while?
[15:46] fictorial has joined the channel
[15:46] mape: (noticed it used in micheils code)
[15:47] _ry: i've heard it sends v8 into the slow case
[15:47] _ry: i.e. should be avoided
[15:48] mape: Ok
[15:49] creationix: hmm, found one for..in loop that returns on the first iteration, it must be checking to see if an object is empty or not
[15:49] micheil: _ry: should it?
[15:49] creationix: I'll have to benchmark to see if object.keys is still faster in that case
[15:49] micheil: _ry: news to me. I was using it purely to destory the buffers
[15:51] inimino: it's delete on objects that sends V8 into a slow path, I think
[15:51] inimino: on object properties, I mean
[15:51] rednul has joined the channel
[15:52] inimino: delete foo.bar; // makes foo a hash-style object, not a hidden class instance
[15:52] inimino: delete foo; // is actually undefined in the language specification
[15:53] inimino: foo=undefined; // probably better
[15:54] inimino: micheil: ↑
[15:54] micheil: okay
[15:54] micheil: unless someone sets undefined
[15:55] inimino: in which case they get what they deserve :)
[15:55] micheil: okay
[15:55] micheil: what about unset?
[15:55] _ry: micheil: I avoid delete wherever possible
[15:55] inimino: unset?
[15:55] micheil: okay
[15:56] inimino: if you're using an object as a hash, it makes sense to use delete on it
[15:56] inimino: especially if you use for-in over it
[15:56] micheil: _ry: could you expose FreeList?
[15:56] inimino: since delete actually removes the property (rather than just making its value undefined) and there's no other way to do that
[15:56] _ry: micheil: where would we put it?
[15:57] micheil: I dunno
[15:57] tisba_ has joined the channel
[15:57] micheil: make it a separate class
[15:57] micheil: it's useful
[15:57] micheil: module rather
[15:57] micheil: I know it's only small, but it's useful.
[15:57] _ry: hm, okay
[15:57] _ry: micheil: want to do that?
[15:58] micheil: okay.
[15:58] _ry: micheil: it's used in http.js too
[15:58] _ry: it has its own implementation at the moment
[15:59] micheil: yeah
[15:59] tilgovi has joined the channel
[15:59] micheil: I'm still slightly unsure of how to exactly use it, but it looks useful
[16:01] micheil: wait. does FreeList even work?
[16:02] micheil: _ry: this.list.length should always be defined, no?
[16:04] creationix: _ry, in http.js line 270, I see you don't have the hasOwnProperty check till after some other logic
[16:04] creationix: are you expecting some data up in the prototype chain?
[16:04] romainhuet has joined the channel
[16:05] inimino: I think we should really just tell people not to extend Object.prototype or use ES5 .defineProperty if they do
[16:05] inimino: adding hasOwnProperty tests everywhere is ugly, IMHO
[16:05] creationix: inimino, agreed, what I'm doing though it to replace the slow for..in calls with faster for loops over the keys
[16:06] creationix: but as a side effect you lose anything in the prototype chain
[16:06] inimino: is Object.keys.forEach faster in V8 now?
[16:06] creationix: much
[16:06] inimino: oh, carry on then :)
[16:06] inimino: good to know
[16:06] creationix: a for loop over the keys is almost 3 times faster than for..in with hasOwnProperty
[16:06] inimino: what about without?
[16:07] bpot has joined the channel
[16:07] creationix: then only just over 2 times faster
[16:07] inimino: ok
[16:08] creationix: I'm hoping these changes will speed up the code a little, though most these loops are for reading configs and stuff
[16:08] creationix: nothing in the critical path
[16:08] RayMorgan has joined the channel
[16:08] aryounce has joined the channel
[16:09] mjr_: That's really surprising to me that for prop in obj is slower than Object.keys().forEach().
[16:09] mjr_: But it's the good kind of surprise, because I much prefer the latter style.
[16:09] creationix: it's because Object.keys can ignore the prototype chain
[16:09] creationix: it's only a win if you were wanting the base keys and had a hasOwnProperty check anyway
[16:10] mjr_: I love using the functional style with keys, forEach, map, etc.  I just assumed that they were all much slower.
[16:11] mjr_: _ry: I'll update the docs for the new repl code, sure.
[16:12] avidal: hrrrrm
[16:15] javajunky: I wonder how many more times someone will ask the same question about the web socket writes ;) 
[16:15] avidal: heh
[16:15] avidal: am i the only one who doesn't really like camelCase?
[16:16] confounds has joined the channel
[16:17] N`: avidal: I do not like camelCase too
[16:18] micheil: _ry: how do I manage to load this new module?
[16:18] mjr_: I like that everybody uses camelcase, because then my own code's under_scores are very obviously different, and then library / system names are easy to distinguish. 
[16:18] mjr_: So keep it up with the camelcase everybody.
[16:19] N`: http://ozmm.org/posts/javascript_style.html is about that
[16:22] micheil: got it. needed to add it to the natives
[16:23] maritz has joined the channel
[16:25] dgathright has joined the channel
[16:26] weepy_ has joined the channel
[16:27] keeto_ has joined the channel
[16:28] micheil: _ry: patching your way!
[16:29] weepy_: Hi -> does anyone know how I would convert "socket.write('\u0000' + data + '\uffff');" for node 0.1.9 ?
[16:29] micheil: umm.. that would work.
[16:29] javajunky: *irony* , yes
[16:29] javajunky: micheil: would it ? 
[16:29] micheil: weepy: when do you need the websocket implementation by?
[16:30] javajunky: socket.write('\u0000'+data+'\uffff', 'binary') would ....
[16:30] micheil: oh
[16:30] micheil: wait.
[16:30] micheil: no.
[16:30] micheil: socket.write("\u0000", "binary"); socket.write(data, "utf8"); socket.write("\uffff"
[16:31] javajunky: socket.write('\u0000','binary');socket.write(data);socket.write(''\uffff', 'binary');  I think would also work…plus there's the buffer approach
[16:31] micheil: , "binary");
[16:31] micheil: yeah, buffers would also work.
[16:31] javajunky: so many web socket server implementations ;) 
[16:31] weepy_: heh
[16:32] weepy_: sorry - im a bit confused now :)
[16:32] micheil: javajunky: I'm working on another to be spec compliant, all others aren't
[16:33] micheil: _ry: incoming! https://gist.github.com/dea62ecdaca7da298347
[16:33] micheil: _ry: we'll document it later I think.,
[16:33] javajunky: I have to admit, socket.io is the one I'd like to see spec compliant, I"m currently using a patched node.ws variant , but socket.io does *so* much it'd be sweet
[16:33] micheil: meh.
[16:33] Gruni has joined the channel
[16:33] micheil: I'm wanting specific things out of it, hence the reason I'm writing one
[16:34] BinaryPie has joined the channel
[16:34] javajunky: isn't that everyone's excuse ;)
[16:34] micheil: yeah
[16:34] micheil: however, I'm wanting it to be just websockets to the spec.
[16:34] javajunky: yeah I've been following your tweets on it ;)
[16:34] micheil: heh heh
[16:35] dnolen has joined the channel
[16:35] weepy_: FYI: => socket.write('\u0000' + data + '\uffff', 'binary'); <= seemed to work
[16:36] micheil: it's not right though
[16:36] javajunky: weepy_ yeah it will, but you may struggle to send true utf8 strings down there wire (its what I'm doing *lazily* in my case as I control the data) .. the other approaches are 'better' I believe..
[16:36] micheil: data should be in utf8
[16:36] inimino: and it's \u00ff that you want to send
[16:37] javajunky: inimino is it ? 
[16:37] inimino: javajunky: yes, it's a single byte
[16:37] javajunky: hmm crap  /me goes off to change some stuff.
[16:37] inimino: there's only 8 bits (two hex digits) in a byte
[16:37] javajunky: ;) yeah thanks for that
[16:38] inimino: binary encoding only uses the low byte anyway
[16:38] micheil: 0xFF
[16:38] steadicat has joined the channel
[16:38] javajunky: hmm lesson learned, don't trust other people's code, always go straight to spec :) .. my bad
[16:40] micheil: latest spec is available here: http://www.whatwg.org/specs/web-apps/current-work/complete.html?slow-browser=1#network
[16:40] micheil: it may crash your browser.
[16:40] inimino: try the multipage version?
[16:40] javajunky: (I'd gone off and read http://tools.ietf.org/html/draft-hixie-thewebsocketprotocol-75#section-1.7 no idea which is more up to date)
[16:40] creationix: I think we all copied the socket.write('\u0000' + data + '\uffff') line from the same example
[16:40] javajunky: quite probably :)
[16:41] inimino: probably, I've seen someone mention \uffff in just about every Web sockets discussion ;)
[16:41] micheil: hmm.. 32bit integers..
[16:42] micheil: that is the default integer type in js isn't it?
[16:44] inimino: ECMAScript doesn't define any integer type at all
[16:44] dgathright has joined the channel
[16:44] indiefan has joined the channel
[16:44] inimino: Number values are guaranteed to support all unique integer values up to 2^53
[16:45] micheil: hmm..
[16:45] inimino: and if an engine uses 32-bit integers as an optimization, you shouldn't be able to detect it or that would be a bug
[16:45] micheil: "Let challenge be the concatenation of part1, expressed as a big-endian 32 bit integer, part2, expressed as a big-endian 32 bit integer, and the eight bytes of key3 in the order they were sent on the wire."
[16:45] micheil: mean anything to you?
[16:45] inimino: hm, not in JavaScript...
[16:46] creationix: why would someone do this? for (var i in addition) if (i && addition.hasOwnProperty(i)) {
[16:46] inimino: there's no way to tell endianness on Number values
[16:46] creationix: can for..in sometimes give falsy keys?
[16:46] inimino: creationix: yes, the empty string
[16:46] creationix: really, as a key?
[16:46] micheil: means something to narwhal. http://github.com/280north/narwhal/blob/master/lib/struct.js#L73
[16:47] inimino: js> o={'':42},a=[]; for (p in o) a.push(p); a
[16:47] gbot2: inimino: [""]
[16:47] inimino: my browser froze
[16:47] creationix: ok, I'll preserve the functionality, not sure why they want to single out empty keys when mixing in values though
[16:48] inimino: oh yeah, I loaded that spec :/
[16:48] inimino: creationix: maybe it's a bug...
[16:48] creationix: it's the mergeObjects function in querystring.js
[16:48] creationix: looks like a simple deepClone to me
[16:50] inimino: probably a bug then
[16:51] creationix: 
ACTION wonders why everyone seems to think that deep clones are always the solution

[16:51] micheil: js> (4146546015 >> 16) + ((4146546015 & 0xFFFF) >> 16)
[16:51] gbot2: micheil: -2265
[16:51] micheil: :/
[16:52] inimino: oh, I should have mentioned
[16:52] micheil: js> (4146546015 >> 16) + (((0 & 0xFFFF) + (4146546015 & 0xFFFF)) >> 16)
[16:52] gbot2: micheil: -2265
[16:52] inimino: all the bitwise operators convert operands to 16-bit integers before doing anything else
[16:53] dandean has joined the channel
[16:53] micheil: so.. to get a 32 bit integer..
[16:54] RayMorgan: They get truncated to 32 bit integers...
[16:54] RayMorgan: https://developer.mozilla.org/En/Core_JavaScript_1.5_Reference/Operators/Bitwise_Operators
[16:54] inimino: sorry
[16:54] inimino: yes, I meant to say 32
[16:56] brianmario has joined the channel
[16:57] robinduckett has joined the channel
[16:57] robinduckett: Hello
[16:57] robinduckett: Anyone alive?
[16:58] atmos: nope
[16:58] robinduckett: Good stuff
[16:58] inimino: micheil: yeah, the endianness stuff in that file seems dealing with string conversions
[16:59] micheil: hmm
[16:59] robinduckett: Small question, what shall I do if I'm getting a segmentation fault?
[16:59] micheil: so, what would be the best approach?
[16:59] creationix: 
ACTION running make test-all after my large patch, wonders how many failures I'll get

[16:59] inimino: micheil: strings are treated as strings of unsigned 16-bit values, so if you are reading 32-bit values out of a string, you have to care about the endianness of the string
[17:00] inimino: so read two characters, left shift one by 16 bits and or them
[17:00] inimino: where 'read' means charCodeAt()...
[17:02] micheil: so.. uh..
[17:03] micheil: inimino: lost me big time, sorry
[17:03] inimino: micheil: what's your goal?
[17:03] gwoo has joined the channel
[17:04] micheil: uh, to do what the spec says
[17:04] inimino: 
ACTION scrolls up

[17:04] micheil: "et challenge be the concatenation of part1, expressed as a big-endian 32 bit integer, part2, expressed as a big-endian 32 bit integer, and the eight bytes of key3 in the order they were sent on the wire."
[17:04] inimino: ah...
[17:04] inimino: I didn't realize that was from WS
[17:04] micheil: yeah
[17:05] micheil: it's part of the websocket security model
[17:05] ssteinerX has joined the channel
[17:08] _ry: it's very important that node script | head -1 works
[17:08] micheil: hmm?
[17:09] micheil: _ry: get the patch?
[17:09] inimino: this Sec- stuff is new since I last looked at Web Sockets
[17:09] micheil: it is new
[17:09] _ry: script.js: while (true) { require('sys').puts('hello') }
[17:09] micheil: last few weeks
[17:09] _ry: ^-- this ought to be killed by head -1
[17:10] _ry: micheil: yes, testing it
[17:10] micheil: k
[17:10] teemow has joined the channel
[17:12] creationix: _ry, got my patch done, and it's passing all the tests except for test-http-head-request
[17:12] creationix: not sure how that one broke, but I'm debugging it
[17:12] _ry: creationix: okay
[17:12] _ry: test-http-head-request is a known issue
[17:12] creationix: really?
[17:12] _ry: yes
[17:12] _ry: http://github.com/ry/node/commit/93913dbdb8f621bef46c61d1f8a824a19a57dfa5
[17:13] creationix: ahh, makes sense now
[17:13] creationix: ok then, sending patch...
[17:14] creationix: ry: should I cc the mailing list so others can review it, or just send it direct to you?
[17:14] creationix: _ry: ^
[17:14] _ry: just paste it here
[17:16] felixge has joined the channel
[17:16] felixge has joined the channel
[17:16] creationix: _ry: http://gist.github.com/363785
[17:19] creationix: I did find a bug in http cat while I was refactoring, it wasn't checking for the case of array style headers when looking for the hostname header to replace
[17:19] inimino: micheil: ok, so you need a MD5 implementation
[17:19] micheil: yeah
[17:20] inimino: micheil: if you find one it will probably take input as a string
[17:20] micheil: in which case, I do: part1+""+part2+""+part3
[17:20] micheil: and md5 it
[17:20] micheil: yeah?
[17:20] inimino: no
[17:20] micheil: ?
[17:20] inimino: part1 and part2 are your 32-bit numbers?
[17:21] micheil: yeah
[17:21] inimino: so it's...
[17:21] teemow has joined the channel
[17:23] Hieu has joined the channel
[17:23] inimino: String.fromCharCode(part1>>>16, part1 & 0xFF, part2>>>16, part2&0xFF) + part3
[17:23] inimino: micheil: should be something like that
[17:23] micheil: okay
[17:24] inimino: hm
[17:24] micheil: js> String.fromCharCode(829309203>>>16, 829309203 & 0xFF, 259970620>>>16, 259970620&0xFF) + ^n:ds[4U
[17:24] gbot2: micheil: Error: SyntaxError: syntax error:  String.fromCharCode(829309203>>>16, 829309203 & 0xFF, 259970620>>>16, 259970620&0xFF) + ^n:ds[4U  ...........................................................................
[17:25] micheil: js> String.fromCharCode(829309203>>>16, 829309203 & 0xFF, 259970620>>>16, 259970620 & 0xFF) + ^n:ds[4U
[17:25] gbot2: micheil: Error: SyntaxError: syntax error:  String.fromCharCode(829309203>>>16, 829309203 & 0xFF, 259970620>>>16, 259970620 & 0xFF) + ^n:ds[4U  .........................................................................
[17:25] inimino: needs more quotes
[17:25] micheil: js> String.fromCharCode(829309203>>>16, 829309203 & 0xFF, 259970620>>>16, 259970620 & 0xFF) + "^n:ds[4U"
[17:25] gbot2: micheil: "ㅮཾ<^n:ds[4U"
[17:25] micheil: :/
[17:25] micheil: I don't think that's right.
[17:25] inimino: hm
[17:26] sveisvei has joined the channel
[17:26] inimino: it's not right
[17:26] micheil: js> String.fromCharCode(0x36, 0x09, 0x65, 0x65, 0x0A, 0xB9, 0x67, 0x33, 0x57, 0x6A, 0x4E, 0x7D, 0x7C, 0x4D, 0x28, 0x36)
[17:26] gbot2: micheil: "6	ee\n¹g3WjN}|M(6"
[17:27] micheil: ugh.
[17:27] micheil: fail. I'm using the wrong data
[17:27] _ry: creationix: i wish we had code review :(
[17:27] inimino: ok
[17:28] creationix: _ry: I know
[17:28] inimino: micheil: are you on the right track, anyway?
[17:28] creationix: are there problems?
[17:28] _ry: creationix: +  for (var socket in activeWatchers) if (activeWatchers.hasOwnProperty(socket)) {
[17:28] _ry: can we make this with braces
[17:28] _ry: around the for
[17:28] _ry: dns.js
[17:28] micheil: inimino: I think so
[17:28] inimino: ok, I'm going to go get lunch bbl
[17:29] creationix: _ry: on two lines or one?
[17:29] _ry: creationix: total nit
[17:29] _ry: creationix: two
[17:29] dnolen has joined the channel
[17:29] creationix: ok, I prefer two lines also, I had left it on one line so I could see the hasOwnProperty check in my search results
[17:30] creationix: but now I think it's the only one left, so it's easy to find
[17:30] micheil: js> String
[17:30] gbot2: micheil: <function String() { [native code] }>
[17:30] micheil: js> String.prototype
[17:30] gbot2: micheil: {}
[17:30] micheil: js> String.toCharCode("㘉eહ3WjN}|M(6")
[17:30] gbot2: micheil: Error: TypeError: String.toCharCode is not a function 
[17:31] micheil: js> "㘉eહ3WjN}|M(6".toCharCode()
[17:31] gbot2: micheil: Error: TypeError: "㘉eહ3WjN}|M(6".toCharCode is not a function 
[17:31] creationix: _ry: updated the gist
[17:31] mjr_ has joined the channel
[17:32] micheil: js> ""㘉eહ3WjN}|M(6"".charCodeAt()
[17:32] gbot2: micheil: Error: SyntaxError: missing ; before statement:  ""㘉eહ3WjN}|M(6"".charCodeAt()  ..^ 
[17:32] micheil: js> "㘉eહ3WjN}|M(6".charCodeAt()
[17:32] gbot2: micheil: 227
[17:32] micheil: "6ee\n¹g3WjN}|M(6".charCodeAt()
[17:32] micheil: js> "6ee\n¹g3WjN}|M(6".charCodeAt()
[17:32] gbot2: micheil: 54
[17:32] micheil: :/
[17:35] piranha has joined the channel
[17:35] Hieu has joined the channel
[17:35] micheil: _ry: are there any encoding converts, eg, Encoder.convert(str, from, to)
[17:35] BiscuitsLap has joined the channel
[17:35] Hieu has left the channel
[17:35] xer0x has joined the channel
[17:36] nsm has joined the channel
[17:36] javajunky has joined the channel
[17:38] micheil: js> String.fromCharCode(0x36, 0x09, 0x65, 0x65, 0x0A, 0xB9, 0x67, 0x33 )
[17:38] gbot2: micheil: "6	ee\n¹g3"
[17:39] micheil: "906585445".charCodeAt()
[17:40] micheil: js> "906585445".charCodeAt()
[17:40] gbot2: micheil: 57
[17:40] micheil: js> "9".charCodeAt()
[17:40] gbot2: micheil: 57
[17:40] micheil: js> "906585445".charCodeAt(2)
[17:40] gbot2: micheil: 54
[17:40] micheil: js> "906585445".charAt()
[17:40] gbot2: micheil: "9"
[17:40] micheil: hmm..
[17:40] _ry: micheil: no
[17:41] micheil: hmm?
[17:41] micheil: oh, right
[17:41] creationix: micheil: why do buffers not work for you?
[17:42] creationix: they should be easier than all this string junk
[17:42] micheil: uhh
[17:42] micheil: "Let challenge be the concatenation of part1, expressed as a big-endian 32 bit integer, part2, expressed as a big-endian 32 bit integer, and the eight bytes of key3 in the order they were sent on the wire."
[17:42] micheil: that's why./
[17:43] creationix: I saw that, but how do strings work better than buffers
[17:43] micheil: because, we're needing to convert things to 32 bit ints.
[17:44] creationix: yes, but you need to store those big-endian integers somewhere no?
[17:45] creationix: Strings will do crazy unicode stuff on you if you're not careful
[17:48] CIA-77: node: 03Tim Caswell 07master * r62d9852 10/ (8 files in 2 dirs): Replace slow and broken for..in loops with faster for loops over the keys. - http://bit.ly/ak7z3M
[17:48] micheil: :/
[17:48] micheil: I'm not sure
[17:48] creationix: micheil, what's the input to your function you're trying to write
[17:48] mjr_: Surely it would be straightforward to add methods to Buffer to slice out things like ints in both endian flavors.
[17:49] CodeOfficer has joined the channel
[17:49] creationix: I'm assuming part1 and part2 are just random 32 bit integers that come across the wire as 32 bits of data
[17:50] creationix: mjr_: and put them back in
[17:50] mjr_: yeah, absolutely
[17:50] mjr_: Buffer is great. 
[17:50] psynaptic has left the channel
[17:51] avidal: is that bot from github or something else?
[17:51] avidal: cia.vc, checking now!
[17:51] avidal: ah, cool
[17:53] mjr_: cia.vc is super fast just clicking around on things.
[17:53] _ry: creationix: thanks for the patch
[17:53] _ry: micheil: still having problems with yours
[17:53] creationix: _ry, no problem, hopefully nothing breaks
[17:54] micheil: hmm..
[17:54] _ry: micheil: http://gist.github.com/363823
[17:54] avidal: apparently the blog is very outdated though
[17:54] avidal: no mention of it working at all for github for instance
[17:55] micheil: _ry: how so?
[17:56] _ry: the http changes are breaking things.. 
[17:56] micheil: oh
[17:56] micheil: yeah, I just implemented for net
[17:56] _ry: but i'll check it out later
[17:56] _ry: gtg now
[17:56] micheil: k
[17:56] _ry: late
[18:09] kriskowal has joined the channel
[18:11] fizx has joined the channel
[18:11] fizx has joined the channel
[18:12] towski has joined the channel
[18:15] piranha has joined the channel
[18:16] frodenius has joined the channel
[18:19] inimino: micheil: yeah, if you can use Buffer it might be easier, but JavaScript strings won't do any "crazy unicode stuff", they are just sequences of unsigned 32-bit integers
[18:19] inimino: erm, 16-bit integers
[18:22] WALoeIII has joined the channel
[18:26] Aduros has joined the channel
[18:30] brianmario_ has joined the channel
[18:30] aryounce has joined the channel
[18:31] charlesjolley has joined the channel
[18:33] creationix has joined the channel
[18:33] mjr_: I guess I don't understand JS Strings then, because I often find myself in crazy unicode land while doing binary stuff.
[18:40] fizx has joined the channel
[18:41] fizx has joined the channel
[18:42] inimino: mjr_: if you just use charCodeAt() and fromCharCode you can treat them just like strings of 16-bit integers and they are completely transparent
[18:43] inimino: if you print them or use them with node, or read UTF-8 data from the network or something, then you have to think about encoding issues
[18:45] _ry: mjr_: yeah, the logic for http client is very crappy
[18:46] _ry: it needs to be refactored at some point
[18:46] _ry: (looking at your error now)
[18:46] mjr_: do you at least get that Stream already opened thing on your machine?
[18:46] _ry: (experiencing it n linux)
[18:46] mjr_: oh, good
[18:46] mjr_: bad that we have this bug, but good that it's not just me.  :)
[18:47] botanicus has joined the channel
[18:53] javajunky has joined the channel
[18:59] mikeal has joined the channel
[19:03] aho has joined the channel
[19:06] malkomalko has joined the channel
[19:07] N` has joined the channel
[19:14] creationix: websockets are popular today http://remysharp.com/2010/02/14/slicehost-nodejs-websockets/
[19:15] creationix: hmm, seems to be an old article, wonder why he just tweeted it
[19:25] ayo has joined the channel
[19:28] aho has joined the channel
[19:29] mape: it's a trap!
[19:30] piranha has joined the channel
[19:32] fictorial: weird thing about that is that slicehost is rackspace
[19:32] fictorial: I wouldn't touch either of them
[19:35] inimino: I'm happy with slicehost
[19:35] inimino: I've been using them since before they were bought
[19:36] CIA-77: node: 03Ryan Dahl 07master * r4681e34 10/ lib/net.js : 
[19:36] CIA-77: node: Fix a race condition or two in net.js
[19:36] CIA-77: node: When making a TCP connection, readyState returns 'opening' while resolving
[19:36] CIA-77: node: the host. However between the resolving period and the establishing a
[19:36] CIA-77: node: connection period, it would return 'closed'. This fixes it.
[19:36] CIA-77: node: This change also ensures that the socket is closed before the 'end' event is
[19:37] CIA-77: node: emitted in the case that the socket was previously shutdown. - http://bit.ly/cYYZS7
[19:37] _ry: mjr_: fixed it, i think---^
[19:37] mjr_: sweet, lemme update and try it out
[19:37] admc has joined the channel
[19:38] mjr_: fictorial: what's your issue with slicehost?  I've been pretty happy with them for the past month.
[19:39] fictorial: I don't trust rackspace; poor track record.
[19:42] Iveaux has joined the channel
[19:42] mjr_: _ry: seems to be fixed on my http proxy, and I can't make test-http.js fail either.
[19:42] mjr_: But it seems to have broken test-repl.js  !
[19:43] mape: is it possible to run websockets crossdomain?
[19:43] _ry: mjr_: test-repl was already broken
[19:43] _ry: unless you're experiencing something else
[19:43] mjr_: It worked for me last night, but stopped working after I just pulled your latest.
[19:44] _ry: what's your error?
[19:44] mjr_: I'll take a look.
[19:44] _ry: http://gist.github.com/363887
[19:44] _ry: is it that--^
[19:44] rictic has joined the channel
[19:44] aryounce has joined the channel
[19:44] mjr_: yeah, where's "invoked" coming from?
[19:45] _ry: the problem is with test-repl is that it's making assumptions about message boundries based on when the 'data' event is fired
[19:45] mjr_: oh, indeed it does.
[19:45] _ry: the stream needs to be buffered and .indexOf('\n') used
[19:45] _ry: then slice off a line
[19:46] mjr_: Right.
[19:46] _ry: https://gist.github.com/a3d0bbbff196af633995 like line 437
[19:46] mjr_: But I might still say that "invoked 333" is unexpected.
[19:47] stephenlb has joined the channel
[19:48] JimBastard: which version of node should module creators be targeting. 1.3 or head?
[19:48] JimBastard: kinda confused
[19:48] _ry: 0.1.90 is nearly API stable
[19:49] JimBastard: awesome, ive been building on 1.9
[19:49] JimBastard: thanks
[19:49] JimBastard: ive been trying to kepp all my projects updated to HEAD
[19:51] mravaux has joined the channel
[20:03] JimBastard: hey creationix 
[20:03] JimBastard: i ran into a problem with step last night
[20:03] creationix: what's up
[20:03] JimBastard: not so much a problem but perhaps a limitation
[20:04] JimBastard: so i needed to read all the files of a directory and append the contents to some variable
[20:04] JimBastard: using step i'd have to be able to dynamically create the steps
[20:05] JimBastard: i got the point where i had an array full of methods in the order i wanted to step them
[20:05] JimBastard: but i wasnt sure how to send that to the Step method as parameters
[20:05] JimBastard: i wound up building a custom method called readDirectoryFiles
[20:05] creationix: for that case you don't need dynamic steps
[20:05] inimino: I like the new docs
[20:05] JimBastard: how would you recommend doing it?
[20:05] creationix: but if you did, just do Step.apply(null, steps)
[20:05] creationix: Step is just a function after all
[20:06] creationix: so what is your goal, to read the contents of a list of files into a single variable?
[20:08] JimBastard: pretty much yeah
[20:08] JimBastard: i knew i was being dumb
[20:08] JimBastard: .apply of course
[20:09] towski has joined the channel
[20:09] JimBastard: http://github.com/Marak/hook.io/blob/master/hookio/protocols/documentation.js -> readAllFiles
[20:09] JimBastard: rough draft from late last night
[20:10] JimBastard: just recurses over the list checking if its the last file
[20:10] Yuffster_ has joined the channel
[20:12] creationix: also I added some significant api to Step the other day that's not documented yet
[20:12] mikeal: JimBastard: why isn't CrockBot in #couchdb :)
[20:12] CrockBot: Douglas Crockford is Chuck Norris' stunt double.    ara_p
[20:12] JimBastard: i can fix that
[20:13] JimBastard: gimme a sec
[20:13] maushu has joined the channel
[20:14] CrockBot has joined the channel
[20:16] creationix: JimBastard: I'm writing you an example using step to see what you think
[20:16] JimBastard: thanks man
[20:16] pedro has joined the channel
[20:16] JimBastard: i aliased Step to hookIO.Step btw
[20:17] JimBastard: the way that hook.io works is that you have access to the hookIO variable everywhere (custom definition and protocols)
[20:17] JimBastard: so now people doing custom hook.io stuff can just call on Step
[20:20] creationix: cool
[20:20] mattly has joined the channel
[20:21] creationix: why does your readAllFiles take so many arguments
[20:21] JimBastard: i know i know
[20:21] JimBastard: it was like 4am man
[20:21] JimBastard: :p
[20:21] JimBastard: ive been up all weekend writing documentation
[20:21] JimBastard: hook.io theBook
[20:22] JimBastard: readAllFiles could be rewritten to make a lot more sense and be a good helper
[20:22] JimBastard: i was thinking maybe readFilesInDirectory or something could be a fs call
[20:22] JimBastard: or readDirectory could return file handles
[20:22] JimBastard: unless there is already an option to do that? i didnt find it in the docs
[20:23] creationix: no, it's too high level for the fs module
[20:23] creationix: I do it all the time in my stuff
[20:23] JimBastard: would be a decent helper
[20:23] JimBastard: hey felixge any status ;-)
[20:24] creationix: JimBastard: how about something that takes a path, a regex, and returns an object with filenames matching the regex and their contents?
[20:25] JimBastard: huh?
[20:25] JimBastard: i want to send in a path and a filter and get back an array of file contents
[20:25] JimBastard: are we on the same page?
[20:26] JimBastard: i'd rather it be a filter method on an array then a regex?
[20:26] creationix: fs.globAndRead(hookIO.PATH + '/docs', /\.md$/, function (err, contents) {...})
[20:26] creationix: well a regex is a function
[20:26] creationix: just with special implementation
[20:26] creationix: you could drop in a function if you needed more capibilities
[20:26] JimBastard: well no, by forcing regex as a parameter you can forcing regex
[20:26] JimBastard: aye
[20:26] JimBastard: exactly
[20:26] JimBastard: creationix is fucking smart
[20:26] JimBastard: ahaha
[20:27] JimBastard: i'm good for now though, the .apply trick for Step should help
[20:27] JimBastard: my current method is OK for now, just not pretty or really reusable
[20:27] creationix: also, have you seen Step.fn from my latest version
[20:28] JimBastard: im using latest version i think
[20:28] JimBastard: it would be cool if you put up some more docs on the using the "advanced" features of step
[20:28] creationix: I don't think Step.fn is documented in the README, but it has a cool comment
[20:28] creationix: I might after I'm done with wheat, it's using step all over the place
[20:29] felixge: JimBastard: I want to kill myself
[20:29] felixge: JimBastard: I'm thinking Java or ColdFusion, can't decide
[20:29] creationix: felixge: wrong answer, try again
[20:29] JimBastard: i have 8 years of professional CF experience, sup?
[20:29] felixge: creationix: Cobol?
[20:30] creationix: now you're talking, I hear that pays well in California
[20:30] felixge: whatever is most painful :)
[20:30] JimBastard: i would highly suggest .net
[20:30] JimBastard: its just the right combination of super powerful and super painful
[20:30] JimBastard: so you kinda get results, but you are stuck in ms hell
[20:30] creationix: no, .net isn't painful, it lulls you into numbness and turns you into a zombie
[20:30] sveisvei has joined the channel
[20:31] JimBastard: and the win32 api slowly chokes you while you are sleeping
[20:31] felixge: alternatively one of you guys could do my accounting, launch that startup, finish dirty & the presentation and work on my mail/email backlog. :)
[20:31] JimBastard: :p
[20:32] JimBastard: its gonna take me about a day or so to rewrite my query library and install dirty
[20:32] JimBastard: so ughh pllleeeeeeeeeeeease
[20:33] JimBastard: ill send you internet dollars
[20:33] mikeal: _ry: I think you fixed my bug :)
[20:33] JimBastard: im not above begging
[20:33] _ry: mikeal: sweet
[20:33] _ry: mikeal: i was hoping that did it
[20:34] mikeal: I'll move on to more testing now so that i can move to 0.1.90 as soon as it's released
[20:34] creationix: so we're all going to be dead tired by the time jsconf gets here from preparing for it, isn't that counter productive?
[20:35] mape: when is jsconf?
[20:35] _ry: this weekend
[20:35] derbumi has joined the channel
[20:35] mikeal: i'm wondering if it's time yet to be bitter about all the people that get to go when I don't :P
[20:35] JimBastard: im an animal creationix ill just sleep on the bus
[20:35] mattly: man i wish i was going to jsconf after all
[20:36] JimBastard: mikeal i dont have a ticket im just crashing the parties
[20:36] mattly: it was great last year
[20:36] brianmario has joined the channel
[20:36] mikeal: i can't put the flight and hotel on my corporate card just to crash the parties :)
[20:36] deanlandolt: JimBastard: you hanging around for most of it?
[20:36] creationix: mikeal: I'm not going either
[20:36] mjr_: mikeal: I think it's time to be bitter.  I've already started.
[20:36] JimBastard: deanlandolt: maybe......friday for sure. if i win that magic ticket ill stay all weekend for sure
[20:36] mikeal: we'll have out own conference mjr_
[20:36] felixge: JimBastard: actually, let me get some beer and fucking do it
[20:36] mikeal: in Oakland
[20:36] JimBastard: i hear the sat night party is dope
[20:36] mikeal: fuck you guys
[20:36] felixge: tonight is ruined anyway
[20:36] felixge: :)
[20:36] mjr_: the 'Town
[20:36] JimBastard: deal felixge 
[20:36] mikeal: it'll be at the Trappist
[20:37] mikeal: and it lasts 3 days :)
[20:37] creationix: felixge: good luck
[20:37] creationix: is anyone in this room from Texas?
[20:37] felixge: creationix: I'm not sure I trust your for var in tests, what were the biggest objects you tested? I saw similar performance of Object.keys() / for var in for huge objects (millions of keys)
[20:37] mjr_: mikeal: surely couchio will host some sort of thing like that, right?  Two conference rooms: one in the office, one at the Trappist.
[20:38] creationix: felixge: my benchmark was 500 keys, I tested it for smaller numbers too
[20:38] mikeal: for serious, we will be doing a CouchCamp in the fall
[20:38] creationix: most the code I replaced was using less than 10 heys
[20:38] creationix: *keys
[20:38] mikeal: but we're going to rent a camp spot in the woods
[20:38] _ry: creationix: where's your gist?
[20:39] creationix: https://gist.github.com/bdd159b4f8f7c900d939
[20:39] mjr_: mikeal: sounds great
[20:39] creationix: uses issacs/node-bench to run
[20:39] mikeal: I think we'll also do another CouchHack that lines up with it
[20:39] mikeal: for people working *on* couch rather than *with* couch
[20:39] fictorial: heh, where do I start debugging this? IOWatcher.callback (net:329:16)
[20:39] fictorial: node.js:814:9
[20:40] fictorial: that's the entire output
[20:40] mjr_: mikeal: don't know what all these CouchWords mean, but I do like the idea of not feeling left out of jsconf.
[20:40] creationix: felixge: that makes sense for objects with large numbers of keys and a shallow prototype chain, the savings you get shows up more when you can feel the cost of the prototype chain
[20:42] felixge: creationix: ok, makes sense. Man I wish there was such a thing as iterators of objects
[20:42] felixge: so I could start iterating, pause, continue, etc.
[20:42] creationix: that would be neat
[20:44] deanlandolt: kriszyp's been pushing on the es group with a proposal that does that
[20:46] felixge: deanlandolt: he's a good man :)
[20:49] creationix: Why did I get this: TypeError: /\.js$/ is not a function
[20:50] creationix: ahh, typeof a RegExp is function, but it's not an instance of Function, that's weird
[20:51] mjr_: isn't that true of all constructors?
[20:51] creationix: JimBastard: this should help https://gist.github.com/9dde3c73d28b445f04bc
[20:51] mjr_: typeof Number === 'function' as well
[20:51] creationix: no, not the constructor RegExp, but an instance of it
[20:51] indiefan has joined the channel
[20:52] creationix: you can call regular expressions as if they were functions
[20:52] creationix: but apparently, you can't pass them as the function callback to Array.prototype.filter
[20:52] mjr_: Oh, I didn't know you could invoke regexps.
[20:53] mjr_: but sure enough, you can.
[20:53] creationix: yep, it's like calling match I think
[20:53] fictorial: guys, I have an uncaughtException handler but I cannot get a stack trace inside of it.  Error.captureStackTrace does not seem to be documented.  I looked at the v8 source and well, I have no idea what the parameters are for.
[20:53] mjijackson has joined the channel
[20:54] isaacs has joined the channel
[20:54] _ry: step is pretty nice
[20:56] JimBastard: sup mjijackson 
[20:56] mjijackson: Hey JimBastard 
[20:57] JimBastard: yeah Step is pretty sweet _ry , makes things simple
[20:57] JimBastard: mjijackson its really cool you released mint today, i created a documentation protocol for hook.io just last night
[20:57] JimBastard: im already autogenerating docs based on the github markdown files
[20:57] JimBastard: so now i just need the other half
[20:57] felixge: fictorial: the stack trace should be captured the moment the error is created for normal stuff, you might be hitting a runtime JS error which I think is handled differently?
[20:57] mikeal: _ry: I still have this error, Error: Stream.close() called already; cannot write.
[20:57] inimino: creationix: it's probably a bad practice to call them (though I've seen someone here do it...) but yeah, they are typeof 'function' (and callable) for historical reasons
[20:57] mikeal: when writing
[20:58] mikeal: and i *know* that it hasn't been closed
[20:58] hellp has joined the channel
[20:58] mikeal: i have listeners on close, end, and error and none of them have fired
[20:58] creationix: inimino: I only do it when I want an api that I can drop in a full function instead of a regexp, like when writing simple lexers.
[20:59] inimino: creationix: I think it's the sort of thing that might go away in some future strict mode, which is why I avoid it, but I guess as long as it works...
[21:00] creationix: yeah, no biggie, I can change my code if the language changes, If I couldn't handle that kind of change, I'd never use node ;)
[21:01] mjijackson: JimBastard: Nice. You're using Mint to do it?
[21:02] mjijackson: Or you already had something else set up?
[21:02] JimBastard: http://github.com/Marak/hook.io/blob/master/hookio/protocols/documentation.js
[21:02] JimBastard: so my docs are its split into multiple MD files in a directory
[21:03] JimBastard: i just concat them together and output as html or whatever
[21:03] mjijackson: Ah, i see.
[21:03] JimBastard: pdf / png /
[21:03] creationix: ok, I'd better close this window or I'll never get my work done today, email me if anyone has questions.
[21:03] creationix has left the channel
[21:03] JimBastard: so now i need the code docs so i can place them inside the output
[21:03] JimBastard: maybe ala Mustache
[21:03] JimBastard: or just dump them into another dir
[21:04] mjijackson: JimBastard: I've grown really fond of Mustache lately. So many template languages don't really help you write clean templates.
[21:04] JimBastard: mustache is just simple
[21:04] JimBastard: and im pretty retarded
[21:04] mjijackson: Mustache actually helps you by preventing you from doing any logic.
[21:04] JimBastard: so it makes sense to me
[21:04] mjijackson: I like code that helps you do the right thing. ;)
[21:05] JimBastard: does Mint have a sample output anywhere?
[21:05] JimBastard: like maybe can you self document mint and post the html online?
[21:07] mjijackson: JimBastard: Yeah. It's on my hard drive. ;) I should have probably created a website for Mint before posting to the mailing list, but I wanted to release early.
[21:07] JimBastard: well really
[21:07] JimBastard: can you update to 1.9 ?
[21:07] deanlandolt: mjijackson: you could just push the generated docs to github :)
[21:08] JimBastard: i could go through and do it but im like insanely backlogged like everyone else this week
[21:08] mjijackson: Right now the output is very similar to the CoffeeScript docs.
[21:08] JimBastard: ive got issues http://github.com/marak/hook.io/issues
[21:09] mjijackson: deanlandolt: Will do.
[21:09] deanlandolt: mjijackson: oh, nice...looks like atul varma's work from a few years back
[21:09] deanlandolt: or, just a year back i guess: http://www.toolness.com/wp/?p=441
[21:11] mjijackson: JimBastard, deanlandolt http://github.com/mjijackson/mint/blob/master/doc/api/mint.html
[21:12] mjijackson: deanlandolt: Wow. That does look similar. I wonder if that's where jashkenas got the idea?
[21:15] mjr_: Man, scripting a test of multiple servers and clients interacting with each other is tricky.
[21:16] kriskowal: mjr_ amen.
[21:16] kriskowal: s/amen/so say we all/
[21:16] mjr_: I keep thinking I've come up with a simple abstraction for it, and then it falls down some other place.
[21:17] kriskowal: should be easier to model in node than most other systems
[21:18] mjijackson: mjr_: The best cure for async madness is a walk in the park. I always have my best ideas when I'm not at my desk.
[21:19] mjijackson: mjr_: That is, of course, if it's making you crazy.
[21:19] drostie has joined the channel
[21:19] joshbuddy has joined the channel
[21:19] mjr_: Yeah, I think I'm going to do something else for a while that promises to be easy.
[21:20] mikeal: _ry: I've done a lot of debugging on this and I kind of need to know what you think at this point
[21:21] _ry: mikeal: ?
[21:21] mikeal: so, (socket._writeQueueLast() == END_OF_FILE) is true before something is pushed to the writeQueue
[21:21] mikeal: but if I check socket._writeQueueLast() before I call .write() on the response
[21:21] mikeal: it's not END_OF_FILE yet
[21:23] _ry: mikeal: maybe it's drained since then?
[21:24] mikeal: it should be blocking on processing between my call to write and making it this far in the stack
[21:24] mikeal: shouldn't it?
[21:24] JimBastard: yo mjijackson those docs look pretty good
[21:25] mjijackson: JimBastard: Thanks! Lots more features to come. As soon as I can get some time to hack on it.
[21:25] mikeal: is END_OF_FILE actually sent as a chunk to the data event in the http client
[21:25] _ry: mikeal: no, it just triggers shutdown()
[21:25] mikeal: ok, then that's not the problem
[21:26] mikeal: nobody has called end or error or anything
[21:26] JimBastard: if you can update to 1.9 i'll hook it up into hook.io this week for sure
[21:26] darkf has joined the channel
[21:27] mjijackson: JimBastard: Yeah. I'll definitely get it onto 1.9 before the end of the week.
[21:28] derbumi has joined the channel
[21:28] mjijackson: Node is a fast-moving target!
[21:29] mikeal: this is driving me mad
[21:29] JimBastard: hrmmm markdown seems to have my ascii art
[21:29] mikeal: it's a blocking call
[21:30] mikeal: but the stream is changing from when i call it 
[21:30] mikeal: grrrr
[21:31] romainhuet has joined the channel
[21:32] _ry: mikeal: i don't really get it - you're callign write() when you've already called end()?
[21:32] mikeal: no, end hasn't been called
[21:32] mikeal: not at least as far as the http response object is concerned
[21:32] mikeal: I'm not adding code to Stream.prototype.end to see if someone else calls end
[21:33] mikeal: but here's the main thing
[21:33] mikeal: I added a reference to the socket from the response object
[21:33] mikeal: and I'm checking if it's sent END_OF_FILE, and when I call .write() it hasn't, but then when it makes it down the stack to the stream write the check says that it has send END_OF_FILE
[21:34] mjr_: _ry: Here's a fix for test-repl.js that isn't clean and slick like I was hoping, but does work reliably:  http://github.com/mranney/node/commit/6439ff9481e132fc81a2847a730694c83e602451
[21:36] mikeal: ok, something called end on the stream
[21:36] mikeal: i know that much :)
[21:37] _ry: mikeal: maybe the http client itslf?
[21:37] mikeal: I'm calling write on the response object
[21:37] mjr_: Can you stick something in end to see who is calling it?
[21:38] mikeal: ok, i'm not calling end, something else is
[21:38] rtomayko has joined the channel
[21:39] javajunky has joined the channel
[21:39] mjr_: Can you stick something in there that throws if it isn't you, so you'll get a stack trace?
[21:39] dgathright has joined the channel
[21:40] CIA-77: node: 03Matt Ranney 07master * rafe3c1c 10/ test/simple/test-repl.js : Buffer partial reads before doing expect match. - http://bit.ly/agd0uI
[21:41] pedro has joined the channel
[21:42] pedro has joined the channel
[21:42] mjr_: It seems to me that if we had an easy way to script tests of multiple clients and servers that problems like these would get noticed more easily.
[21:43] rednul_ has joined the channel
[21:43] xla has joined the channel
[21:48] _ry: mjr_: perhaps
[21:50] dgathright has joined the channel
[21:51] isaacs_ has joined the channel
[21:51] isaacs_ has joined the channel
[21:58] mattly has joined the channel
[22:01] tekky has joined the channel
[22:04] mikeal: _ry: i think i figured it out
[22:06] _ry: mikeal: cool
[22:06] _ry: what is it?
[22:06] mikeal: so, .write() calls flush()
[22:07] mikeal: which calls flushQueueMessage
[22:07] mikeal: and that is calling socket.write
[22:07] mikeal: wait, i'm confused, hold on
[22:07] mikeal: dammit i thought i had it
[22:08] _ry: heh - yeah the http client logic sucks
[22:08] _ry: it really needs to be refactored
[22:08] mikeal: this is kind of insane
[22:08] _ry: the problem is trying t share code between the client and the server
[22:08] _ry: which mostly works
[22:08] mikeal: i'm trying to find a non-blocking call somewhere in the write logic that is getting pushed behind
[22:08] _ry: but makes for some craziness
[22:09] mikeal: the thing is
[22:09] mikeal: since it does this whole queueing thing
[22:09] mikeal: there isn't a good way for me to check at my point in the code if I can call .write
[22:09] mikeal: because end hasn't fired yet, i know that
[22:09] mikeal: and writing is throwing an exception saying it's been closed
[22:10] binary42 has joined the channel
[22:13] charlesjolley has joined the channel
[22:16] mikeal: _ry: is there a reason that OutgoingMessage.end doesn't emit("end")
[22:16] mikeal: ?
[22:18] JimBastard_ has joined the channel
[22:18] _ry: mikeal: writable streams don't emit 'end'
[22:19] _ry: end() sends an eof
[22:19] _ry: 'end' recvs it
[22:19] mikeal: ok
[22:30] sh1mmer has joined the channel
[22:32] towski has joined the channel
[22:34] mikeal: _ry: I really think that a writable stream needs to emit an event after is ends EOF
[22:35] mikeal: otherwise there is no way to protect yourself from writing to it
[22:47] dnolen has joined the channel
[22:50] dnolen has joined the channel
[22:50] _ry: mikeal: hm
[22:50] _ry: mikeal: well supposedly you have full control of it
[22:51] _ry: in the case of a response from a http server, nothing else should end() it
[22:51] mikeal: that's what i would thing
[22:51] mikeal: er think
[22:52] _ry: but response != socket
[22:53] gf3 has joined the channel
[22:53] towski has joined the channel
[22:56] mikeal: i know
[22:56] mikeal: i put a print when the socket end happens an when i call response.end
[22:56] mikeal: and there are ton of socket.end calls that i don't initiate from response.end
[22:59] _ry: mikeal: hmm
[22:59] _ry: mikeal: could it be that you're responding to messages out of order?
[22:59] mikeal: i seriously doubt it
[22:59] _ry: that someone connecting is pipelining a bunch of requests
[22:59] _ry: and you answer the last before the first?
[22:59] mikeal: that's possible
[23:00] mikeal: if they were pipelining then the server would handle breaking it up in to different responses
[23:00] mikeal: wouldn't it?
[23:00] _ry: yes
[23:00] _ry: supposedly
[23:00] _ry: i think there is a test for that even
[23:01] _ry: but in that case the server will end() the connectin
[23:01] _ry: once it's sent out the last response
[23:03] creationix has joined the channel
[23:07] mjr_: mikeal: I think your problem is too many ifs.
[23:08] isaacs: mjr_: XD
[23:08] mjr_: cross-channel humor: guaranteed funny
[23:14] mikeal: _ry: this is what is breaking me
[23:14] mikeal: http://github.com/ry/node/blob/master/lib/http.js#L627
[23:14] mikeal: that is closing out the stream I'm trying to write to
[23:14] mikeal: or at least the fd matched
[23:15] hassox has joined the channel
[23:15] mikeal: and this is for a http.ServerResponse object
[23:17] _ry: mikeal: i thought we were talking about servers..
[23:18] mikeal: we are
[23:18] mikeal: the fd matches
[23:18] mikeal: this should never be called on a clientResponse
[23:19] tekky has joined the channel
[23:19] mikeal: sorry, ServerResponse
[23:22] _ry: no, it should never be called on a server response
[23:22] mikeal: but I'm print the fd before it does it
[23:23] mikeal: and the fd before my exception in Stream.write and they match
[23:23] mikeal: they are both 9
[23:23] lifo has joined the channel
[23:25] _ry: ok..hm
[23:25] towski has joined the channel
[23:27] _ry: i don't see how that's possible
[23:27] _ry: you're writing to a server response?
[23:31] mikeal: ok, more info
[23:31] mikeal:     at IOWatcher.callback (net:335:31)
[23:31] mikeal: that is what attaches Client to my fd
[23:32] mikeal: the Client.ondata event is firing and that's what triggers the close
[23:33] mikeal: oh, and this is calling close long before my server gets the clientResponse
[23:33] mikeal: sorry
[23:33] mikeal: ServerResponse
[23:35] _ry: mikeal: are you repoducing this regularly?
[23:35] _ry: mikeal: is it possible for me to get in there and check it out via ssh?
[23:35] mikeal: yeah, i have to trigger couchdb replication through my proxy to do it
[23:35] mikeal: but i can reproduce it every time
[23:36] CIA-77: node: 03Ryan Dahl 07master * r1efd261 10/ doc/api.markdown : doc typo - http://bit.ly/c4M2mR
[23:36] CIA-77: node: 03Ryan Dahl 07master * rb7947e4 10/ lib/http.js : Recycle http client parsers - http://bit.ly/aWAOWS
[23:41] mikeal: ssh would be a little hard
[23:41] mikeal: you can probably do it locally
[23:42] mikeal: oh no wait, that won't work either
[23:42] _ry: it'd be really good to get a test case :/
[23:43] mikeal: so, this is happening on a long running stream that doesn't have a content-length
[23:44] mikeal: i could write a test if i had any idea what is actually causing it
[23:45] _ry: doesn't have a content-length
[23:45] _ry: tell me more
[23:45] _ry: is it transfer-encoding: chunk ?
[23:45] mikeal: chunked
[23:45] mikeal: yup
[23:45] _ry: can you tell me if there are multiple messages in the stream?
[23:45] _ry: i.e. is it using keep-alive?
[23:46] _ry: and this stream is coming into a Node http server?
[23:46] _ry: or is this the response from couchdb?
[23:47] mikeal: here are the headers http://gist.github.com/364133
[23:47] mikeal: i'm proxying the request/response
[23:47] _ry: okay
[23:47] mikeal: and the error happens when I'm writing a chunk I got from the remote request to the ServerResponse object
[23:48] mikeal: somewhere over 500K it breaks
[23:48] mikeal: the response is a lot longer than that
[23:49] _ry: hmmmm...
[23:49] _ry: maybe you should tcpdump your connection
[23:50] _ry: to the backend
[23:50] _ry: both directions
[23:50] _ry: maybe we'll see if a FIN packet is being sent in?
[23:51] confounds has joined the channel
[23:52] mikeal: i really doubt it
[23:52] mikeal: without the proxy this all works fine
[23:53] mravaux has joined the channel
[23:57] towski has joined the channel