@piotrek in SO pointed out that I should be able to extend ring.core.protocols/StreamableResponseBody to do this

(extend-protocol protocols/StreamableResponseBody
  LinkedBlockingQueue
  (write-body-to-stream [output-queue _ ^OutputStream output-stream]
    (log/error "starting")
    (with-open [out (io/writer output-stream)]
      (.write out "starting")
      (loop [chunk (.take output-queue)]
        (log/errorf  "got chunk %s" chunk)
        (when-not (= chunk ::EOF)
          (.write out (str chunk))
          (.flush out)
          (recur (.take output-queue)))))))

though I'm trying to figure out whey that handler doesn't actually run for me

and setting the jetty output buffer size to one does make it "work" though that's a bad idea for other reasons i suppose

It might be because Jetty has a set buffer size that doesn’t support flush.

ok so it works if I jump into ring.util.servlet and reload the buffer with the repl

then it starts calling my extension to ring.core.protocols/SreamableResponseBody

Reloading protocols in the REPL does have some caveats I believe. Not sure if that affects records only or existing class extensions as well.

I don't normally need to reload the library that uses a protocol to have it notice that I extended that protocol right?

I don’t believe so.

It does seem to work when all the protocol is extended in a namespace that's loaded initially, so this is a REPL only situation whatever it is that's going on here

can anyone point me to a good article explaining the viewpoint of not using namespace aliases?

@wei of not using them? I'm not sure that I've seen any. What problems are you anticipating?

where namespace alias is an alternative to unqualified import perhaps?

@tbaldridge the arguments I’ve heard for using fully-qualified names for every function call are related to readability. interested to see the argument fleshed out in an article and I was wondering if anyone has written about it

I’ve been preferring one-letter namespace aliases for conciseness, so I’m interested in hearing the opposite view

I've been playeing with one letter namespaces

so stuff is named things like c for client

and client/gui.cljs instead would be c/g.cljs

then, when I import the namesapce, it's just [c.g] and its' short enough to not need to alias

why would you do that

use last part of namespace as aliasename , example: (:require [http://aa.bb.cc :as cc] [xx.yy.core :as yy])

Try to keep the alias globally unique, just like best practices for database naming

also convenient to generate notepad++ autocomplete file.

is it possible to refer to a gensym generated with the # syntax from inside ~?

ok

just quoting it with backtick returns the generated symbol

@stathissideris no

you'll have to manually use gensym

auto-gensyms only exist within the backtick context they were created in

@bronsa thanks, I thought I was missing something 🙂

if you

`[a# ~`a#]

this will generate 2 different gensyms, one per backtick context

got it. It’s just that I end up using gensym “manually” most of the time and I thought that maybe there was a way to avoid it that I didn’t know about

@ctford Hi there. I’m hoping you could help me understand something with Leipzig (which I order by the way), I’m passing in nils to act as breaks in phrases, but when I use live/play it silently fails and stops playing entirely when it gets to the nil. Is this behaviour you’ve experienced before?

@wei, I haven't heard convincing arguments to not use namespace aliases. Most programs will have a convention to them that makes their aliases make sense in the context of the project. As a consultant I read a lot of code written by other people, and the aliasing of namespaces has never been a problem for readability. Problems with readability I've encountered are much more around things like overuse of higher order functions, macros, etc.

if I call a function from within a core.async go block and then try to use the >!, <! channel macros within that function it is considered "not in go(....) block". What's the rationale here?

answering myself. Found a discussion around go block semantics at https://github.com/clojure/core.async/wiki/Go-Block-Best-Practices

timeout returns the same object, for two different calls, I know it's memory efficient, but it's not the Clojure idiomatic way I guess

is there a version of with-out-str that returns both the stdout string and the value that would have been returned from the expression?

I want to basically do

(let [[elapsed result] (with-out-str (time (some-function)))] 
    (log elapsed)
    result)

@tjtolton I don't think there's anything like that in core, but if you check out the source, it's a really simple modification

yeah, the example even has a simple version of it

called with-out-str-data-map

@tjtolton fyi, the example should have (do ~@body), before it bites you 🙂

when does that bite one?

I'm not asking because I disagree, I'm asking because that's valuable information to have

try (with-out-str-data-map 1 2)

ohhhh

I see

hahaha

if you are on 1.9, you should get a really nice readable error message (jk)

nice. great catch.

I haven't run it yet

but, in figuring out where to put that line, I realized the problem

hi o/ what is the best way to test clojure developement under mswindows ?

I know boot is not fully supported

is the poison pill pattern considered an anti-pattern for core async pipelines? i.e. sending down a special value down a chain of pipelines to shut the "network" down

i wouldn't say it's an anti-pattern at all, but usually you close the channel instead of poisoning it

You'd use a different poison pill if you want to keep the channels open after poisoning

does anyone have any conventions around using some?? I can’t think of when I’d ever use it over just checking the truthiness of the form instead, or using seq and empty? in the case of sequences

core.async

some? was essentially added to test for closed core.async channels

see also if-some

interesting. because you might want false as a value?

yes

well thanks!

I'm seeing a weird stackoverflow occur and I'm having a really hard time reasoning about it

https://gist.github.com/fe1b3f7baa0a28849d3d8ddbe84b1216

I have no idea why this is overflowing. My understanding is that this should be tail call optimized, so this sort of overflow shouldn't happen right?

cursory glance: concat in a loop: https://stuartsierra.com/2015/04/26/clojure-donts-concat

he's not concatenating over the results of the iteration

ah thanks for the clarification for me

I see link-snapshot is either (get-as-hickory ...) or its an error. But either way you call (data-parser-fn link-snapshot) on it. Is it possible you fail above and then call data-parser-fn on data that doesn't make sense?

case fails to match ==> exception cond failes to match ==> it just returns False ?

nil

not false

i think the reasoning is that cond is more like an if .. else if ... else if .... where case is more like a switch

I see; it just surprised me as I'm used tocase throwing xceptions (which I then debug) on not matching, whereas cond just "silently return snil"

if anything case is the exception here, most things return nil in clojure on not found/no match

@shayanjm so is that error happening inside (data-parser-fn link-snaphsot) ?

if so, I bet data-parser-fn is doing something funky with nested lazy operations in a loop

yeah, case is enumerating options, cond not necessarily

@noisesmith that's a great question

It's definitely happening in the link parser but didn't drill down into it

I included a full trace which refers to hickory quite a bit

I suspect you're likely right

hello. I’ve just used https://github.com/clojure/core.match in my re-frame pet project and it adds some clarity to my code. I wonder why it’s stuck at alpha version?

Well, to quote some maintainers of Clojure: " Because we don't really associate any emotional significance to the version number." 🙂

wellllll, the whole spec.alpha seems to telegraph feelings about the significance of the alpha designator, if not the version number

But last I heard (and I can't find the talk) there's also some open questions about how core.match should evolve in a language like Clojure that prefers open dispatch.

dnolen is also a busy guy

I.e. in a perfect world we could extend a match statement with new clauses from a different namespace, and yet not loose any of the performance or features of the original statement.

there’s also the fact between for, doseq, match, and core.async you can use two, maybe three if you are lucky, in one nested scope, before your methods are too large for the jvm

sometimes you can only get away with one, depending on how well the code is factored…

ok, but I mean it’s so useful

doseq is a neat nerdsnipe https://gist.github.com/ghadishayban/fe84eb8db78f23be68e20addf467c4d4 even with the neat approach there you still generate a fair bit of code

very cool!

heh that code output doesn't even take into account chunked seq or the let/while/when clauses

99% of doseq's in the wild tho, do not use let/while/when

and most have a single binding

yeah, it is a rabbit hole, which makes it a neat puzzle, you think "oh this is simple, I'll just ..." and then five hours later you have something that frustratingly still doesn't cover all the edge cases

the compiler nerd side of me hates that Clojure isn't as simple as Scheme. Every other side of me is thankful that it's not.

What are people using for data visualization from a clojure repl? 🙂

haven’t used it, but this looks impressive: https://atom.io/packages/proto-repl

@darwin Its a good suggestion. When I was using Atom as my editor I was using proto-repl combined with this proto-repl-charts. Since switching over to Emacs and Cider I was not able to get proto-repl-charts to place nicely.

Sorry, first time slack user. Is it better to reply in the thread or in the main channel?

that’s pretty new feature, I guess most people don’t care

So far I'm aware of the following ideas. Incanter uses Jfreechart, which I find creates very old fashioned looking plots, but its good you can use it from a repl. Proto-repl-charts is using C3.js as a backend, built on top of D3.js which is cool which creates plots in a browser window. Similarly gorilla-repl plots using a similar endeavour called vega,

Both Vega and C3 are very healthy looking projects. https://github.com/c3js/c3 https://github.com/vega/vega

i’ve used gorilla-repl a bunch (http://viewer.gorilla-repl.org/view.html?source=github&user=wiseman&repo=leaflet-gorilla&path=example-worksheet.cljw is one of the more involved examples, with a map renderer i wrote), but i would probably look to see what the state of jupyter/clojure integration was if i were doing that again.

thats cool 🙂

anyone know how to pronounce timbre? https://github.com/ptaoussanis/timbre

it’s a logging library, so I imagine it’s pronounced “timber”, as someone would yell when chopping down a tree; in the logging industry.

timbre and timber are completely different. lumberjacks have timber, musical instruments have timbre ("TAM-br").

assuming it was intentional, it's a beautiful pun. different logging libs have different timbres, even tho they all log.

I have a question about excluding dependencies when pushing to clojars. When I'm working on some code, I tend to have more "development dependencies" in my project.clj file than are needed for users when I push to clojars. For example I might have some testing dependencies or some visualization dependencies that I use when I'm writing and testing the code, but these do not need to be in the jar that I distribute. Is there a flag to list some dependencies as development, so that lein deploy clojars does not add these as dependencies for others?

so like the test scope during maven build?

because i think you can do that with profiles

Yes, people use the dev profile for this in lein

Ah, this is what I'm looking for. Thank you Alex