@bradford you can just use ExecutorService directly if you like

regular clojure functions are Runnable already

as a general rule, if the host facility is good, clojure doesn’t try to replace it

'aite, thx

is there an alternative to -> or ->> where it's (----> arg f1 f2 f3 f4) where the fi's are all of the form #( .... % ... ) ?

i.e. after the first arg, all the exprs result in functions which take just 1 arg

and I want to thread the arg through all the functions

@qqq (doc as->)

@bbloom: this works; thanks!

@bradford You may also be interested in https://github.com/TheClimateCorporation/claypoole this lib.

@mingp Thanks! I was looking at that but didn't see anything for timeouts

Besides Red5/Java, are there any other libraries for building live video streaming (like twitch ) in Clojure?

src/bar/foo.clj <-> src/bar/foo_test.clj is becoming really annoying; is there any downsides to using test/bar/foo_test.clj ?

Not sure what you're asking there @qqq -- most people have tests in the test subtree and source in the src subtree...

@seancorfield: I guess I was asking "is there anything bad if I stop kicking this rock" ;; I originally had impl/code in the same directory due to a problem with cider/emacs having difficult jumping between the two; but now they're in separate dirs, and I'm much happier

I would ask, why would you have tests in the src tree?

Pretty much everyone puts their tests in the test tree...

After all, when you build production artifacts, you don't want test included.

@seancorfield : because I'm working on hobby projects, and getting C-c p t to work was more important than directory structure at the time (until it started getting overcrowded)

I was going from "no unit testing -> unit testing" not "good practice -> bad practive"

there's a lesson there somewhere... 😉

I think multiple lessons are waiting in the backlog 😄

The greatest gain is from "no unit tests" to "unit tests" -- anything else is pre mature optimization. 🙂

Thank you @pesterhazy, that works

Anyone know what clojure.lang.AReference is for or about?

I have now these specs :

; specs for validatimg input to find rhe ids(s/def ::artObject (s/keys :req-un [::objectNumber]))(s/def ::artObjects (s/coll-of ::artObject))(s/def ::body (s/keys :req-un [::artObjects]))(s/def ::response (s/keys :req-un [::body])) 

and use a merge to test for both ?

@futuro it's a base class for reference type implementations

@alexmiller I must admit I'm a little confused by what a reference type is

I see in the source that clojure.lang.ARef extends clojure.lang.AReference, and so does clojure.lang.Namespace, and then c.l.Atom, c.l.Agent, c.l.Ref, and c.l.Var extend c.l.ARef; which at first glance makes me think c.l.AReference is a super-type of just about everything

ah, but a little searching helps me; am I correct in understanding that vars, refs, agents, and atoms are all reference types?

yes

a reference type is generally a "pointer" to a single value, and you can use deref to pull the value out

Cool, thank you @gfredericks and @alexmiller 🙂

When writing middlewares for nrepl, is there a dynamic var bound to the id of the response I'm working on?

I'm registering a handler for writing to java.lang.System/out. Its registered once but the middleware puts the id of the registering request rather than the request it is handling further down

I'm actually a little surprised that I can't leave the Id off and let the middlewares further up merge it in if its nil

@dpsutton there's a dynamic var with the whole req I think

that might be only accessible during eval though

ah

CIDER registers the out redirects on id 7

and it looks like if you eval (.println (java.lang.System/out) "hi") it has that binding but its bound to id 7

nrepl middlewares always give me a headache

yeah

and its so slow to test

as you modify stuff, lein install, and see what happens

I think I tested faster than that

if you run lein repl out of the project with the middleware you should be able to do it that way

still a pain to test, just didn't have to go through maven

I have a file. It has ~ 20 functions. 4 are public. The rest are private. When unit testing, do I only test these 4 functions (because the rest are implementation details) or do I unit test all of them (helps me check bugs when refactoring) ?

what do you prefer?

do you need to test your private functions?

Last night, I went from s/foo/bar.clj and src/foo/bar_test.clj to a separate "test" directory; so I want to follow best practices.

best practice is test what needs to be tested

I'm a big fan of not 100% testing coverage but testing the bits that need to be

but that's orthogonal to private/public functions

Is there any chance of https://github.com/clojure-emacs/cider/blob/c698df1c905127f190efee3403822b2b09a811c3/doc/indent_spec.md becoming an official clojure thing? I think it’s a good standard, but standards like these are only as good as how much they’re used

I need a way to (1) dump out all function names and (2) for each function, print the number of arguments it takes (or if it's var arg). What libraries / functions should I use to do this?

@qqq Do you mind if I ask, what are you trying to build?

Do you want to operate at the level of source files or loaded namespaces?

@mingp: In haskell, because functions have fixed arity, we can do application without using ()

I would like to write a clojure macro (!! ... ) which does something similar

it looks at its arguments, figures out which ones are functions, and their input arity, then puts () at the right palces [this would ignore var arg functions and functions with multiple arity]

@qqq i did something similar in cider-nrepl

@dpsutton: awesome! can you please point me to the github lines? 🙂

https://github.com/clojure-emacs/cider-nrepl/blob/master/src/cider/nrepl/middleware/refresh.clj

check out the function zero-arity-callable

when refreshing a repl, it used to invoke functions that you set up for before and after

but only if they were zero arity

but this prevented functions where the args where entirely optional

(:argslist (meta func)) looks like the key

is the :arglists field of meta something that (1) you setup yourself or (2) something clojure does for you?

i believe that the defun form does it

so if you define a func with a (def func (fn [] ...)) it may be missing

not sure but i have seen some forms without it

(defn foo [a b] [a b]) (meta foo) is returning 'nil' for me

(meta #'foo) <-- shows it now

is there a way to force correct method dispatch on a java method in clojure? basically, if i do Paths.get("/tmp") works in java, but (Paths/get "/tmp") does not work in clojure. Here Paths` is java.nio.file.Paths. https://docs.oracle.com/javase/8/docs/api/java/nio/file/Paths.html

I get the following error:

@nux java varargs are a compiler illusion

there are two overloads: (string, []string) and (uri)

sorry, this is the exception: java.lang.ClassCastException: Cannot cast java.lang.String to java.net.URI

yeah, you need to explicitly pass an empty array

or maybe nil works too

@bbloom nil did not work, let me try with empty array

this works: (java.nio.file.Paths/get "x" (make-array String 0))

@bbloom thanks

using ns-interms, I have gotten a kv map of "symbol" -> "var object for symbol" for namespaces I care about. From this, how do I get the corresponding source of the symbol?

@qqq try (source source)

@bbloom: is that available in general, or do I need a ruynning repl?

that seems to be from the repl namespace

sure, so (use ‘clojure.repl) if your repl doesn’t already

1. source is returhing "source not found" even though the repl can see the var. 2. This doesn't seem to be how spectrum https://github.com/arohner/spectrum/tree/master/src/spectrum is doing it

any idea how spectrum is getting the source to analyze ?

try a fully qualified name

I used a full qualified name; let me show full interaction

user> [g.h0.fl/foo (source g.h0.fl/foo)] Source not found [#function[g.h0.fl/foo] nil]

so it sees the symbol, but can't get the source

source works by looking on the file system

so a file that the source looks up has to be evaluated via require or whatever it does that adds the :file metadata to the var

try (meta #‘nc) on your var, does it have :file, :line, :column?

(keys (meta #'g.h0.fl/foo)) (:arglists :line :column :file :name :ns)

and the values there are non-nil

are the values sensible?

yes, they're correct

the :file name exists

and is the file that contains the code of the module

well then i’m not quite sure - you’re going to have to dig 🙂

can anyone think of a performant way to query ~250 sqlite file databases that will be read from more-or-less equally ?i'm using a third party data source that happens to be structured in individual DBs... i've considered writing a script to merge them into postgress or something but it complicates the process of updating the resource dbs individually.

right now i've got a map of :id => db connection (using conman for a connection pool) and it works, but i think the overhead is probably huge (havent figured out how to measure it yet)

I guess one nice thing about loop is you can write recursive anonymous functions easily without having to use combinators

one thing I wish I could do with loop is (loop [obj] .... (modify obj) .... (modify obj) .... (recur))

instead of having to "compute the entire new obj all at once at the place of recur"

atom?

I was hoping "haskell state monad"

ive run into that before and figured i wasnt thinking functionally enough

didnt really know how to fix it though

sometimes, it's imperative though, like I have one part which is: increment counter add element ot raw data store update indexes

and clojure forces mt to compute all 3 at one spot

(with-local-vars [v1 "foo" v2 :bar] ...)

cool

@gfredericks How does using a local var compare with using a volatile or a plain atom? When is each preferable?

I've never seen anybody use a local var in my life, so I have no idea.

I think, but am not certain, that local vars are "cheaper" than atoms, because atoms are concurrency safe, and local vars should only be used in one thread?

Happy 2017

I am pprint on a large data structure. Is there a way to tell pprint to only print 20 lines?} I could possibly use (subs (with-out-str (pprint ...))) ... but then it has the problem that I end up printing it as a string

whereas, I'd want the \n to be actual newlines

Thanks!

@qqq You can use *print-level* and *print-length* to control that

for example (binding [*print-level* 5 *print-length* 5 ] (clojure.pprint/pprint {:a 1 :b {:c 2 :d (range 20) :e 3} :f 4}))

(adjust both values to see the effects)

@seancorfield: print-level is even more better

thanks!

(binding [*print-level* 3 *print-length* 2 ] (clojure.pprint/pprint {:a 1 :b {:c 2 :d (range 20) :e 3} :f 4}))
user=> {:a 1, :b {#, #, ...}, …}

clojure.tools.analyzer.jvm is the most verbose output I have yet to see

[xpost] Is there a way in honeysql or pure clojure.jdbc to cast a Postgres parameter? I receive :

PSQLException ERROR: column "personal" is of type jsonb but expression is of type character varying
  Hint: You will need to rewrite or cast the expression.
  Position: 66  org.postgresql.core.v3.QueryExecutorImpl.receiveErrorResponse (QueryExecutorImpl.java:2182)

I am passing a parameter as jsonb