@shader you won't be able to just do to-date/from-date, as that will mean when it deserializes a util.Date it wouldn't know whether it's a util.Date or joda date. but anyway I think you just need to extend Freezable2 https://github.com/ptaoussanis/nippy/blob/master/src/taoensso/nippy.clj#L316

and add something to custom-readers https://github.com/ptaoussanis/nippy/blob/master/src/taoensso/nippy.clj#L253 🙂

the macros extend-freeze and extend-thaw (https://github.com/ptaoussanis/nippy/blob/master/src/taoensso/nippy.clj#L1413) would be better than directly extending Freezable2

oh there you go,thanks tanzoniteblack

I understand (let [{:keys [a b]} ...]) however, what does

(let [{:keys [foo/a foo/b]} ...]}

the "foo/" part confuses me.

It binds the name a to the value at the key :foo/a, and likewise for b.

So, like, the map you'd be destructuring should look something like this: {:foo/a 1, :foo/b 2}.

(let [a (:foo/a obj) b (:foo/b obj)] ...) ?

Right.

Little more explanations https://clojure.org/guides/destructuring#_namespaced_keywords

@tap: some day, I need to read that end ot end instead of section by secton lazily 🙂

I think it’s ok. Lots of people are here to help

can someone explain please what does wrong during this back and forth number conversion?

(let [n 9007199254740995
      converted-back-and-forth (long (bigdec n))]
  (- converted-back-and-forth n))
=> 1

it's nowhere near Long/MAX_VALUE and BigDecimal is supposed to be arbitrary precision... 😕

9007199254740995 is 0x20000000000003 which is just 7bytes, 54bits

static public long longCast(Object x){
	if(x instanceof Integer || x instanceof Long)
		return ((Number) x).longValue();
	else if (x instanceof BigInt)
		{
		BigInt bi = (BigInt) x;
		if(bi.bipart == null)
			return bi.lpart;
		else
			throw new IllegalArgumentException("Value out of range for long: " + x);
		}
	else if (x instanceof BigInteger)
		{
		BigInteger bi = (BigInteger) x;
		if(bi.bitLength() < 64)
			return bi.longValue();
		else
			throw new IllegalArgumentException("Value out of range for long: " + x);
		}
	else if (x instanceof Byte || x instanceof Short)
	    return ((Number) x).longValue();
	else if (x instanceof Ratio)
	    return longCast(((Ratio)x).bigIntegerValue());
	else if (x instanceof Character)
	    return longCast(((Character) x).charValue());
	else
	    return longCast(((Number)x).doubleValue());
} 

ah, good point, i should have just looked up the definition of long sorry for the noise

looks like it is a case where it falls under the last else branch

(.doubleValue (bigdec 9007199254740995)) => 9.007199254740996E15

Floating point arithmetic looks like what is to blame here

(let [n 9007199254740995
      converted-back-and-forth (.longValue (bigdec n))]
  (- converted-back-and-forth n))
=> 0

actually my problem arose because i was just using clojure.core// on a BigDecimal, but i should just drop down BigDecimal methods and use .divideToIntegralValue...

How could I read a file and exit with true when a defined string is found?

@danielgrosse

(defn file-includes?
  [filename substring]
  (with-open [buf (BufferedReader. (FileReader. filename))]
    (if (some #(clojure.string/includes? % substring) (line-seq buf))
      true
      false)))

there are other ways like slurp but this will buffer, so you can use on larger files if desired. also, this is one line at a time and does not let you look for multi-line substrings, but this can be done too of course, just not with line-seq

@joshjones thank you

I have to access java objects, but want to write a macro to make it more convenient. But now I struggle with the quoting of the object name. How could I rewrite a JavaClass.object to (JavaClass/Object)so the object name could be passed as variable?

interesting challenge

shouldnt be hard

even for me 🙂

giving it a go

but bear with, Im a noob

@danielgrosse As a variable, then your only chance is reflection which is super slow. But a macro, yes then you could write (inv-java System.currenttimeMilis).

(defmacro inv-java! [x]
  (let [[c m] (str/split (str x) #"[.]")]
    (list (symbol (str c '/ (symbol m))))))
(inv-java! System.currentTimeMillis)

aww. too fast for me

was busy on the third line

had too google too many things

Btw, I read the question a bit differently, so for interests sake, this is what I did (btw it supports class names with dot notation)..

`(alias 'str 'clojure.string) (defmacro java[obj] (let [t (str/split (str obj) #"[.]")] (read-string (str (str/join "." (butlast t)) "/" (last t))))) (macroexpand '(java Java.lang.Math.abs)) (java Java.lang.Math.abs)`

paste fail, but w/e

I read the question as calling a static on it

Just to make it clear: I think all this is actually a terrible idea. 🙂 Good editors will understand (Xyz/zyx) and provide proper help/hints/arity-checks.

agreed 😄

but yay, my first answer to a macro question. noob feel happy.

Thank you both. Maybe I wrote the question in a wrong way. I try to achieve something like literal notation in javascript. So I could do something like

(defmacro get-object
 [objectname]
 (Javaclass/~objectname)
)

Currently I made a map where I associate the Object to a keyname.

how do you know what hte java class is? or is that baked into the specific macro?

my java is non-existent, so maybe Im being blonde

what would you expect to get out of , say (get-object abc) with your pseudocode?

What is the best DB technology for using with clojure ?

And why ?

imo, as a noob, mongodb

because the impedance of the document structure matches the forms in clojure very well

but again, Im a noob, so take with grain of salt

And why with example. Thanks you all guys. :)

Monger examples:

' (mc/insert-batch db "documents" [{ :first_name "John" :last_name "Lennon" } { :first_name "Paul" :last_name "McCartney" }])'

dammit how do you block quote code here

As with everything, it depends on your use case but I also think mongo is very good, specially for prototyping or smaller projects. You rarely need to think about performance and since every clojure datastructure can be serialized you can really just focus on building features.

;; returns documents with year field value of 1998, as Clojure maps (mc/find-maps db "documents" { :year 1998 })

@scknkkrer I prefer for Datomic

Monger has a DSL as well...

(with-collection db "movies" (find { :year { $lt 2010 $gte 2000 } :revenue { $gt 20000000 } }) (fields [ :year :title :producer :cast :budget :revenue ]) ;; note the use of sorted maps with sort (sort (sorted-map :revenue -1)) (skip 10) (limit 20) (hint "year-by-year-revenue-idx") (snapshot))

@scknkkrer I would not depend on the programming language when deciding for a database. These are two different things. When looking for a DB you might want to consider other stuff, like read / write performance, scalability, costs, support, do you need a key / value store, etc..

true that

and mongo is fairly good in most languages anyway, because most languages are not set based constructs

but based on hierarchical / tree data

haven't tried Datomic, but I hear good things

again, use case matters

With all these criterias, I am looking for idiomatic technology for clojure. Any cassandra user ?

afaik, Datomic is the idiomatic choice for clojure, I think

why do you want to choose based on idiomatic tech for the lang though

@scknkkrer yes

With datomic made by Rich Hickey I would also say that datomic is a natural choice. However, its not an open source license, except the free version.

@scknkkrer there are quite a few c* users, and we have quite decent bindings for c* as well: shameless plug > https://github.com/mpenet/alia

My primary db is C*, I can highly recommend alia+hayt

been meaning to look into cassandra one day, thanks for that 🙂

Super easy. Normally not a fan of clojure-DSL to represent SQL queries, but in case of C* it makes a lot of sense since the SQL dialect is so minimalistic. So I'm using hayt DSL with it.

then again if you hesitate between mongo, datomic and c*, maybe you need to have a clear idea of what are you requirements, these are all quite different in many ways

Is there any http library which expose data transfer rate information?

Or at a minimum have a hook to capture time of the first byte received and last byte received.

@mpenet, so, Is cassandra good choice for long-time ? I heard a bit of cassandra. And I looked a little bit.

depends on your requirement, it's not a silver bullet.

Can you give me a real-world example for fascinating me.

Yes, nothing is a silver bullet. That said, another team in our company is quite entrenched in cassandra, and it's a big system.

tell me what you want to do instead 🙂

With clojure.

they use cassandra for big data for a credit bureau

I don't know. Pick something.

if you don't need clustering, dealing with large volumes of data (or writes), maybe something else would fit better

(in short)

and mongo is a bit hybrid , so if you don't need the extra expressiveness of the queries, and hte sharding (which C* is better at anyway), then it's also not the best

that said, it's great to get going with

Yea, but early in project, I need implementation simplicity. Late in project, we will have really large datas...

How can I speel c*, any git repo. ?

I meant Cassandra.. assumed C* was the same thing (noob again)

for Cassandra look at alia and hayt like mentioned earlier

c* is cassandra

Wow. I need to read about cassandra, I think. Thanks you all.

has anyone had problems with the slingshot library and clojure 1.9?

I’m getting the following error: Caused by java.lang.UnsupportedOperationException Can't type hint a primitive local when I try to use the try+ and throw+ macros provided by the library

Anybody familiar with JDBC for MS SQL? Can't seem to get windows authentication working - copied ntlmauth.dll to my java library path as requested, and it still complains with "SSO Failed: Native SSPI library not loaded. Check the java.library.path"

fighting it for a while now

Im on a windows machine

http://stackoverflow.com/questions/13267764/i-o-error-sso-failed-native-sspi-library-not-loaded/16429490

they mention writing out the java.library.path and just confirm it is what you think it is

yup I followed that 🙂

aw bummer

i use ms sql with both the microsoft driver and now jtds

one thing I didnt do though, from that blog, let me try

but each time we use connection string login

k

still nothing. one thing is that the first answer says I should place 32 bit version in program files (x86) and the 64 bit in program files. but I dont have a 64 bit java folder under Program Files

so that doesnt really add up

and I checked hte library path, found that C:\Program files (x86)\ is in there.. tried that, nothing

I bet rebooting will fix. sigh

dll might not load if the dlls that it depends upon cannot load. Its sometimes helpful to run depends.exe to analyse dll dependencies, and tell you what’s missing

I made the above macro that wraps defmethod, it works fine in the repl during development but when running my program with lein run or running the jar I get"No method in multimethod 'execute'" for the tasks I defined

I should mention that the deftask calls are made in different namespaces, and they are required in the core namespace

tx @benha . good point there

it seems like it is not related to the macro but rather the fact that the defmethods are not in the same namespace since it doesn't work either when I use the result of the macro-expension directly

i want to subclass a java class to customize one of its methods, whats the easiest way to do this from clojure?

hi I'm using clj-http.client and cljs-http.client and getting 403 (Forbidden) error in server when using the methods PUT and POST

GET is working well, btw

@thomaas Maybe you are missing the CRSF token?

@sveri, could you paste me some link about that?

@chrisblom have you tried reify? https://clojuredocs.org/clojure.core/reify

reify only works for interfaces or protocols, i need to extend a class

public interface Lock

@tomaas Hm, in general? I think that is very good to be found by google. In case you use luminus here are the docs about it: http://www.luminusweb.net/docs/security.md#cross_site_request_forgery_protection

reify the relavant interface that Reentrant Lock implements?

Its just a guess that this might be the problem as GET requests dont expect a CSRF token, but POST / PUT / ... do

@dpsutton that not really what i'm looking for, i need a subclass of ReentrantLock, as other parts of my code expect a ReentrantLock or a subclass

@chrisblom I think you don't need to write this in the parameter vector for lock and unlock

hello everyone, is there a way to create a type alias in clojure?

id like to do the following:

(s/defschema Benefit
  {:cpf s/Str
   :nomeSegurado s/Str
   :emailSegurado s/Str
   :atendimentoId s/Str
   :referenciaId s/Int
   :clausula s/Str
   :dataInicioBeneficio s/Str
   :dataFimBeneficio s/Str
   :motivo s/Str})

(defn my-function [^String bla ^Benefit foo] ...)

my guess is that this gains you nothing

and also that you want spec

but my understanding of compiler hints is that is prevents reflection warnings

since you don't have an actual underlying type you won't prevent this, so you get nothing out of the annotation

could definitely be wrong though

id like to anotate with something, and the compiler complains of this code .. maybe i should try to integrate compojure with spec instead of prismatic

what does the compiler say?

@plins If you want something to check your types at compile time you need core.typed

Unable to resolve classname: Benefit

yeah

there is no class called Benefit

@sveri i just want to document things a little bit, is meta the right way?

docstring?

should i include types?

@plins from clojure 1.9 spec is the way to go

hmm ok thanks 🙂

if you want programmatic guarantees, use spec and if you want hints to the programmer use the doc string

apparently using the :require [ns1 [subns1 [a] [b] [c]]] for in the ns declaration was my problem for the multimethods not being present at runtime

but i doubt an end consumer would ever look in the meta of your function for usage hints

but would greatly appreciate documentation explaining types and an example

im thinking about my team members looking at my code

so im going for docstrings i guess 😃

speaking of type hints, does anyone know how to generate an anonymous fn form with a type-hinted return value?

I've figured out from the docs that you can add type hints to params by putting {:tag type} metadata on the param symbols, but not sure what to do about return types

docs (https://clojure.org/reference/java_interop#typehints) say to put a ^type tag before the arg vector...

does that mean the metadata should go on the arg vector?

what would a type hint on the return value do?

you presumably want to give a type hint to the consumer of that value, yes?

good question

Hi all, is there a way to do deep clone of hash ?

I guess type hinting on both the return value and params might be unnecessary...

so just type hint the function that takes it as a param can work for you?

also, check this out to make sure you are solving an actual problem

https://clojuredocs.org/clojure.core/*warn-on-reflection*

the functions that will take this are java methods, so I guess they're "type-hinting" by default

my understanding is that type hinting does less than you think, so check if you are getting reflection warnings in the first place

I've already gotten my interop code to a place where it has no reflection warnings

then my understanding (and i could be wrong) is that type hinting will gain you nothing

I guess you're right

sorry, not very familiar with java's type system or strongly typed languages in general

me either. it's just what i've picked up from hre

alex miller corrects a lot of misunderstandings of type hinting a lot so i've picked up a few things from him

@dpsutton a type hint on the return value makes it so that consumers of your function don't have to type hint it every time

ah. thanks for the correction

also if it's a primitive type hint, it will optimize the emitted code to potentially avoid extra boxing

will he see a benefit if he's already generating no reflection warnings?

@bronsa so should the type metadata go on the arg vector then?

yes

.. except for deftype methods

but for everything else, yes

if you want to type hint the return value

if you need to type hint the return value

@dpsutton not necessarily, but possibly if he's using primitive math

thanks for clearing that up

I guess i was helping to spread the confusion about it 🙂

if neither *warn-on-reflection true nor *unchecked-math* :warn-on-boxed complain

then the type hint is usually useless

this is the context, in case anyone's curious:

(defn- method->fn
  "Takes a method and returns a type hinted anonymous `fn` form that
  calls the method."
  [{:keys [parameter-types declaring-class name]
    :as method}]
  (let [params (-> parameter-types
                   count
                   (repeatedly gensym))
        params-with-types (map (fn [param type]
                                 (with-meta param {:tag type}))
                               params
                               parameter-types)
        obj `obj#
        obj-with-type (with-meta obj {:tag declaring-class})]
    `(fn [~obj-with-type ~@params-with-types]
       (. ~obj ~name ~@params))))

type hinting anonymous functions is useless

the hint will be ignored

type hinting is a compile time thing

anonymous functions exist at runtime

without the type hints though, I get reflection warnings

well, type hinting the return value of anonymous functions, I should say

the args type hints are still useful

ah ok

is type hinting the return value still useless if the anonymous function is stored as a constant?

yes

type hinting the return value of a function is only useful if you use defn

using (def x (fn ^Foo [] (Foo.))) rather than (defn x ^Foo [] (Foo.)) will make the return type hint useless

ah, gotcha

@bronsa This yields 1 vs 1.0. I'm confused

(let [x (fn ^long [x] x)]
  (x 1.0))
(let [x (fn [x] x)]
  (x 1.0))

yeah..

it's confusing

so, here's how it works

I lied before by saying that return type hints on anon functions are useless -- what I meant is, they don't get propagated outside the fn

so in your last example, that type hint acts as a cast for the return value of that function

but

any function that uses (x 1.0) as its argument

won't know that x returns a long

because it's only local info

I see.

type hinting has a lot of quircks

esp wrt primitive types

So the type hint will still properly work for the function itself. Wrt, the compiler avoiding reflection for that anonymous function, right?

so, if you had

public class Foo { public void a (long x) { System.out.println("1");} public void a (Object x) {System.out.println("2")}}

I agree, type hinting with primitives took me a while to get right. I wish there was more documentation about it. Cause me nice bug.

that effect of casting the return value will only happen for primitive type hints

and you invoked

(defn a ^long [] 1) (Foo/a (a))

but if you tried with

(def a (fn ^long [] 1)) (Foo/a (a))

Also, if I run:

(binding [*print-meta* true]
  (pr-str
    (macroexpand-1
      '(defn xy ^String [x] x))))

The defn just expands to def + fn.

what you don't see is the metadata that gets attached to the Var

use prn rather than pr-str

you'll see the arglists metadata has the type hint

Oh yes, now I got it.

oh my bad, pr-str works just fine

That all makes sense now.

but :arglists is what the compiler uses

which it only looks up on vars

and only attaches to vars created via defn

also what @hiredman said, type hinting only acts as casts for primitives

.. and only in some scenarios

in other scenarios it forgets to cast and just causes bad bytecode to be emitted

in others it causes the compiler to complain and not compile your code at all

That's new...

¯\(ツ)/¯

anyone know if you can (or even should be) redefining with-refdefs a java method like java.utils.UUID/randomUUID?

you can't

with-redefs only works with clojure Vars

and even then, it's usually not a good idea

and it won't work all the time

thanks, we'll probably define another function that calls out and redefine our function

why are you testing side effects like a random uuid?

I would pass that in as input so the function maintains purity

I’m working with a web app (via luminus). I’m currently trying to apply access rules via the buddy-auth warp-access-rules middleware. Though it seems my changes only take place when I restart my server. Is this expected?

@mruzekw if you take a look at https://github.com/funcool/buddy-auth/blob/master/src/buddy/auth/accessrules.clj#L320 you can see that it compiles your rules and then returns an anonymous function that is actually being used as the middleware, that means that any changes you make to your access rules won't take effect until they've been recompiled and your server has the new function handed in as a middleware

Thanks @tanzoniteblack I looked around but wasn’t able to make full sense of the code yet.

But that makes sense now

so yes, that's expected. You should be able to work around this though if you do something like (def my-access-rules (wrap-access-rules <rules>)) and then use the symbol my-access-rules in your middleware

that should cause your rules to take effect whenever you redefine my-access-rules

but that might depend on which http server you're using 😄

Using immutant

Thanks, I’ll try this out

don't actually know the subtleties of how immutant works well enough to tell you if that should or shouldn't work, but I'd be interested to hear back what you find

While (subseq sc >= s <= e) returns a lazy seq, (subseq sc >= s) does not - returns a PersistentTreeMap$Seq. Does anyone know if the latter behaves similar to lazy seq? That is, does it walk the tree step by step as asked for values or does it just make the whole thing up front?? Thanks

apologies if this is the wrong place to ask - is there a recommended clojure postgresql package?

i'm looking at clojure.java.jdbc

that or http://funcool.github.io/clojure.jdbc/latest/

does anyone here use immutant/undertow as an http server?

@joshjones: I do, but I wouldn't be unbiased :) What's up?

I'm just trying to figure out how to read the :body of a POST request. The body in the request is:

#object[io.undertow.io.UndertowInputStream 0x46598eaa "io.undertow.io.UndertowInputStream@46598eaa"],

@piotr2b what do you mean by multi-tails? are you saying you cannot have more than one tail position in clojurescript? because that is not correct

Do I really need to directly use the read methods, etc., in this UndertowInputStream class? Or is there a more straightforward way to do it? @tcrawley -- further, I see that the stream is closed, so I don't know how to get the data

your algorithm is not tail recursive, so recur won't work, has nothing to do with clojurescript

@joshjones what type of data is the body? if it's a string, you can just (slurp (:body request))

@tcrawley thanks so much, that did it -- geez louise, I did not notice that it extends java.io.InputStream... derp! 🙂

happy to help

also, is tail recursion optimization beneficial outside of strict evaluation? if this is a lazy structure, does it matter if its tail recursive or not?

@dpsutton good point -- seems that it can just be constructed with the current lazy-seq / cons pattern of constructing a lazy seq

no

generate a lazy sequence is almost never properly tail recursive

generating

or I should say, generating a sequence

(actually the second code doesn’t work, there is an infinite loop in it)

because you have cons in the tail position

because you are building something up

your function is almost never in the tail position

your docstring is also incorrect about laziness

@catilac I'm currently working on a project that uses a postgresql database and I'm using clojure.java.jdbc and also com.layerware/hugsql

well

ok, it is sort of correct, sort of not correct

@piotr2b would recommend to take a look at https://clojure.org/reference/lazy#_recipe_how_to_write_lazy_sequence_functions_in_new_model , see if it might fit your use case

and you definitely cannot use recur like you are

Actually my question is whether this can be rewritten with tail recursion

Oops, I’m sorry about the mistake in docstring. Could you elaborate on this?

every recursive function can be rewritten to be tail recursive

it is sort of lazy because you do call map, but nothing in your function is lazy

but i think we were thinking tail recursion and lazy are kinda at odds with one another

you call to recur in the second example is nonsense

#() is a new function, so you are recuring in that new function

you almost certainly don't want it to be tail recursive

(that function)

because to be properly tail recursive it can't be lazy

constructing a lazy sequence (that is really lazy) and recur are mutually exclusive things

@joshjones thanks for your links, I gonna read it

http://dev.clojure.org/jira/browse/CLJ-1906

ok thanks for referencing this ticket @hiredman.

I’m starting a referral program for introducing new Clojure and other “Functional” Engineering Professionals for new gigs. Working with several customers hiring. Anyone interested in learning more, send me a message.

hi all, actually I also have a project with Clojure (Datomic+Cassandra as well) if anybody in the EU and interested msg me

#jobs is more appropriate for job announcements

hoops

my bad

is there a version of (some) that works on a (clojure.walk/postwalk) instead of (map) ?

i basically want to walk through a datastucture and perform an update on the first element that meets a predicate while leaving all the other elements alone

that isn't some

that is if

(if (p node) (f node) node)

you could shorten it, if that is your thing

yea but if i use that with postwalk, itll update all the elements that match the pred

i only want to do it on the first on ethat matches

which isn't how postwalk works

you may want to look at zippers

@yedi have you looked at https://github.com/nathanmarz/specter ?

ooo specter looks promising, i might try that. though ideally i wouldn't have to pull in another lib

ill check out zippers and see if i can leverage those, if not ill just use specter

you could make your predicate match only once but that seems like a waste of resources

with zipper you can walk the structure until pred matches, modify that location, then return the structure

a shudder, stateful predicate

apparently hiccup has built-in zipper helpers, so im thinkin this is the way to go