Is there a way to somehow iterate over lexical bindings inside body of a function or a let form (that should happen in the context of macro expansion of course)?

I'm pretty sure I have seen some Clojure debugger-type library thingy that might be implemented by doing that, or something similar. Let's see if I can remember which one it is...

I haven't tried this, but the code that includes a definition of 'local-bindings' in this StackOverflow Q&A might be relevant? https://stackoverflow.com/questions/2352020/debugging-in-clojure

That code looks to me to be specific to Clojure/JVM, and wouldn't work for ClojureScript, because it is reaching into internals of Clojure/JVM compiler specific to it.

@andy.fingerhut Thanks, this is exactly what I was looking for.

Google search for clojure.lang.compiler/LOCAL_ENV turns up other places it has been used for similar purposes, too, in case you want other examples.

i was playing with alex and george's debug repl recently - fwiw, the latest version doesn't use compiler internals. still i didn't find it to work as-is for cljs -- though it seems likely it would work for clojure clr. i believe it uses &env now: https://github.com/GeorgeJahad/debug-repl/blob/master/src/alex_and_georges/debug_repl.clj#L29_L46

fwiw. the speculation bit there seems fairly in line with what i saw combing through past clojure irc logs 🙂

@sogaiu thanks, I just learned about &env

yes &env I think is what you're looking for

unfortunately, using &env breaks using macroexpand to test 😕

Hello folks! When I run my function in the REPL I get > Execution error (StackOverflowError) at (REPL:1). but the stack trace contains nothing of my code, only many clojure.core$concat, $seq and clojure.lang.LazySeq. How do I troubleshoot this? How do I find out where in my code am I creating this??? The stack trace:

[[clojure.lang.LazySeq sval "LazySeq.java" 42]
 [clojure.lang.LazySeq seq "LazySeq.java" 51]
 [clojure.lang.RT seq "RT.java" 535]
 [clojure.core$seq__5402 invokeStatic "core.clj" 137]
 [clojure.core$concat$fn__5493 invoke "core.clj" 725]
 [clojure.lang.LazySeq sval "LazySeq.java" 42]
 ... (these few repeated many times over)
 [clojure.lang.LazySeq seq "LazySeq.java" 51]
 [clojure.lang.RT seq "RT.java" 535]
 [clojure.core$seq__5402 invokeStatic "core.clj" 137]])

can you show the code or the error message at least ? 😄

only having the stack trace is very little to go on

ah, now i get it, the stackoverflow is your error message ... any chance you built endless recursion there ?

https://stuartsierra.com/2015/04/26/clojure-donts-concat this might give you some insight. if you built up a sequence using concats, you end up with too many lazy thunks to realize the sequence later and run out of stack

Thank. I am pretty sure I do not build a sequence manually anywhere but I do operate on data using lazy sequence operations. I cannot really share the code as there is few hundred lines of code involved. I just call my (fetch-and-process-all-invoices). The error message is there ☝️ and it is useless: Execution error (StackOverflowError) at (REPL:1).

The only hypothesis is that the error actually happens after my function run, when REPL tries to display its output, that could explain why there is only clojure.* in the stacktrace.

if that was true, you would see a different error message (not Execution error)

Execution error indicates it happening during the E of REPL

You are right. But why doesn't the stack trace have anything of my code? How do I find where is this happening?

do you use concat anywhere?

Yes, at a few places, I will look into that / alternatives. (And reread the article!) Ah, I guess I am hitting this > and the accumulated stack frames of seq prevent us from seeing where the error originated

Hm, looking through the code I do not see any suspicious use of concat with multiple arguments, mostly just (merge-with concat ...) etc. I have to re-check...

are you doing that in a loop / recursion?

why clojure doesn’t have a list literal or a symbol literal? > Clojure has no literal representation for Lists. If you write a List out in your program, it is interpreted as code. At first glance, it appears that quoting the List is a good solution. However, if you quote a List, the elements of the List won’t be evaluated. That means that a quoted List is only useful for Lists of literals. https://purelyfunctional.tv/guide/clojure-collections/ Clojure is data oriented, I was wondering why there is no literal for a symbol or a list of data. One supposition would be homoiconicity could not be achieved, but I could not explain why.

#inst "1985-04-12"
=> #inst"1985-04-12T00:00:00.000-00:00"
#{:a}
=> #{:a}
[1 2 3]
=> [1 2 3]
{:a 1}
=> {:a 1}
'(1 2 3)
=> (1 2 3)
(1 2 3) 
=> error

I don't understand the question. '(a b c) is a list literal containing 3 symbol literals

It needs quoting to avoid being evaluated in a REPL, unlike a map or set literal, that is true, because code is inside of lists, too.

Are you asking why it needs quoting at the REPL to avoid evaluation, and maps and sets do not?

Maybe a little more -- all of those expressions are being read, then evaluated, and the result of evaluation printed, at the REPL.

maps and sets are read, and during evaluation, they evaluate to themselves, as do keywords, numbers, and #inst values. All subexpressions of the map and set literals are also evaluated, recursively, too, so if a map or set literal contained a parenthesized expression, it would be evaluated, and present the same difference in behavior as if the parenthesized expression were typed as the whole expression.

If lists also always evaluated to themselves, like maps and sets did, then they could not be used to contains code for evaluation -- you would need some other syntax for that? (not sure if that makes sense)

yes that make sense! Seems like it would be handy* to have another syntax for that

but I was wondering if it could be possible

I should be a little more precise, even without the quote in front (a b c) is a list literal containing 3 symbol literals. If you only do read on that, no eval, you get a list.

It is the eval that causes an error if any of a b or c are not defined with values, or if the value of a is not something that can be called as a function or macro.

user=> (read-string "(a b c)")
(a b c)
user=> (eval (read-string "(a b c)"))
Syntax error compiling at (REPL:1:1).
Unable to resolve symbol: a in this context

As opposed to this, where the set evaluates to itself, and so do 1, 2, and 3:

user=> (read-string "#{1 2 3}")
#{1 3 2}
user=> (eval (read-string "#{1 2 3}"))
#{1 3 2}

In the next example, reading returns a set containing 3 symbols. Without eval, no problem, you get back such a set. Eval of the set returns a set, but also recursively calls eval on all elements, which is when it finds that a is undefined:

user=> (read-string "#{a b c}")
#{a c b}
user=> (eval (read-string "#{a b c}"))
Syntax error compiling at (REPL:1:1).
Unable to resolve symbol: a in this context

Reading a list or set or map are all pretty consistent behavior, and the syntax for each are all literal syntax for those data structures. The rules for eval behavior on a list are different than the rules for eval behavior on a set or map -- the first element of the list is special.

I am stuck trying to proxy an java.io.OutputStream. It has two different write methods, one overloaded with a byte array and one overloaded with an integer. I seem to type hint things correctly to tell the compiler how to implement the write methods.

My solution right now is to write a java class that derives from output stream and names those methods differently as abstract methods.

@chris441 you may have to test types inside the method impl

That would lead to an abstract method error as you will have only overridden one of the two methods.

Most likely the byte[] one.

that isn't true

(def x
 (proxy [java.io.OutputStream] []
  (write
    ([o] (println (class o)))
    ([o s len]))))
#'user/x
user=> (.write ^java.io.OutputStream x 2)
java.lang.Integer
nil
user=> (.write ^java.io.OutputStream x ^bytes (byte-array 0))
[B

I've proxy'ed outputstreams a few times, and it is always the same, type check in the write method

yup

huh. Well, you are overriding the object method (byte[]) one I think. I wonder how that is working or if just no one calls the int method.

like, ghadi literal posted a transcript of a repl session showing you it working for both the int write method and the byte[] write method

stop saying it doesn't work

We don't know what the reflected method called.

we

we do

1. it isn't reflected

2. even if it was, it prints out the type so we know

So then when you proxied did it override all methods with one signature and call your code for both?

in the output you can see that it's calling the code I provided in both cases of input

Yes for sure, agreed.

clojure's proxy is strictly name based

I think so -- overrides all by the same name

e.g. you give it a function body (which may have multiple arities) and all methods with that name will call that function

@lukaszkorecki Any chance you have some sample code? This stuff is confusing

@josh_tackett sure, let me find a version I can share

@josh_tackett here you go: https://gist.github.com/lukaszkorecki/f978452d1b3e88c3d88832d3defd5e74 - usual disclaimer, code is as-is etc etc etc

👍

also, this works only with java 8 - because it relies on javax.xml.bind DatatypeConverter

I'm going to try to extend buddy-core to handle :aes256-cbc-hmac-sha256

so hex-pair-to-byte would have to be tweaked for java 11 I think

got it, i will circle back to this if I can't extend buddy-core

I don't think I ever used it - usually BouncyCastle does everything I needed, including PGP and the API is pretty straightforward, what's confusing is the overall cryptography stuff 😉

beware -- neither BouncyCastle nor buddy are "misuse resistant" libraries like libsodium or google's tink

pgp tends to be less straight forward once you get to multiple objects, subkeys etc. 🙂 at least if you do it with bouncy castle, there are lots of pitfalls

then again ... it does the job ... so i don't complain 🙂

Luckily the PGP stuff I was working on was pretty light weight and in a controlled environment, implementing production-grade, customer facing features is indeed a task that needs a lot of attention and verification.

@lukaszkorecki I'm getting this error: java.lang.ClassNotFoundException: javax.crypto

@lukaszkorecki actually it can't find any of the libs I'm trying to import haha

because you aren't using the his ns verbatim, you are trying to turn it into an (import ...) call in the repl, but failed to quote the imported classes

hah

or not

apparently recent enough clojure versions don't require the quoting, which is wild

what a world

(:import java.security.SecureRandom (javax.crypto Cipher SecretKeyFactory) (javax.crypto.spec SecretKeySpec IvParameterSpec PBEKeySpec) (javax.xml.bind DatatypeConverter) (org.bouncycastle.jce.provider BouncyCastleProvider) (java.security Security)))

so he does it with an import

but that isn't import

that is :import

:import is a directive to the ns macro, import is a macro you can call at the repl

(ns skyscraper.scrape.encryption
  (:import javax.crypto Cipher KeyGenerator SecretKey
           javax.crypto.spec SecretKeySpec
           java.security SecureRandom
           org.apache.commons.codec.binary Base64))

(defn bytes [s]
  (.getBytes s "UTF-8"))

(defn base64 [b]
  (Base64/encodeBase64String b))

(defn debase64 [s]
  (Base64/decodeBase64 (bytes s)))

(defn get-raw-key [seed]
  (let [keygen (KeyGenerator/getInstance "AES")
        sr (SecureRandom/getInstance "SHA1PRNG")]
    (.setSeed sr (bytes seed))
    (.init keygen 128 sr)
    (.. keygen generateKey getEncoded)))

(defn get-cipher [mode seed]
  (let [key-spec (SecretKeySpec. (get-raw-key seed) "AES")
        cipher (Cipher/getInstance "AES")]
    (.init cipher mode key-spec)
    cipher))

(defn encrypt [text key]
  (let [bytes (bytes text)
        cipher (get-cipher Cipher/ENCRYPT_MODE key)]
    (base64 (.doFinal cipher bytes))))

(defn decrypt [text key]
  (let [cipher (get-cipher Cipher/DECRYPT_MODE key)]
    (String. (.doFinal cipher (debase64 text)))))

ya I'm using :import

no

that :import is incorrect

ok, got it. how do I fix it?

you want (:import (PACKAGENAME SIMPLECLASSNAME1 SIMPLECLASSNAME2 ))

you removed all the grouping of classes in packages

(ns skyscraper.scrape.encryption
  (:import (javax.crypto Cipher KeyGenerator SecretKey)
           (javax.crypto.spec SecretKeySpec)
           (java.security SecureRandom)
           (org.apache.commons.codec.binary Base64)))

correct

👍 thank you

you don't need apache commons for base64, java has a built in class for it

https://docs.oracle.com/javase/8/docs/api/java/util/Base64.html

is it a bad idea to be using vectors like so?

(ns skyscraper.scrape.encryption
  (:import [javax.crypto Cipher KeyGenerator SecretKey]
           [javax.crypto.spec SecretKeySpec]
           [java.security SecureRandom]
           [org.apache.commons.codec.binary Base64]))

either is fine

I prefer vectors

@jumar a compatible alternative to jasypt in clojure: https://github.com/pyr/jcipher