Clojurians Log v2

json doesn't have a date type, so how is a json decoder supposed to know which string fields to treat as special dates?

If you could provide value transformer function somehow... (fn [x] (if (inst-str? x) (Instant/parse x) x)) I'll probably use clojure.walk on decoded edn value

you could use Jackson API directly:

(.readValue j/default-object-mapper (j/write-value-as-string (java.util.Date. 0)) java.util.Date)
; => #inst"1970-01-01T00:00:00.000-00:00"

… or describe your data with a schema/spec and derive the transformation using it. with #malli:

(require '[malli.core :as m])
(require '[malli.transform :as mt])

(def User
  [:map
   [:name string?]
   [:created inst?]
   [:tags [:set keyword?]]])

(def json->user (m/decoder User (mt/json-transformer)))

(def user
  {:name "liisa",
   :created #inst"2021-01-01T00:00:00.000-00:00"
   :tags #{:clj :cljs}})

(-> user
    (j/write-value-as-string)
    (j/read-value j/keyword-keys-object-mapper)
    (json->user)
    (= user))
; => true

it’s also much faster than a generic walk over the data.

I'll try, thx

Has anybody heard of Android development being done with Clojure? Just wondered ..

I think there have been several attempts over the years but I think Clojure's dynamic nature and startup time have made it difficult. I think this is the most recently-maintained effort https://github.com/clojure-android/neko ?

I think people mostly go down the ClojureScript path and target React Native under the hood?

soon enough we would have clojuredart (people are working on it)

that would allow us to use flutter

Building a microservice for kubernetes. Where should I be starting so basic things are properly plumbed. Logging, signal handling, request handlers etc?

I'm poking with futures, in this example:

(comment
  (def f3 (future (do (Thread/sleep 5000) 1)))
  (deref f3 3000 :timeout))

(comment
  (def f2 (future
            (for [x (range 5)]
              (do
                (Thread/sleep 1000)
                x))))
  (deref f2 3000 :timeout))

for is lazy.

ah, right! thanks! Wrapping in doall does the trick.

i need some inspiration. i have a bunch of validation to do involving Java classes that can throw, and I would like to both preserve any error message, and still have my is-thing-valid?functions return truthy/nil.

the "obvious" solution is perhaps to lift the return of is-thing-valid? into a map like {:valid-thing? false :message "An error has ..."}

Hey @alexmiller I just saw your talk with ClojureD about tools.build, and it looks really good! I do have one question about it: are there any plans to be able to include source dependencies and "prepare" them by enabling or disabling aliases? I saw there was some way to run an alias to do like a build step, but in one of my projects I have different native dependencies which need to be included based on the platform, and I'd like to be able to provide those by specifying aliases with the dependency.

I'm trying to wrap slingshot's try+/catch in a macro and it doesn't work for some reason. If I just call macroexpand-1 and capture and eval its output it works fine. But trying to eval it directly produces an error my.ns/_e - failed: map? at: [:bindings :form :map-destructure] spec: :clojure.core.specs.alpha/map-special-binding What am I doing wrong?

(defmacro test-it [name options params & body]
  `(defn ~name ~params
     (try+
      (println "Test it")
      ~@body
      (catch map? _e (throw (RuntimeException. "error" (:throwable &throw-context)))))))

;; this throws an error:
;; my.ns/_e - failed: map? at: [:bindings :form :map-destructure] spec: :clojure.core.specs.alpha/map-special-binding
(test-it my-test {} [x y]
         (/ x y))

;; this can be evaluated without problems
(defn my-test [x y]
  (try+
    (println "Test it")
    (/ x y)
    (catch
        map?
        _e
      (throw
       (java.lang.RuntimeException.
        "error"
        (:throwable &throw-context))))))

Your problem is that _e is being expanded to a qualified symbol.

If you want a symbol that's local to that part of the code and can't be referenced by users of the macro, you should append # to the end of the symbol name, which will turn it into an auto-gensym.

This is because you're quoting the form with `

If you want a symbol with a specific name (maybe to allow users to reference it), you'd need to do it like so: ~'_e

Got it: I For this one, using ~_e` and ~'&throwable works. Now I need to find out the reason for another issue I have when I wrap it with a hystrix's defcommand function O:-)

If _e isn't supposed to be used, then prefer _e#

This is so that if you change the macro and allow user code inside the block where that's bound, it won't shadow any bindings made by the user.

Yes, that's a good idea!

from the ex-datadoc: Returns exception data (a map) if ex is an IExceptionInfo. Otherwise returns nil.

in what cases would it be nil? would it be nil for any typical java exception, which doesn't implement IExceptionInfo?

could or should it be used to separate unexpected crash-type errors from other types of errors?

> would it be nil for any typical java exception, which doesn't implement `IExceptionInfo`? Yes. Or any subtype of IExceptionInfo that allows nil to be used for additional data (`clojure.lang.ExceptionInfo` does not allow that). With that being said, the answer to your last question depends on what "expected" means. If you rely on some specific field in ex-data, then perhaps. But if you just want to check for IExceptionInfo, then use instance?.

it comes up in handling HTTP responses since f.ex. clj-httpstuffs the error response in there. but should it be some unexpected crash while handling the request it's nil, and I have to handle that and return a 500myself.

you mean aliases that vary depending on the project using it?

I am interested in understanding what the community and Clojure core team do to track and manage potential security issues with Clojure core and Clojure libraries. Are there any particular tools and other resources that are helpful? I appreciate that libraries can be kept up to date with lein ancient or clojure -M:outdated . Is there any recommended ways of tracking issues that are discovered / reported? I see that there is the Open Web Application Security Project https://owasp.org/ for general info and tracking I'm really just looking to have a better understanding of security around Clojure, the libraries I use and the applications I build. Thank you.

Are you looking for a tool like https://github.com/rm-hull/lein-nvdlein-nvd ?

I guess the link should be: https://github.com/rm-hull/lein-nvd

Yes, specifically I have a game engine that depends on native libraries. I'd like to be able to write an application that depends on the game engine and tells it which dependency set to include based on an alias in the dependent application.

The way I have this working now is that the game engine has two aliases, :windows, and :linux, and these two aliases include native dependency jars. I'd like to be able to say io.github.IGJoshua/s-expresso {:git/tag "v1.0.0" :deps/aliases [:windows]} or something to be able to "enable" those dependencies.

And I want to know if there are plans to enable something like this within tools.build or new releases of tools.deps. It seems like it'd be within the scope, but I don't know if it's a direction you're intending to go.

lein-nvd has worked pretty well for me over the last couple years :) recently I contributed to it a new mode when using Lein as it can be subject to some problematic ambiguities. default deps.edn usage is less prone (but not free) of them it's all reflected in the README

this is a bit beyond what will be provided initially - the alias to use for prepping is fixed, so either you would need two wrapper projects that did the right thing, or a prep program that knew based on some other information (presence of something on the classpath, env var, etc) what to do

hello! I have a kinda stupid question, I'm trying to use honeysql an it's helopers, I required them with [honeysql.helpers :as sql] and called like (sql/select ...)

I see. That'd mean for now it'd be easier to just include a snippet to tell the user to include in their deps.

but when I tried to use (sql/select-distinct-on ..) it was crashing saying that this function does not exist in the package

so I went to the documentation and saw that the package was actually honeysql.sql.helpers

tried to require that way, but now I'm getting Could not locate honeysql/sql/helpers__init.class, honeysql/sql/helpers.clj or honeysql/sql/helpers.cljc on classpath.

what am I doing wrong?

@matheusashton There's also a #honeysql channel where you can ask questions about it

check out how the ns is actually called, you got a small typo there https://github.com/seancorfield/honeysql/blob/develop/src/honey/sql/helpers.cljc

oh sorry, didn't know, I'll ask there

oh, there was a major version bump, I just saw it now due to your comment, that may be it, thanks!

no worries, you can ask here too, but the honeysql-knowledge is more concentrated there ;)

@matheusashton The 2.0 version uses different namespaces (and different coordinates) so that people can use both versions side-by-side which makes it easier to migrate piecemeal from 1.0 to 2.0.

yep, I've just noticed, thanks

random thought, but it would be so nice if java.util.concurrent.ExecutionException told you the id of the thread that threw the exception (e.g. from a (future)) would speed up some debugging

yes, lots of exceptions are like that: this is wrong, type x. yes, but perhaps tell me the concrete x?

j.u.c.EE always wraps another exception, which should point to context pretty well

it points to the context in terms of a stacktrace N threads can share a very similar (or even identical?) stacktrace. Then thread id would make all the difference (assuming intentful naming of threads)

better to catch -> annotate -> rethrow

oh yes I'm happy this has improved with spec and related libs

thread ID is one of infinitely many different things that can enrich an error context

good call, I should just make a wrapper if I'm so inclined :)