Clojurians Log v2

I’ve always used parinfer. What’s a good alternative I should learn?

https://github.com/DogLooksGood/parinfer-mode

if you emacs …. with vim I use https://www.vim.org/scripts/script.php?script_id=3998

lispy for emacs https://github.com/abo-abo/lispy

I saw this solution for clearing the terminal. It works on Linux (gnome terminal) but I am curious if it is portable:

(print "\033[H\033[2J")
  (flush)

no mention of paredit? that's the gold standard

It works on macOS as well, I will try in windows cmd.exe...

doesn't work in cmd.exe with the native console, but it does work if you use the new Terminal app on Windows

search "vt100 clear screen"

so it seems to be fairly portable, assuming everyone will use the new terminal on Windows

Great, otherwise the alternative is to detect the OS, then run the appropriate shell command... This quasi oneliner is preferable

@adam678 On windows I think you can detect the new terminal usage using an environment variable

user=> (System/getenv "WT_SESSION")
"c1d08d7a-69ae-4dc2-93e6-8689e6789dd1"

I take it if you're NOT in the new terminal, this will return an empty string instead of a UUID, correct?

it will return nil probably

Yeah depends what editor you use. In my case it's Emacs/EVIL, so you go with Lispy/Lispyville. Lispyville integrates nicely with EVIL.

When developing I'd find it useful to have a get/`get-in` that errors in case the key isn't found. Is something like that built in?

get m k <pivot>) check pivot

or find first but it does not work for get-in

I guess I could just (get m k (Exception.)).

Nope, did not work.

That will return a new exception

since it’s dev-time only you could do something like:

(alter-var-root #'clojure.core/get-in
                  (fn [v]
                    (fn [m ks]
                      (let [v (v m ks ::not-found)]
                        (if (= v ::not-found)
                          (throw (ex-info "" {}))
                          v)))))

Cool! I was hoping for something like that. I guess I could add it to my dev/env.clj :thumbsup:

providing not-found in get can be quite elegant: https://github.com/clojure/clojure/blob/clojure-1.10.1/src/clj/clojure/core.clj#L7203

It should break libraries that use get-in

I noticed take always returns a collection, while take-last might return nil. For instance when the input is empty;

(take 1 [])
=> ()
(take-last 1 [])
=> nil

from a seq point of view, these are the same

Good point.

Sure thing, i ran into this because I changed some code and now the api returned nil instead of an empty collection. Unfortunately not all our clients use languages with good nilpunning 😅

in general, you should guard against this by using (seq ...) in the if condition (or seq'ing the result) if passing it on

What I've seen some people do is to just define sget and sget-in, which do the above. The "s" meaing "safe".

Might not be the most elegant solution, but you can use a library like timbre, for example:

(let [ret (get-in {:a {:b 42}} [:a :z] :not-found)]
 (timbre/spy ret)
 ret)
2021-06-25T15:38:24.354Z MBP.local DEBUG [ss.experimental.scratch:2] - ret => :not-found
=> :not-found

(timbre/set-level! :trace)

I have this super-weird behavior in my multithreaded code: When my error message is the following:

error-message (count missing-files)

I've tried reproducing what happens here:

(require '[babashka.process :as p :refer [process]])
(require '[<http://clojure.java.io|clojure.java.io> :as io])
(def bb-process (process ["bash" "-c" "echo hi!"] {:shutdown p/destroy}))
(def java-proc (:proc bb-process))
(defn post-process []
  (throw (Exception. "Hi")))

(.thenApply
           (.onExit java-proc)
           (reify java.util.function.Function
             (apply [this p]
                 (post-process))))

(defn read-process [proc] ;; babashka/process
  (try
    (with-open [rdr (io/reader (:out proc))]
      (binding [*in* rdr]
        (loop []
          (let [line (read-line)]
            (when (not (nil? line))
              (println line)
              (recur))))))
    (catch Exception e
      (println "ooops\n" (pr-str e)))))
(read-process bb-process)

(type missing-files) ;; clojure.lang.APersistentMap$ValSeqs and (count missing-files) ;; 1 work, but when I try to touch the contents of missing-files for example with (type (first missing-files)) my function goes mum again and nothing is printed.

What do you mean by "use of `tap>` is not the problem either"? Have you tried using it with e.g. Reveal?

to be clear, tap> does not display any stack trace if it fails, it just silently fails

you need try/catch inside your tapping function if you need to know it failed

But if I remove the tap and just add a print it also fails.

I should try to simplify my code. I just wanted to quickly throw together something that worked but I guess I need to simplify it to find the error.

Thanks for the tip about reveal! I'll look into it

it sounds like one of the values in your collection is throwing an exception in its toString method

ins)user=> (defn try-string [o] (try (.toString o) (catch Exception e (pr-str e))))
#'user/try-string
(ins)user=> (try-string (reify Object (toString [_] (/ 1 0))))
"#error {\n :cause \"Divide by zero\"\n :via\n [{:type java.lang.ArithmeticException\n   :message \"Divide by zero\"\n   :at [clojure.lang.Numbers divide \"Numbers.java\" 188]}]\n :trace\n [[clojure.lang.Numbers divide \"Numbers.java\" 188]\n ..."

works like regular str on values that don't throw

(ins)user=> (try-string 1)
"1"

how can i parse a string to timestamp that has AM/PM in the string? I tried clj-time, but it seems to ignore the aa param in the formatter, Read instant date doesn't like the syntax either (def custom-formatter (f/formatter "yyyy-MM-dd HH:mm aa")) (str (f/parse custom-formatter "2021-06-28 03:00 pm")) => "2021-06-28T03:00:00.000Z" (str (f/parse custom-formatter "2021-06-28 03:00 am")) => "2021-06-28T03:00:00.000Z"

it's not hard to do it with raw interop

(cmd)user=> (java.text.SimpleDateFormat. "hh:mm a" java.util.Locale/US)
#object[java.text.SimpleDateFormat 0xc4c0b41 "java.text.SimpleDateFormat@3264901b"]
(cmd)user=> (.parse *1 "11:30 AM")
#inst "1970-01-01T19:30:00.000-00:00"

I added the explicit locale because "AM" and "PM" are not universal to all locales

oh, your issue is that the resulting string doesn't have am/pm - you need to use the date format a second time to get that

I would prefer to use java.time in 2021

(cmd)user=> (java.text.SimpleDateFormat. "hh:mm a" java.util.Locale/US)
#object[java.text.SimpleDateFormat 0xc4c0b41 "java.text.SimpleDateFormat@3264901b"]
(cmd)user=> (.parse *1 "11:30 AM")
#inst "1970-01-01T19:30:00.000-00:00"
(ins)user=> (str (.format *2 *1))
"11:30 AM"

that's fair

anyway, it's doable with dateformat, you just need to use the format method to get the format back out in a string is all

or, more helpful probably, you need to use the formatter function of whatever library you are using, which I assume must be there

@jdkealy clj-time calls the function you want unparse, but I agree with @borkdude that you should use something based on java.time, clj-time is deprecated

(`clj-time` is pretty clearly deprecated and has a big, bold notice at the top of the readme recommending folks use something based on Java Time instead!)

I'm glad I went with java.time in babashka

I'm glad we have stopped using clj-time altogether at work!

we're still using it at work because I'm extremely lazy when it comes to rewriting something which works quite well for a long time already

Thanks y'all 🙂

Straight from the source 🙂 https://github.com/clj-time/clj-time/commit/26e0f3ad066612336fce3050a3ea845072cf26aa

@seancorfield oh hah, I didn't realize you used to maintain that lib ;)

Yup, for years 🙂 because we used it at work. But recently I put in a "tech debt" Jira ticket to get us off it, mostly to raw Java Time, but some clojure.java-time wrapper code still exists. And we still have some date-clj code in play as well (I have another Jira ticket to get rid of that).

We have clojure.java-time as well, but lately we’ve been using plain interop more and more. I don’t know, plain interop seems a bit more transparent to me

there's also cljc.java-time (bb compatible): https://github.com/babashka/babashka/blob/master/doc/projects.md#cljcjava-time

I really wish assert returned its argument, since (assert (get ...)) feels really natural. Of course you can write your own

Not sure if I'm asking in the right channel; I want to know how do people upgrade the clj tool on Linux systems. Just download the new installer version and run?

Yep.