not looking for re-matches or anything like that, my question is more what's wrong with my actual regex.

(I do use a str/replace in my actual code with a matching group to extract that part out)

You're missing a +

it should be \[\w+\]

there's still a weird thing going on that I'm not following...

I'm doing https://exercism.org/tracks/clojure/exercises/log-levels/edit and i'd think that code would just work fine: (defn log-level "Takes a string representing a log line and returns its level in lower-case." [s] (str/lower-case (str/replace s #"\[(\w+)\]: .*" "$1")) ) but I'm getting "error\r\n" as a result.

I know I can use trim-newline, but I don't understand why I'd have to do it 😞

Regex witt only a dot normally does not match new line or carriage return characters

What are you hoping to be returned from this function?

"error"

Are you sure your input doesn't have newlines in it? If you're slurping a file it might have a trailing newline

it does have a newline in it. that's part of the test

that's the input string to this function

[ERROR]: \t Corrupt disk\t \t \r\n"

You are telling it to replace everything except any new lines after the right square bracket with $1, so the carriage return remains in the replaced string

Then don't use replace, instead extract the level with match and use that

Ah that makes sense, thanks Andy!

Agreed. If you only want whatever matched the \w+ part, then replace is an odd choice

thanks valerauko, I'll do that next then. I was just eh replace all the string with the thing I matched... Just couldn't understand why it was adding the \r\n.

Yeah, it was not adding those characters. You were, without realizing it, telling it to leave them there

Thanks 🙂 still learning.

Would this be an idiomatic way to write it? (str/lower-case (get (re-find #"^\[(\w+)\]:" s) 1))

Looks good to me

I recommend using an actual profiling tool, but if you want something with absolutely zero dependencies, then simply use this macro:

(import com.sun.management.ThreadMXBean)
(import java.lang.management.ManagementFactory)

(defn get-usage []
  (let [thread-bean (ManagementFactory/getThreadMXBean)]
    (System/gc)
    (.getCurrentThreadAllocatedBytes ^ThreadMXBean thread-bean)))

(defmacro with-memory-statistics [& body]
  `(let [start#     (get-usage)
         result#    (do ~@body)
         end#       (get-usage)]
     (array-map :result result# :diff (- end# start#) :start start# :end end#)))

following your example, you would then evaluate (with-memory-statistics (fib 300))

Thanks a ton, @U0479UCF48H

Doesn't the end# call gc before measuring? Is that intended?

My intention is indeed to do something like

(System/gc)
(measure)
(run the body)
(System/gc)
(measure)

Wouldn't that measure memory leak instead of overall usage?

ah yeah you’re right… I just took a look at the docs for getCurrentThreadAllocatedBytes, and this returns total # of bytes allocated in the thread (as in, a running total). So running GC is unnecessary at best and, at worst, only measures memory leaks.

so just deleting the (System/gc) call should be good enough. the start value is going to be strictly increasing, even if the java heap shrinks, since it measures how many bytes the JVM tried to allocate in total for that specific thread

Perhaps worth considering: (System/gc) is advisory and is no guarantee that any GC will actually happen at that point.

@U04V70XH6 That explains why I got meaningful results for both variations of the script. Thanks, good to know!

https://github.com/taoensso/carmine translates clojure data structures iirc

https://github.com/tolitius/redclaw I think this may be similar to what you describe: creating redis-backed data structures that implement the expected interfaces.

What does "really fast" mean? What does a "network service" mean? Are these really the only metrics?

TCP, low latency per request

With such a tacit description, I doubt there's anything better than java.net.ServerSocket. :)

aleph supports raw tcp/udp services, and it's rather performant, but I suppose you know about it 🙂

You can also interop with Aeron's Java implementation, I think Onyx did that "back in the days".

Java catching up to where Clojure started :)

https://openjdk.org/jeps/457 is exciting

I look forward to using it in 15 years

https://openjdk.org/jeps/461 feels a bit like transducers for streams...?

it is, and I absolutely think you could write something that took a transducer and implemented a gatherer, which is a very interesting idea to bring our whole transducers library into java streams

https://openjdk.org/jeps/456 is _ as an "unnamed" variable 🙂 Feels like they've been getting quite a bit of Clojure inspiration here...

It's a rising tides lift all boats kinda thing

I think most changes coming and that have come to Java strengthen Clojure - less so for Kotlin and Scala which kinda do seem sillier

well several of us in the Clojure core are friendly with Brian Goetz so we try to influence him :)

Is there one thing 'missing' in the JVM/JDK with no known JEP for it?

personal opinion: an equivalent to {:deps {lib/name {:mvn/version "..."}}}

I wish they’d just leave it alone mostly :)

the JDK?

Java in general

I agree. The java stdlib is pretty huge as-is and I find it very useful. Outside of tweaks to the JVM and GC, I don’t see much things that interest me much

20 years later: java.llm.adapters.ChatGptToGeminiAdapter.

https://openjdk.org/jeps/454 seems like a big deal

now u can program zig-style in java too 😉