@cureadvocate there may be others, but I think https://github.com/clojure/math.numeric-tower has functions that deal nicely with the various numeric types

ty, jaawerth. I didn't think that worked for BigInteger, but it does. Yay!

any one get graal native-image working with clojure spec?

Is it possible to find out if a value is serializable? Like is there an interface/protocol serializable values implement?

What do you mean by serializable?

Java has a Serializable interface but not sure if that’s actually what you want

@alexmiller encodable to transit-json

thanks for responding!

@jesse.wertheim There's some hidden gems under the clojure/ umbrella, that I think would benefit from more publicity. Like data.avl, data.priority-map, core.memoize, algo.generic, math.combinatronics, core.match, etc.

How do you get the reference to an object? Say I want to know the reference to 10 ?

As opposed to its value

Hum, I guess maybe its not possible in Java to do that

This is as close as it gets (System/identityHashCode [1 2])

@didbus I believe I once found a low level method to get the current address/pointer to an object in Java, but realize that such a value can change on every GC, and the times of GCs are unpredictable.

It looks like I experimented a few years ago briefly with a library called jol for "Java Object Layout" here: https://openjdk.java.net/projects/code-tools/jol/ It has a method VMSupport/addressOf that can return the current address of an object, after doing (:import (org.openjdk.jol.util VMSupport)). Note: I do not all the possible caveats of this method or what all might go wrong with it, but there is at least the issue that the value can change on every GC.

is there a way I can ask clojure to reload data_readers.clj from the repl?

I don’t know if there is a straight way to do it, but this should work:

(set! *data-readers* (into *data-readers* (map (fn [[tag f]] [tag (resolve f)])) (read-string (slurp ( "data_readers.clj")))))

Trying this example from https://clojuredocs.org/clojure.core/re-groups

(def phone-number "")
(def matcher (re-matcher #"((\d+)-(\d+))" phone-number))
(re-groups matcher)

Execution error (IllegalStateException) at java.util.regex.Matcher/group (Matcher.java:536).

what’s wrong here?

seems like you have to call (.find matcher) first

I get same error with Clojure 1.9.0 and JDK 1.8.0_192. Not clear to me yet whether that ever worked, or is changed behavior.

I tested all the way back to 1.1.0. maybe they changed something in Java?

user=> (doc re-matcher)
-------------------------
clojure.core/re-matcher
([re s])
  Returns an instance of java.util.regex.Matcher, for use, e.g. in
  re-find.

Doubtful. More likely that the example never worked

^^ designed for use with re-find

alright, I’ll edit the example 🙂

user=> (re-find matcher)
["456-3212" "456-3212" "456" "3212"]

@borkdude Interested in correcting the example on http://ClojureDocs.org?

then after that, you can:

user=> (re-groups matcher)
["456-3212" "456-3212" "456" "3212"]

oh wait, it calls re-find on it first. I figured yes

but it requires you to have previously called re-find, yes

thanks all

Java Matcher objects are mutable.

so call sequence is important.

yeah

this stuff doesn’t exist in clojurescript (re-groups, re-matcher)

ClojureScript regex's probably have some way to get multiple "capture" substrings from a single regex match, or at least most regex libraries nowadays do, but I haven't looked. Just surprised if it doesn't enable that functionality.

That requires a global regex in JS and that’s generally avoided

you may have heard that mutable objects are bad :)

make a table of tradeoffs 🙂

yeah. I’m writing a spec for re-groups now, that’s why I bumped into this. never actually used it

@andy.fingerhut interestingly via writing a spec for subs, I got into this: https://dev.clojure.org/jira/browse/CLJS-2979

no trade off there, only fix + speedups 😉

Is that ClojureScript-specific behavior, perhaps?

yes

Just checked that Clojure/JVM throws exception on (subs "" 1) as I would have expected, although the exception message contains -1 instead of 1 which is a bit weird, but at least the exception is there.

Simultaneous fix + speedup for the win!

Does it make much sense to write a spec that checks return values of functions that take and modify mutable objects? I would guess that would be fraught with trouble.

I do it for completeness sake. you can unstrument it if you want.

good or bad: function/var names with : ? ex: find-thing:source-a find-thing:source-b (no I don't want/need to use a multimethod, it's just a naming thing here)

Bad

ok, that's clear, the reader doc states something different I think

I didn’t say it was invalid

> Symbols beginning or ending with ':' are reserved by Clojure. A symbol can contain one or more non-repeating ':'s.

I just don’t like it

yeah I saw it recently during a review and I was wondering

You asked good/bad :)

Not valid/invalid

understood

isn’t that how they namespace things in common lisp?

cider highlighting doesn't like it much

yay, I can now see the return “type”:

user=> (doc re-groups)
-------------------------
clojure.core/re-groups
([m])
  Returns the groups from the most recent match/find. If there are no
  nested groups, returns a string of the entire match. If there are
  nested groups, returns a vector of the groups, the first element
  being the entire match.
Spec
  args: (cat :matcher :speculative.specs/matcher)
  ret: (or :string :speculative.specs/string :seqable :speculative.specs/seqable-of-string)
nil

hm lein uberjar seems to completely ignore that some dependencies are marked as scope "compile" or "provided"

Dispatching on specs: (defpoly jump (partial s/valid? dog-spec) [animal-obj] ...) 😉

what’s defpoly?

@borkdude I think that's @john’s predicate dispatch stuff?

Yeah, I'm not beholden to the defpoly name

I'm curious to learn more about deploying a single-process Clojure app with zero downtime. Not sure if that qualifies as blue-green deployment but something in that direction. Are there any tools to make this easy? How did you solve this?

if you're insane you could use tools.namespace/refresh and hot load the new code in the prod process

I think the real answer is usually some variant of load-balancing and service discovery

which would not be different between clojure and another language as I understand it?

Single process? Like you mean given one and only one process?

Normally you have a fleet of hosts, and you rotate them in and out as new changes are deployed

I was thinking that perhaps there's something particularly easy to setup. Currently I'm just running a systemd process and nginx reverse proxying to a hardcoded port.... so service discovery and swapping processes sounds like it'll require a lot more components...

right - to limit downtime you need multiple processes (at least for a minute or two at a time)

and then load balancing or service discovery to make the swap seamless

> Normally you have a fleet of hosts, and you rotate them in and out as new changes are deployed Yeah, that particular app doesn't follow an "immutable infrastructure" style approach where you just cycle entire nodes (the node is stateful)

Right multiple processes for the transition time would be expected.

I think this is part of why people end up using things like redis and kafka - to pull the stateful stuff out of the apps

at @dayjob we use kafka to accumulate runtime state, then marathon to manage app instances, and the kafka topics are a mechanism of service discovery of sorts

A statefull node? Like in memory state inside the process? Or disk persistence?

disk persistence

Yeah, definitely agree with the general motivation of having state in separate services but the project is OSS and I want to keep setup as easy as possible in the interest of contributors (using SQLite at this time)

Ok, well, you can probably pull it off if you can dynamically change the nginx port. You could start a new process on a different port and then do the port swap, and then kill the old process.

what about abstracting the sql layer, so people can use eg. postgres for seamless roll over, or sqlite for simplicity?

you might also look at something like envoy, likely overkill but I suspect doing the swap in envoy would be easier than in nginx

> NGINX configuration is captured in configuration files. When you change a configuration file and restart NGINX to pick up the new configuration, it implements a “graceful restart”. Both the old and new copies of NGINX run side by side for a short period of time. The old processes don’t accept any new connections and terminate once all their existing connections terminate.

oh - so you could roll over nginx with an alternate port config

Seems like that could work

I'm also looking at Caddy which seems like a nice reverse proxy with automatic HTTPS: https://caddyserver.com/docs/proxy It supports an upstream option that can be a port range... > try_duration is how long to try selecting available upstream hosts for each request. By default, this retry is disabled ("0s"). Clients may hang for this long while the proxy tries to find an available upstream host. This value is only used if a request to the initially-selected upstream host fails.

This could also be interesting https://stackoverflow.com/a/32001397/172272

Anyways, I think you got the gist

I did, thanks for your suggestions 🙂