Clojurians Log v2

hola @volrath

hallo @pesterhazy !

@richiardiandrea hmm find-ns happens to trigger an interesting shadowing bug, opened https://dev.clojure.org/jira/browse/CLJS-2088

Identifying a repl:

(cond
  (find-ns '<http://clojure.java.io|clojure.java.io>) :clj
  (find-ns '<http://clojure.clr.io|clojure.clr.io>) :cljr
  (find-ns 'lumo.repl$macros) :lumo
  (find-ns 'planck.repl$macros) :planck
  (find-ns 'cljs.core$macros) :cljs-js
  (find-ns 'cljs.core) :cljs-jvm
  :else :unknown)

@cgrand find-ns is bootstrap only

for repl sniffing it works enough in non-bootstrap

says so in the docs at least

don’t use that in a lib though (advanced compilation...)

hmm you are right, actually works in normal as well

but sending that before every interaction would suck

why not once per connection?

rlwrap nc localhost 8201
shadow-cljs - REPL - see (help), :repl/quit to exit
[13:0]~shadow.user=> (shadow/node-repl)
[13:1]~cljs.user=> :repl/quit
:repl/quit
[13:0]~shadow.user=>

@pesterhazy starts out as CLJ, then becomes CLJS, then CLJ again

if only there was a protocol which clearly tells when an upgrade occurs 😇

I know right? 🙂

any ideas how to do the upgrade process properly for cljs yet?

not self-hosted that is

still not totally on-board with unrepl yet but thats mostly due to tool support

guess I can’t get around nrepl for now

Some ideas start to click together. The current cljs repl infrastructure is not ok but it can be remixed (that is: reuse environments and so on but recreate an upgradable infrastructure to replace https://github.com/clojure/clojurescript/blob/master/src/main/clojure/cljs/repl.cljc)

shadow-cljs doesn’t use any of that, maybe its simpler to integrate there

is there an unrepl client I could test this with?

I dunno, ask @pesterhazy or @richiardiandrea

my current idea is to consider the loop on the js-side as its own repl

it starts as a framed JS repl but is upgradable

this is the loop for the CLJS REPL at the moment

https://github.com/thheller/shadow-cljs/blob/master/src/main/shadow/cljs/devtools/server/repl_impl.clj#L91-L130

it reads off *in* and prints to *out*

the actual JS eval is RPC

@thheller I'd test with netcat first

so it should be really simple to add the upgrade

once you have a working protocol, you can try unravel

this is where the blob is concatenated at the moment: https://github.com/pesterhazy/unravel/blob/b687edb188aaab58b9ac0ce7a143172dc3e256a6/src/unravel/loop.cljs#L115

it's not updated yet to use the "official" unrepl blob

and the blob is only for clj

bootstrap soon

of course the blob would have to be completely replaced

yeah still not sold on the blob idea

I’m fine with just having a dependency in the server

@thheller how do you handle the print part in shadow repl?

the client decides how to print

currently it just does pr-str but I could make that configurable

it does not send a wrapped form over the wire

so if you eval (js/alert "foo") the client actually just evals alert("foo") nothing else

not the wrapped stuff the normal CLJS REPL sends

@cgrand, I did a presentation on unrepl/unravel here in Berlin yesterday. Here are the slides: http://pesterhazy.karmafish.net/unravel-your-repl.pdf

How was it received?

pretty well. I had a few people express interest in contributing after the talk, which is awesome

^ yes, pretty well 🙂

great! welcome @volrath

> Rich capabilities, frugal dependencies I like this wording!

bonus points for using “transclusion”

heheh

[15:1]~cljs.user=> [1 2 3]
(] 3   2   1 [)
[15:1]~cljs.user=> (set! shadow.cljs.devtools.client.env/repl-print-fn #(reverse (pr-str %))
)
(] " }
 ; ) ) _ P R A H S _ 5 2 6 5 4 _ _ 1 p , l l u n ( l l a c . r t s _ r p . e r o c . s j l c , l l u n ( l l a c . e s r e v e r . e r o c . s j l c   n r u t e r
 { ) _ P R A H S _ 5 2 6 5 4 _ _ 1 p (   n o i t c n u f "   n o i t c n u F [ t c e j b o #)
[15:1]~cljs.user=> [1 2 3]
(] 3   2   1 [)

not the result I had in mind but fun 😛

@cgrand, yup, I've learned a new word 🙂

[15:1]~cljs.user=> (set! shadow.cljs.devtools.client.env/repl-print-fn #(clojure.string/upper-case (pr-str %)))
#OBJECT[FUNCTION "FUNCTION (P1__45626_SHARP_){
RETURN CLOJURE.STRING.UPPER_CASE.CALL(NULL,CLJS.CORE.PR_STR.CALL(NULL,P1__45626_SHARP_));
}"]
[15:1]~cljs.user=> [:hello :world]
[:HELLO :WORLD]

cgrand: thanks! I really like the whole idea. I told @pesterhazy yesterday that I went through the documentation two nights ago, to prepare for the talk, and I think I can contribute a very small PR with some minor fixes to the README (commas, and a couple typos) and then maybe I can help with documenting more stuff (missing tags like :started-eval, and so on)

that would help with my understanding of the protocol as well

if that's ok with you

that’s more than totally ok

is there a CLJS impl for the unrepl print style yet?

Don’t hesitate pinging me (in private or on the channel)

thank you! will do

yes there is, @plexus made one https://github.com/plexus/unrepl/blob/lumo_support/src/unrepl/print.cljc

I plan to merge it when I got time

hmm

[15:1]~cljs.user=> (load-file "/Users/zilence/code/shadow-cljs/src/dev/unrepl/print.cljc")
[:result {:type :repl/error, :message "cond-splice not in list", :data {:type :reader-exception, :line 15, :column 5, :file "unrepl/print.cljc"}, :causes []}]

indeed, remove the @

and add a do

another one at 170

ok got it

reader discrpencies: LispReader is ok with that, tools.reader not

async-reader should be fine too

hmm heeh

tools.reader issue?

it cannot print itself

[18:0]~shadow.user=> (shadow/dev :browser)
[18:1]~cljs.user=> (require 'unrepl.print)
[18:1]~cljs.user=> (set! shadow.cljs.devtools.client.env/repl-print-fn unrepl.print/edn-str)
nil
[18:1]~cljs.user=> {:foo "bar"}
{:foo "bar"}

TypeError: Cannot read property 'cljs$core$IFn$_invoke$arity$1' of undefined
    at cljs.core.MultiFn.cljs$core$IFn$_invoke$arity$1 (cljs.core.cljs:10295)
    at Function.unrepl.print.ednize.cljs$core$IFn$_invoke$arity$3 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:457:119)
    at Function.unrepl.print.ednize.cljs$core$IFn$_invoke$arity$2 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:425:28)
    at Object.unrepl$print$print_on [as print_on] (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:610:33)

doesn’t seem to work really

I haven’t had time to test it, I was doing some preliminary work on the upgradable cljs repl

(range 0 10000) also fails

TypeError: Cannot read property 'cljs$core$IFn$_invoke$arity$1' of undefined
    at cljs.core.MultiFn.cljs$core$IFn$_invoke$arity$1 (cljs.core.cljs:10295)
    at Function.unrepl.print.ednize.cljs$core$IFn$_invoke$arity$3 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:457:119)
    at Function.unrepl.print.ednize.cljs$core$IFn$_invoke$arity$2 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:425:28)
    at unrepl$print$print_on (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:610:33)
    at Function.unrepl.print.print_vs.cljs$core$IFn$_invoke$arity$5 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:571:113)
    at Function.unrepl.print.print_vs.cljs$core$IFn$_invoke$arity$3 (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:516:30)
    at Object.unrepl$print$print_on [as print_on] (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:653:23)
    at unrepl$print$edn_str (eval at shadow$cljs$devtools$client$browser$do_js_load (shadow.cljs.devtools.client.browser.cljs:48), <anonymous>:698:18)

damn js stack traces are ugly

ah wait I have a debugger for that 😉

i may have an idea, in the clojure version you need to set some dynvars around it

(keep calm it’s synchronous so plain binding is ok)

it would explain the undefined

*ednize* seems unbound

(defmulti default-ednize-fn class) <- no class in cljs

why isn’t that a protocol?

trying to remember — I dislike multimethods so there must be a reason

Error: No method in multimethod 'unrepl.print/default-ednize-fn' for dispatch value: function Function() { [native code] }

just changed it to type

uhm (try try try try try try try try try try #unrepl/... nil) 😛

almost?

[20:1]~cljs.user=> (range 0 1000)
(try try try try try try try try try try #unrepl/... nil)

🙂

where does the try come from? 😛

re: protocol, really I don’t see why anymore, maybe to have the opportunity to prefer impls in case of conflicts?

there are only two “try” one in the expansion of a macro, I think it’s the one

ok .. I gotta do some other stuff .. will try again later

on line 234 blame-seq which is a macro is called like a function (because cljs) and then its head (`try`) is used for v

[20:1]~cljs.user=> (range 0 1000)
(0 1 2 3 4 5 6 7 8 9 #unrepl/... nil

indeed

#?(:cljs (:require-macros [unrepl.print :refer (blame-seq)])) in ns

#?(:clj
   (defmacro ^:private blame-seq [& body]
     `(try (seq ~@body)
           (catch Throwable t#
             (list (tagged-literal 'unrepl/lazy-error t#))))))

oh noes, you just broke it for self-hosted

hihi

macrovich to the rescue!

heh?

(btw David applied my patch)

“was wondering why $macros should it find either way” ?

cgrand: lol let me rephrase

Why did you switch to detecting core$macros instead of core

because in self-hosted cljs, (find-ns 'lumo.repl) cripples access to this ns macros (open lumo, enter this find-ns and after that try to use doc)

Oh gotcha temporary fix before your patch is officially in

and until non-patched cljs are rare enough

@thheller I think @plexus was writing an Emacs mode