Clojurians Log v2

Wow, I really shuffled things around, thx for the heads up. Instructions for running rxTrak are here: https://github.com/kennytilton/web-mx-sampler/tree/main/src/tiltontec/example The source for rxTrak is here, but no README: https://github.com/kennytilton/web-mx-sampler/tree/main/src/tiltontec/example/rxtrak So it is weird that the Matrix wiki is documenting a Web/MX project, but MX makes more sense as a UI state solution, so I will stick with that. But I should add a readme to the rxTrak source directory since I am linking to that. Thx!

Have you tried [com.tiltontec/matrix "5.0.1"]? There was definitely stuff like that getting cleaned up as I moved to non-snapshot releases. Hmmm. Maybe I should get the hood up on those snapshots, since there is no way to delete them from Clojars.

Could not find artifact com.tiltontec:matrix:jar:5.0.1

Running with figwheel works fine:

clojure -M -m figwheel.main --build todomvc --repl

I bascally copied all the files in the todomvc dir fix ns added deps:

[funcool/bide "1.7.0"]
                [com.taoensso/tufte "2.4.5"]
                [com.tiltontec/matrix "5.0.1-SNAPSHOT"]
                [com.tiltontec/web-mx "2.0.1-SNAPSHOT"]

Oy, I have been juggling TodoMVC's all over the map. Did you find the one under web-mx-sampler, or the old mxtodomvc repo? I'll see if I can find which from the deps. web-mx stable is 2.0.1, btw.

Ah, to make things worse, in web-mx-sampler I have both project.clj and deps.edn, and the latter is out of sync. This could be the problem. My apologies for that. I will fix that now.

https://github.com/kennytilton/web-mx-sampler

hmm project.clj use 5.0.1 which maven insists isn't there

Hmmm. Maybe I installed locally and never on clojars. On Clojars it is def not there. tbh, I am just now adjusting to semver and stable vs snapshot and all that. the sampler is using

[com.tiltontec/matrix "5.0.1-SNAPSHOT"]
                 [com.tiltontec/web-mx "2.0.1-SNAPSHOT"]
                 [com.tiltontec/mxxhr "2.0.0-SNAPSHOT"]

I started with those vers, and they had the mentioned error

But running with figwheel works?

yep which is super weird

OK, coffee is sinking in. I thought I recognized un-stopped as sth I had tossed -- I could, in fact -- but it is still part of MX. And it is a macro, and we are in CLJC-land, which is a dicey combo. Do you have a little shadow set-up you can share that fails on that?

I am def having deja vu on this shadow vs figwheel thing.

just a sec

Of course we also have the issue of CLJ moving away from refer-macros, which I am glad to see what that is weird, too, I will have to remind myself on that.

Just cleaned it up for you from all my noise

I started refactoring the requires and planed to clean that up too

What is the shadow command you use to launch?

yarn watch matrix or npm run watch matrix

OK, I have recreated the error.

[:matrix] Build failure:
------ ERROR -------------------------------------------------------------------
 File: jar:file:/Users/kennethtilton/.m2/repository/com/tiltontec/matrix/5.0.1-SNAPSHOT/matrix-5.0.1-SNAPSHOT.jar!/tiltontec/cell/integrity.cljc:1:1
--------------------------------------------------------------------------------

   1 | (ns tiltontec.cell.integrity
-------^------------------------------------------------------------------------
Invalid :refer, var tiltontec.cell.base/un-stopped does not exist
--------------------------------------------------------------------------------
   2 |   (:require
   3 |     ;#?(:clj [taoensso.tufte :as tufte :refer :all]
   4 |     ;  :cljs [taoensso.tufte :as tufte :refer-macros [defnp p profiled profile]])
   5 |     #?(:cljs [tiltontec.util.base
--------------------------------------------------------------------------------

you mean with the refactoring?

Yes. Or do you see anything wrong with my requires in re macros and CLJC and all that?

This is the kind of thing I get fixed and then forget, in part because I cannot forn a mental model around it. Like when we were supposed to have X.clj and X.cljs and require one from the other.

My working premise is that figwheel.main is being nice and bailing out some unsupported thing I am doing that shadow is correctly rejecting.

Hmmm. Back in March someone persuaded me to get current with the requires in re macros. Maybe I did it wrong but figwheel handles it. And by then I was not doing shadow. Still digging...

All I was doing was cleaning unused requires and was planning on moving macros to normal refers. I don't think this is directly related to the issue at hand

Your hypothesis is much more likely. Maybe figwheel handles some clj-cljs relation some other way than shadow. Maybe @thheller can help us out 😀

I think I found the syntax I need:

(ns tiltontec.matrix.api
  #?(:cljs
     (:require-macros [tiltontec.matrix.api]))
  ....)

I didn't follow the whole thread but the main difference when it comes to macro is that shadow "warns" you about possibly incorrect macro setups. figwheel is not as strict. The "problem" is that macros need to be explicitely required. so, as above the :require-macros for itself. That ensures the compiler will load this namespace and provide the macros properly.

if this is not done you are relying on something else having loaded the macro namespace, which may happen by coincidence, but is not guaranteed. so if nothing required the macros, you get weird runtime errors.

so there might never be a problem, but making sure macros are properly provided just makes it safer

Ah, thx for jumping in! I was going to thrash away for a few hours before sending up a flare. Someone had suggested we could ditch the explicit require, sparing users of my library knowing which symbols were macros. I'll revert all that and go wwith explicit. Thx again, @thheller.

hmm? maybe you misunderstood what I meant

YOU as the library author do (ns library.ns (:require-macros [library.ns]))

the USERS of the library don't care

they never :require-macros anything, or do anything else macro specific

the self :require-macros in your library enable the compiler to figure out the macro stuff, it is the hint that tells the compiler "this namespace has macros". exactly so that the users don't have to specifcy whether they meant the macro or the function

for example in the snippet above you have [taoensso.tufte :as tufte :refer-macros [defnp p profiled profile]]

If taoensso.tufte sets up macros correctly (which I don't know) then this could just be :refer instead of :refer-macros, thus making it work exactly like CLJ. making the reader conditional redundant even

basically if there is a :require-macros that is not the same symbol as the ns itself thats a "bug"

https://code.thheller.com/blog/shadow-cljs/2019/10/12/clojurescript-macros.html

Thx for the extra trouble! That ^^^ is what I got talked out of doing! 🤦 The change seemed to work so I went with it. I will revert to that as soon as I finish up this smaller brush fire. 🙂

OK, @lidorcg, I am grinding thru the reversions. I'll let you know when your project runs under shadow.

So @lidorcg, I tried building against the versions I had in early March where the set-up was a little friendlier to shadow, but then the rest of the code you grabbed for the exercise is a problem because that assumes the overhaul I did getting things ready for a wider audience. The only way forward is to get the current MX compatible with Shadow. A good idea anyway, but I was hoping for a quick workaround. I will head over to #shadow-cljs to sort this out, since it might benefit others to have the excahnge there.

@hiskennyness Thank you so much for your efforts 🙏🏼! And @thheller thank you for your guidance, we learned a lot

In the meantime we'll play with the working figwheel setup, we still have some hacking to do and see if what we're aiming for is achievable through matrix. If it is, we'll focus on making the shadow setup work 😄

Awesome. I am hard at it now, will keep you posted.

@lidorcg, right now I am stumped on getting shadow to build MX, pending info from the Clojure standard itself, which I cannot find. 🙂 Maybe tmrw we can get that from the cognoscenti--I asked late in the day. I did try the recommended split of macros into their own CLJ, and got past the un-stopped symbol, but on the next build failure that did not help--with a different error. It occurs to me I have prolly only ever built with shadow on RN projects, so this may not have ever been known to work. I like your plan of soldiering on with figwheel until you know if you even get to POC, then worrying about shadow. Me, I will keep an eye on the issue, and go drown my sorrows in ClojureDart. 🙂

Hi @hiskennyness I tested the shadow setup and it compiled 🥳 However I noticed a little bug: 1. When refreshing the page 2. Trying to add task 3. result in this error:

Uncaught TypeError: pdom is null
    <anonymous> html.cljs:183
    cljs$core$IFn$_invoke$arity$5 core.cljs:11368
    cljs$core$IFn$_invoke$arity$3 watch.cljc:16
    tiltontec$cell$evaluate$propagate evaluate.cljc:526
    tiltontec$cell$evaluate$c_value_assume evaluate.cljc:384
    tiltontec$cell$evaluate$calculate_and_set evaluate.cljc:223
    propagate_to_callers evaluate.cljc:580
    cljs$core$IFn$_invoke$arity$2 integrity.cljc:71
    cljs$core$IFn$_invoke$arity$1 integrity.cljc:66
    tiltontec$cell$integrity$finish_business integrity.cljc:80
    tiltontec$cell$integrity$call_with_integrity integrity.cljc:182
    tiltontec$cell$core$cset_BANG_ core.cljc:301
    tiltontec$cell$core$c_reset_BANG_ core.cljc:306
    tiltontec$model$core$mset_BANG_ core.cljc:89
    cljs$core$IFn$_invoke$arity$variadic core.cljc:113
    cljs$lang$applyTo core.cljc:112
    cljs$core$IFn$_invoke$arity$5 core.cljs:3997
    cljs$core$IFn$_invoke$arity$variadic api.cljc:163
    todo_entry_field core.cljs:34
    setProperties dom.js:549
    forEach object.js:25
    setProperties dom.js:531
    createDom_ dom.js:925
    createDom dom.js:900
    cljs$core$IFn$_invoke$arity$4 core.cljs:3947
    cljs$core$IFn$_invoke$arity$4 core.cljs:3989
    cljs$core$IFn$_invoke$arity$2 html.cljs:110
    dom html.cljs:113
    2 core.cljs:4770
    2 cljs.core.js:17787
    sval core.cljs:3462
    cljs$core$ISeqable$_seq$arity$1 core.cljs:3519
    cljs$core$seq core.cljs:1236
    2 core.cljs:3817
    sval core.cljs:3462
    cljs$core$ISeqable$_seq$arity$1 core.cljs:3519
    cljs$core$seq core.cljs:1236
    cljs$core$INext$_next$arity$1 core.cljs:3294
    cljs$core$next_STAR_ core.cljs:3925
    cljs$core$IFn$_invoke$arity$4 core.cljs:3948
    cljs$core$IFn$_invoke$arity$4 core.cljs:3989
    cljs$core$IFn$_invoke$arity$2 html.cljs:110
    dom html.cljs:113
    2 core.cljs:4770
    2 cljs.core.js:17787
    sval core.cljs:3462
    cljs$core$ISeqable$_seq$arity$1 core.cljs:3519
    cljs$core$seq core.cljs:1236
    2 core.cljs:3817
    sval core.cljs:3462
    cljs$core$ISeqable$_seq$arity$1 core.cljs:3519
    cljs$core$seq core.cljs:1236
    cljs$core$IFn$_invoke$arity$4 core.cljs:3989
    cljs$core$IFn$_invoke$arity$2 html.cljs:110
    cljs$core$IFn$_invoke$arity$1 api.cljs:37
    <anonymous> core.cljs:73
    globalEval app.js:430
    evalLoad app.js:1387
    <anonymous> app.js:1618
html.cljs:183:25

#(when (= (.-key %) "Enter")
                          (let [raw (form/getValue (.-target %))
                                title (str/trim raw)]
                            (when-not (str/blank? title)
                              (mswap! (mget @matrix :todos) :items-raw conj
                                      (make-todo {:title title})))
                            (form/setValue (.-target %) "")))

I think the issue now is how the shadow version is kicking off the app. I glossed over before the bit about matrix-build! being the init-fn:

:builds {:matrix     {:target     :browser
                       :output-dir "resources/public/js/compiled"
                       :asset-path "/js/compiled"
                       :modules {:app {:init-fn matrix-todo.core/matrix-build!}}
                       :devtools {:preloads []}
                       :dev {:compiler-options {:closure-defines {}}}}}}

(let [root (dom/getElement "app")
      app-matrix (matrix-build!)]
  (set! (.-innerHTML root) nil)
  (dom/appendChild root
    (tag-dom-create
      (mget app-matrix :mx-dom)))

  (when-let [router-starter (mget app-matrix :router-starter)]
    (router-starter)))

Ah, got this from ChatGPT, as a way to add a shadow build alongside a fig.main build:

{:builds
 {:app {:target :browser
        :output-dir "public/js"
        :asset-path "/js"
        :modules {:main {:entries [my.app.core]}}}}}

This is missing dependencies, there is a way to tell shadow to use dependencies from lein https://shadow-cljs.github.io/docs/UsersGuide.html#Leiningen So your shadow-cljs.edn should be:

{:lein true
 :builds
 {:app {:target :browser
        :output-dir "public/js"
        :asset-path "/js"
        :modules {:main {:entries [my.app.core]}}}}}

@mynomoto the full edn looks like this:

{:nrepl {:port 8777}

 :source-paths ["src" "test"]

 :dependencies [[funcool/bide "1.7.0"]
                [com.taoensso/tufte "2.4.5"]
                [com.tiltontec/matrix "5.0.1-SNAPSHOT"]
                [com.tiltontec/web-mx "2.0.2-SNAPSHOT"]
                [com.tiltontec/mxxhr "2.0.0-SNAPSHOT"]]

 :dev-http {3000 "resources/public"}

 :builds {:matrix     {:target     :browser
                       :output-dir "resources/public/js/compiled"
                       :asset-path "/js/compiled"
                       :modules {:app {:init-fn matrix-todo.core/matrix-build!}}
                       :devtools {:preloads []}
                       :dev {:compiler-options {:closure-defines {}}}}}}

Oh, sorry. I saw the message on the channel and failed to get the context.

Quick fix did not work. 😞 But I am pretty sure we need to do sth different kicking off a shadow version.

Progress: https://clojurians.slack.com/archives/C6N245JGG/p1683735522125549

@lidorcg, if you really need to use shadow, I suggest starting from https://github.com/shadow-cljs/quickstart-browser and moving the toplevel code into the start function, as I did over in the #shadow-cljs post. I might try it myself at some point, but not sure when I will get to it.

Thanks again @hiskennyness 🙏🏼 I'm sure we'll be able to make this work. I'll send here once I have a fully working shadow-mx-web starter

Hey, I went ahead and gave it a shot, works fine on one of the w/mx examples. I will send you a zip shortly. btw, not seeing hot reload work. Does shadow do that? Might need more work, but we have a start.

Hey, @hiskennyness The web-mx-sampler won’t run, i think it’s related to the matrix commit yesterday (mxxhr cleanup). Changing the matrix version 5.0.1-SNAPSHOT to 5.0.0-SNAPSHOT in the deps.edn file, fixed it:

Thx, @aviv846, I'll get on that now. 🙏

Ah, I neglected to push the sampler changes that accommodated the tweak to MX 5.0.1-SNAPSHOT. I just pushed the web-mx-sampler changes, and tested in a fresh clone after clearing out my .m2, it should be ok now. FYI, the relevant new requires:

[com.tiltontec/matrix "5.0.1-SNAPSHOT"]
[com.tiltontec/web-mx "2.0.2-SNAPSHOT"]
[com.tiltontec/mxxhr "2.0.0-SNAPSHOT"]

That is the src dir and shadow-cljs.edn. Had to tweak a few things once the shadow build succeeded, so check the source closely! I'm here if you need anything. Thx again, @thheller!

ps. I was careful to clear out my .m2 so you should be good pulling from clojars. OTOH, GitHub is having an outage so the repos themselves have not been updated.

yep, was having the same error, and I assumed the lib haven't updated yet

@hiskennyness thank you so much for the immediate attention and solution 🙏

np! And thx for surfacing the issues that got me up to speed on :require-macros!