goog.require('shadow.js.shim.module$fs');

["fs" :as fs])

=>> node target/main.js
SHADOW import error /Users/chen/repo/cumulo/cumulo-workflow/server/target/shadow-cljs/builds/app/dev/out/cljs-runtime/server.main.js

/Users/chen/repo/cumulo/cumulo-workflow/server/target/shadow-cljs/builds/app/dev/out/cljs-runtime/server.main.js:19
})());
  ^
ReferenceError: fs is not defined
    at /Users/chen/repo/cumulo/cumulo-workflow/server/target/shadow-cljs/builds/app/dev/out/cljs-runtime/server/main.cljs:14:10
    at Object.<anonymous> (/Users/chen/repo/cumulo/cumulo-workflow/server/target/shadow-cljs/builds/app/dev/out/cljs-runtime/server.main.js

looks strange. I thought it's compiled to require('fs')

check /Users/chen/repo/cumulo/cumulo-workflow/server/target/shadow-cljs/builds/app/dev/out/cljs-runtime/shadow.js.shim.module$fs.js

I might have broken that accidentally

it should contain shadow.js.shim.module$fs = require("fs");

+ goog.provide

server/main.cljs:14:10 whats there?

maybe you do (js/fs.something ...)?

should be (fs/something ...)

tested, not broken

https://github.com/Cumulo/cumulo-workflow/blob/master/server/src/server/main.cljs

yes, fs.existsSync is not valid .. use (fs/existsSync ...) (and the others)

does that not produce a warning?

I don’t think we have covered this before.. I need to debug a javascript library that is included in my build. Is there a way to include a js file/package that is not in node_modules, but have it otherwise be treated the same? (somehow I need to include my own fork of Keypress.js with a minimum of fuss)

goog.provide("shadow.js.shim.module$fs");
shadow.js.shim.module$fs = require("fs");

code is there..

... my code was using fs.x...

working now.. so it's fs/existsSync

shadow-eval?

@jiyinyiyong it’s just a sample project using the latest :bootstrap build ability of shadow-cljs

now shadow-cljs bootstraps?

yes 🙂 it’s great. just working out the final bugs/limitations but it is already quite capable

https://github.com/mhuebert/shadow-eval

not see docs XD

I just added a readme to that repo

& updated the readme again.. I am not entirely sure if it will ‘just work’ but you can have a try

if you can follow the above directions & then see the same thing as the screenshot I posted, then it is working the same as on my machine

sounds cool anyway

(defn eval-str [source cb]
  (cljs/eval-str
    c-state
    source
    "[test]"
    {:eval cljs/js-eval
     :load (partial boot/load c-state)
     :ns   (symbol "shadow-eval.user")}
    cb))

what does "[test]" mean in this example?

the cljs compiler just needs some string provided as a ‘name’ for the current thing being evaluated. in normal builds this would be the filename. in self-host, there is no “file” so you can specify this to be whatever you want. it will show up in stacktraces etc.

oh, a fake file name

@mhuebert :resolve {"keypress.js {:target :file :file "path/to/file/from/project-root.js"}}

will look into the other issue

@thheller oh great, that’s simple. Just had a friend test a fresh install (never had shadow-cljs installed before) using the new instructions in the shadow-eval readme. It all works exactly like on my machine 🙂 including the errors in the first two blocks.

if shadow-cljs is installed globally, and not run using yarn run shadow-cljs .., does it still use the version specified in package.json?

or does one always have to use yarn run ... to get the specified version

`------ ERROR -------------------------------------------------------------------
 File: /Users/zilence/code/tmp/shadow-eval/src/shadow_eval/user.cljs:1:1
--------------------------------------------------------------------------------

   1 | (ns shadow-eval.user
-------^------------------------------------------------------------------------
   2 |   (:require re-view-hiccup.core
   3 |             [cells.cell :refer [cell]]
   4 |             [cells.lib :as cell
   5 |              :refer [interval timeout fetch geo-location with-view]
   6 |              :refer-macros [wait]]
--------------------------------------------------------------------------------
Invalid :refer, var shapes.core/fish does not exist

shadow-cljs will use the version from the project IF installed, otherwise global

it prints the version when starting

ok, I can just remove the fish then it compiles with the cljs.js errors at runtime

investigating

oh i think i just pushed an update for that

maybe a minute after you tried

fish was removed

from shadow-eval.user

why do you require cljs.js in the first place?

the issue is that cljs/js.clj can never be loaded

but it seems like require tries to load it no matter what

not sure why

https://github.com/clojure/clojurescript/blob/master/src/main/cljs/cljs/js.clj#L13 this line prevents it from ever being compilable

i require cljs.js just because i want to use shadow-eval to sort out all these issues that are also present in maria, without all the complexity of maria at hand

and it is very useful to be able to use all the stuff in cljs.js

but why does maria need the require? the shadow-eval.user already ensures that it is loaded?

that works, its only the eval require that doesn’t work

the user might create a namespace and want to refer it

ah

(require '[cljs.js :as cljsjs]) (cljsjs/eval-str ...)

or whatever

and i think this will happen with any namespace that requires macros which don’t work in self-host. but we still want to be able to use those namespaces.

so bootstrap within boostrap, my head hurts already 😉

so the only thing I can think of is to ignore the load request

and just pretend that its loaded

but that just delays the problem until the user tries to use a macro from that namespace

hehe yeah. i think this all comes down to functioning :exclude or :no-follow or :ana-only or whatever options, ways of somehow saying “i have these namespaces which are preloaded and i just want to load the cache”

but that’s a transitive dep

hmm

it’s not (require-macros cljs.js)

doesn’t matter, its asks for the macros since it has a self require

i think the point of :no-follow or whatever is that the user doesn’t need transitive macro deps if cljs.js is already compiled?

it was compiled using those macros, but we don’t need those macros to use its compiled fns

just like we should be able to use other compiled namespaces which used other non-selfhost-friendly macros at compile time?

yes, :exclude is ignored in this case. I can fix that

however .. if the user directly tries to use a macro

that will lead to an error

am wondering what exactly :exclude does or if that is the right name

yes

:exclude excludes a macro namespace from compilation

it does NOT exclude it from loading (currently)

it does exclude it when loading via transitive deps

it does not exclude when loading directly

and thats the issue here

i think in cljs-live :exclude had the effect of only loading the analysis cache for the thing

it has no problem with shadow-eval.user loading the cljs.js, it does the correct thing here

the issue is who tries to load

but it does have a problem with (ns my.app (:require [cljs.js :as cljs]))

yes, but that again is the self-host compiler trying to directly load it

shadow-eval.user is loaded by shadow.cljs.bootstrap.browser which is aware of the exclude

long story short. I can work around this easily

but it WILL fail if the user tries to use a macro

that is unavoidable

well not unavoidable

we could just fix cljs/js.clj 😉

oh i mean for all the other use cases 🙂

right

i should set up an example of a namespace that uses the go macro

with the right exclude, it should be usable

no, that will never work

(the precompiled namespace, not go in self-host)

right ok, precompiled yes

let me fix the exclude issue

fixed in 2.0.11

fantastic 🎉

works for me

@thheller re: webserver w/ ring handler - that would be great. i used that functionality of figwheel like so: https://github.com/mhuebert/figwheel-pushstate-server

one thing, with bootstrap’s release mode, i think it may need to be combined with a clean command of some kind. otherwise the bootstrap directory can’t really be deployed without deleting it / rebuilding manually, because it fills up with multiple versions

in my experience thats not always a bad thing

I used to delete all old versions too but sometimes someone would access my site with a badly implement cache

that had cached html but not cached js

so they would attempt to download “old” js which had been deleted

never figured out how that was caused but it happened often enough to make me keep old version forever

granted forever is a bit extreme

but yeah some type of clean should probably exist

I’m always hesitant to delete user files though

yeah I should probably just deploy these files on a different system than the main app. then it makes sense to be ‘additive’

@mhuebert 2.0.12 -> :devtools {:http-root ... :http-handler shadow.http.push-state/handle} always returns index.html if files don’t exist

or point it to any other clojure fn you like

feels a bit rushed but maybe it works for you

off to bed, gn8

@thheller wow that was fast! 🙂 i think content-type headers need to be set to html, if i navigate to something like localhost:8702/intro it downloads as a file instead of opening