Removed foreign declarations yield a warning.

It turned

exports.a = "❄";

into

exports.a = "�";

after running zephyr on this recipe after clicking on plus or minus it will throws an error.

the problem is inside the halogen free monad hooks, but I could not find the culprit.

I don't see the option referenced anywhere except in the argument parser.

Hello!

When attempting to build both zephyr-0.3.0 yesterday and zephyr-0.3.1 today the following errors have come up both times:

zephyr> build (lib + exe)
Preprocessing library for zephyr-0.3.1..
Building library for zephyr-0.3.1..
[1 of 7] Compiling Language.PureScript.DCE.Constants
[2 of 7] Compiling Language.PureScript.DCE.Errors
[3 of 7] Compiling Language.PureScript.DCE.Foreign

/mnt/c/code/hs/zephyr/src/Language/PureScript/DCE/Foreign.hs:135:21: error:
    Not in scope: data constructor ‘JSAsyncFunction’
    |
135 |     isUsedInStmt n (JSAsyncFunction _ _ _ _ es _ (JSBlock _ ss _) _) =
    |                     ^^^^^^^^^^^^^^^

/mnt/c/code/hs/zephyr/src/Language/PureScript/DCE/Foreign.hs:165:21: error:
    Not in scope: data constructor ‘JSAwaitExpression’
    Perhaps you meant one of these:
      ‘JSAssignExpression’ (imported from Language.JavaScript.Parser.AST),
      ‘JSCallExpression’ (imported from Language.JavaScript.Parser.AST),
      ‘JSClassExpression’ (imported from Language.JavaScript.Parser.AST)
    |
165 |     isUsedInExpr n (JSAwaitExpression _ e) = isUsedInExpr n e
    |                     ^^^^^^^^^^^^^^^^^


--  While building package zephyr-0.3.1 using:
      /home/me/.stack/setup-exe-cache/x86_64-linux/Cabal-simple_mPHDZzAJ_2.4.0.1_ghc-8.6.5 --builddir=.w/home/wsl1/dist/x86_64-linux/Cabal-2.4.0.1 build lib:zephyr exe:zephyr --ghc-options " -fdiagnostics-color=always"
    Process exited with code: ExitFailure 1

On investigation, the two mentioned constructors JSAsyncFunction and JSAwaitExpression only appear to have been included in Language.JavaScript.Parser.AST.JSExpression in language-javascript-0.7.1.0 (and were not present in 0.7.0.0).

The stack.yaml specifically refers to language-javascript-0.7.0.0 and the zephyr.cabal file has a version bound of language-javascript ^>=0.7.

For reference, I'm building with stack build --copy-bins using stack-2.3.1 on WSL1 Ubuntu 18.04.

Bumping the version in the stack.yaml to 0.7.1.0 allows everything to build fine. :)

thomashoneyman/purescript-halogen-realworld#65

add-path was deprecated today.

https://github.com/coot/zephyr/blob/4aa399b9055266e8ef463687b491979f468bcd93/.github/workflows/posix.yml#L21

Needs to be replaced with this syntax.

https://docs.github.com/en/free-pro-team@latest/actions/reference/workflow-commands-for-github-actions#setting-an-environment-variable

echo "action_state=yellow" >> $GITHUB_ENV

Simply running

purs compile --codegen corefn (spago sources)
zephyr Main

gives me this error:

zephyr: output/Affjax/foreign.js: copyFile:atomicCopyFileContents:withReplacementFile:copyFileToHandle:openBinaryFile: does not exist (No such file or directory)

And ideally psc-package.

I tried running zephyr on the Spork todo-mvc example https://github.com/natefaubion/purescript-spork/blob/master/package.json#L6

yarn example:todo-mvc -- -- -g corefn
...
zephyr Todo.Main.main

And zephyr ate up about 30min of CPU time across 6 cores, after which I killed it. It seems unlikely that it was doing productive work for that amount of time.

$ zephyr -f Main.main

Error: Failed parsing:
  output/Data.TraversableWithIndex/corefn.json
        output/Data.Semigroup.Last/corefn.json
        output/Control.Monad.Rec.Class/corefn.json
        output/Data.Monoid.Generic/corefn.json
        output/Data.Show.Generic/corefn.json
        ... (139 more)

The repo mentions "tree shaking tool" a lot, but grammatically it should be compound adjective as "tree-shaking tool" in the README and other places through the repo.

From Grammar Monster:

You can use a hyphen (or hyphens) to link the words in a compound adjective to show it is a single adjective. (A compound adjective is a single adjective that is made up of more than one word.) For example:

• two-seater aircraft
• never-to-be-forgotten experience

An existing example is in Rollup's documentation

I've been testing the spork todo-mvc example, and it is failing when using zephyrs output. The app loads, but I think there is a bad transform, particularly around pattern matches.

For example, some original output:

  var update = function (model) {
      return function (v) {
          if (v instanceof None) {
              return Spork_Transition.purely(model);
          };
          if (v instanceof UpdatePending) {
              return Spork_Transition.purely({
                  pending: v.value0,
                  fresh: model.fresh,
                  todos: model.todos,
                  visibility: model.visibility
              });
          };
  ...

And the corresponding output from zephyr:

  var update = function (model) {
      return function (v) {
          if (v instanceof None) {
              return Spork_Transition.purely(model);
          };
          if (v instanceof UpdatePending) {
              return Spork_Transition.purely({
                  pending: v,
                  fresh: model.fresh,
                  todos: model.todos,
                  visibility: model.visibility
              });
          };
  ...

In particular this diff:
pending: v.value0,
pending: v,

This is giving me a dreaded [object Object] 🙀

This might actually be intentional and we will need to update the FFI in the "offending" module. I'm using react-basic-hooks.
And that has code like this:

exports.useEffect_ = function (eq, deps, effect) {
  var memoizedKey = exports.useMemo_(eq, deps);
  React.useEffect(effect, [memoizedKey]);
};

I don't use useMemo_ (in fact I can't since it's not exported) but I use useEffect which in turn uses useEffect_ which then can't find useMemo_.

Would you say this is a bug or should the FFI be rewritten in a way that avoids this?

Edit: Yeah I just had a look at dce-output and that's what's happening. I will try to rewrite the FFI.

Hey @coot

Happy new year 🥳

I think it'd be great if zephyr worked with the new version of the compiler when it gets released.

I've tried to incorporate some older changes from @kl0tl but something seems amiss.

It'd be great if you could have a look at the changes here: https://github.com/working-group-purescript-es/zephyr/pull/1/files

#10, other than turning off evaluation we could

• keep two ASTs - evaluated and the original
• use the original one for printing js modules
• use the evaluated to perform dce

Currently this is still pointed at 0.13.8, but 0.14.x has been out for a while. It appears to work, but a banner comment appears at the top of the file saying it's generated with PureScript 0.13.8 which raises some red flags to me.

As mentioned in #64, Nix makes setting up a build/dev environment much easier.

I'm new to the pureacript ecosystem, what is the best way to use this with spago and parcel?

zephyr: An internal error occurred during compilation: "Missing value in mnLookup: App.Domain.
Route"
Please report this at https://github.com/purescript/purescript/issues
CallStack (from HasCallStack):
  error, called at src/Language/PureScript/Crash.hs:24:3 in purescript-0.11.6-E9oYDASU5kVKQAjU
68tzan:Language.PureScript.Crash
  internalError, called at src/Language/PureScript/CodeGen/JS.hs:128:35 in purescript-0.11.6-E
9oYDASU5kVKQAjU68tzan:Language.PureScript.CodeGen.JS

The cause is that the eval function will evaluate Var{} expressions. When evaluating Var{} expressions this should be limited to only those that points to number, character or boolean literal, but preserve string (because it might be repeated many times), array and object literals (this is the direct cause of the failure: they may contain references to other modules).

Hey there. After updating to PureScript 0.15 zephyr fails with

$ zephyr -f Main.main
Error: Failed parsing:
  output/Control.Comonad.Traced/corefn.json
        output/Control.Monad.State/corefn.json
        output/Control.Monad.Error.Class/corefn.json
        output/Control.Monad.Identity.Trans/corefn.json
        output/Main/corefn.json
        ... (346 more)
error Command failed with exit code 1.

This is the offending branch https://github.com/dnikolovv/purescriptify/tree/purescript-0.15. You can run yarn run bundle to reproduce.

Being able to install via NPM would probably be nice and lower the barrier to using zephyr

Downloading the binary for zephyr 0.3.2 fails on my Arch Linux system with the following error.

/zephyr: error while loading shared libraries: libtinfo.so.5: cannot open shared object file: No such file or directory

I worked around this issue by simply symlinking my libtinfo.so.5 library to libtinfo.so.6.

sudo ln -s /usr/lib/libtinfo.so.6 usr/liblibtinfo.so.5

I don't know if this work-around will go belly-up later (I haven't tested it), but I thought I'd pass it along in case it is helpful.

Haskell and PureScript newbie here; I ran into this problem when trying to build purescript-halogen-realworld, which failed due to #38. Any insights are greatly appreciated!

• macOS 10.15.5 (Catalina)
• cabal 3.2.0.0 (via homebrew)
• ghc 8.10.1 (via homebrew)

cabal build exe:zephyr   
Warning: Requested index-state2020-06-09T00:00:00Z is newer than
'hackage.haskell.org'! Falling back to older state (2020-06-08T22:11:02Z).
Resolving dependencies...
cabal: Could not resolve dependencies:
[__0] trying: zephyr-0.3.0 (user goal)
[__1] trying: purescript-0.13.8 (dependency of zephyr)
[__2] next goal: base (dependency of zephyr)
[__2] rejecting: base-4.14.0.0/installed-4.14.0.0 (conflict: purescript =>
base>=4.11 && <4.13)
[__2] skipping: base-4.14.0.0, base-4.13.0.0 (has the same characteristics
that caused the previous version to fail: excluded by constraint '>=4.11 &&
<4.13' from 'purescript')
[__2] rejecting: base-4.12.0.0, base-4.11.1.0, base-4.11.0.0, base-4.10.1.0,
base-4.10.0.0, base-4.9.1.0, base-4.9.0.0, base-4.8.2.0, base-4.8.1.0,
base-4.8.0.0, base-4.7.0.2, base-4.7.0.1, base-4.7.0.0, base-4.6.0.1,
base-4.6.0.0, base-4.5.1.0, base-4.5.0.0, base-4.4.1.0, base-4.4.0.0,
base-4.3.1.0, base-4.3.0.0, base-4.2.0.2, base-4.2.0.1, base-4.2.0.0,
base-4.1.0.0, base-4.0.0.0, base-3.0.3.2, base-3.0.3.1 (constraint from
non-upgradeable package requires installed instance)
[__2] fail (backjumping, conflict set: base, purescript, zephyr)
After searching the rest of the dependency tree exhaustively, these were the
goals I've had most trouble fulfilling: purescript, zephyr, base

For the record, a workaround is fixing the purescript version to 0.13.6 in purescript-halogen-realworld/package.json:

"purescript": "0.13.6"

zephyr binary isn't compatible with musl and requires glibc (what about uclibc?). The errors I was getting trying to use Alpine Linux while trying to use the v0.3.2 release binary wasn't very clear but eventually I tracked it down to this incompatibility. It could be useful to mention this somewhere in the docs.

At work we use Zephyr built from source with PureScript 0.13.3 (after #21 landed) and it works fine with files compiled with other PureScript 0.13.x versions.

However, I recently switched us over to use the Zephyr which is included in easy-purescript-nix, which is the last release: 0.2.1. Running this version on a project compiled with PureScript 0.13.6 on Linux causes this error:

zephyr: dce-output/Web.Fetch.Integrity/externs.json: openBinaryFile: does not exist (No such file or directory)

I put together a reproduction of most of our PureScript dependencies, which uses version 0.13.6 of the compiler and version 0.2.1 of Zephyr. But this reproduction fails when compiled and run on Mac OS with a different error:

zephyr: output/Web.HTML.HTMLTableHeaderCellElement/corefn.json: openBinaryFile: resource exhausted (Too many open files)

This is despite setting a large upper limit on file handles on Mac OS, as described in #9.

I think these errors are red herrings and that projects running on 0.13.x ought to use versions of Zephyr compiled with PureScript 0.13.x. I have been having trouble updating the reproduction to compile Zephyr from source on the master branch using PureScript 0.13.3, but I'll test that out tomorrow and make an update here.

That said: I believe it's worth making a new release compiled with PureScript 0.13.3 which could be added to easy-purescript-nix for everyone to use, regardless of whether I can fully reproduce this issue outside our codebase here.

purs compile bower_components/purescript-*/src/**/*.purs src/**/*.purs test/**/*test && zephyr -O1 Test.Main.main && node -e "require('./dce-output/Test.Main').main()"

[x] purescript-argonaut
[x] purescript-argonaut-core
[x] purescript-profunctor-lenses
[x] pulp (test suite & pulp itself)

So it seems like the latest purescript version doesn't generate json externs anymore so running zephyr breaks with errors like:

zephyr: output/Affjax/externs.json: openBinaryFile: does not exist (No such file or directory)

Is it possible to integrate zephyr with purs loader?

I thought I would try out Zephyr and see if I could improve my build sizes, but I seem to be running into a problem when I run zephyr Main.main. Here's what I did.

First I checked out the zephyr repo and ran stack install, which added zephyr to my path. I have Stack Version 1.5.1 x86_64 hpack-0.17.1, PureScript 0.11.6, and Pulp 11.0.0. Then I went to my project's root and ran pulp build -- --dump-corefn which compiled my project. Then I ran zephyr Main.main and I got this error...

Error
Failed parsing:
  output/DOM.File.FileReader/corefn.json
        output/DOM.Event.WheelEvent/corefn.json
        output/Data.HeytingAlgebra/corefn.json
        output/DOM.HTML.HTMLMediaElement.NetworkState/corefn.json
        output/Control.Monad.Writer.Class/corefn.json
        output/Data.Bifunctor.Clown/corefn.json
        output/DOM.HTML.HTMLOptGroupElement/corefn.json
        output/Data.List.Lazy.NonEmpty/corefn.json
        output/Data.Distributive/corefn.json
        output/Data.Interval/corefn.json
        output/Control.Monad.Gen/corefn.json
        output/Data.Int.Bits/corefn.json
        output/Data.Monoid.Dual/corefn.json
        output/Data.Time.Component.Gen/corefn.json
        output/Data.Identity/corefn.json
        output/Data.Monoid/corefn.json
        output/DOM.HTML.HTMLTableElement/corefn.json
        output/Data.Char/corefn.json
        output/DOM.XHR.Types/corefn.json
        output/DOM.Node.NonElementParentNode/corefn.json
        output/Data.Traversable/corefn.json
        output/Control.Comonad.Store/corefn.json
        ....

I don't think there's anything unusual about the project I'm compiling. This is the bower.json file with my project dependencies.

Hi there,

It would be great if zephyr had a macos-arm64 (aka aarch64-darwin) build like purescript itself does it would allow to provide this architecture for the easy-purescript-nix tooling which doesn't currently support it because of https://github.com/justinwoo/easy-purescript-nix/blob/master/zephyr.nix#L20

Cheers,

Jun

PureScript's 0.13.8 release changed the format of externs to use a binary encoding PR here.

If you try to use Zephyr on a project compiled with 0.13.8 you'll see an error like this:

zephyr: output/Affjax/externs.json: openBinaryFile: does not exist (No such file or directory)

There are no longer externs.json files. Instead, this would be a externs.cbor file in the same location in the output directory.

We'll need to update Zephyr to accommodate the new format.

(As a related note, there has been some discussion (1, 2) about moving Zephyr into the compiler, which would help with staying up-to-date with big changes in the compiler. Do you have any interest in this?)

Thank you very much for this nice tool. I use it for purescript-native cpp code generation:
PureScript Shared Objects
It is essential in the sence that the ffi implementations are not complete and zephyr allowes me to generate only the code which is just needed and does not pull in references of unimplemented functions. One issue i stumbled across is that zephyr creates new files for each run, regardless if the contents has changed or not. This will trigger all depending compilations which should not necessary at all. I worked around this by comparing the file sizes with the previous run, if they match then the new file gets the timestamp of the older file. (zephyr.bash in the mentioned repo)
The second issue is that I would like to get the dependencies for groups of entry points. I use this for automatic assembling of the libraries and applications. I worked around this by calling zephyr for each group for different output directories. This should be possible in one run too.
If this makes sense for you I would like to hear from you. I could try to implement this if you give me some guidance. Thanks a lot!

It would be nice to remove dead branches of case expressions.

This is important for if statements:

if (Utils.production == "production")
  then ...
  else ...

For this case there could be a flag to provide value of Utils.production so zypher could infer that this is a production build.

I tried running zephyr on macOS 10.13 Beta (17A362a) in a clone of the pux-starter-app and got this:

± zephyr -f Main.main 
zephyr: output/Data.DateTime/corefn.json: openBinaryFile: resource exhausted (Too many open files)

You can work around the problem with these commands:

sudo sysctl -w kern.maxfiles=1000000
ulimit -S -n 8000

I was surprised to see that using zephyr can sometimes increase bundle sizes. Is there a mistake in my workflow?

git clone https://github.com/purescript-halogen/purescript-halogen-template.git bundle-test
cd bundle-test

# Workflow with just spago bundle-app and parcel (no zephyr)

npm run build-prod
du -hs prod/index.js dist/prod.*.js
# 316K    prod/index.js          # bundle-app output
# 128K    dist/prod.0a4f58ae.js  # minified by parcel

# Including zephyr

rm -r dist
spago build --purs-args '--codegen js,corefn' && zephyr Main.main --dce-foreign
purs bundle -o prod/index.js -m Main dce-output/**/*.js
parcel build prod/index.html
du -hs prod/index.js dist/prod.*.js
# 332K    prod/index.js          # zephyr + bundle-app output
# 132K    dist/prod.08b89072.js  # minified by parcel

https://discourse.purescript.org/t/recommended-tooling-for-purescript-in-2020/1615/4