Wasm_of_ocaml support #11093

vouillon · 2024-11-04T11:50:01Z

@hhugo @rgrinberg This is an alternative design. What do you think about it?

This adds a wasm mode. All the js_of_ocaml options are duplicated as wasm_of_ocaml options. There is a (js_of_ocaml (enabled_if <blang expression>)) and (wasm_of_ocaml (enabled_if <blang expression>)) option in env and executable stanzas to control whether the js and wasm modes are enabled.

I have not updated the documentation yet.

samoht · 2024-11-04T13:01:39Z

doc/wasmoo.rst

+.. code:: console
+
+   $ dune build ./foo.bc.wasm.js
+   $ node _build/default/foo.bc.wasm.js


Why do I need to specify the submodes here? Dune could find out what to build for this wasm.js target on its own, isn't it?

The documentation has not been updated. That should be the following above:

(executable (name foo) (modes wasm))

But yes, you can just use dune build.

Not that this was just copied from the Js_of_ocaml part of the doc:

dune/doc/jsoo.rst

Lines 38 to 44 in b409963

And then request the ``.js`` target:

.. code:: console

$ dune build ./foo.bc.js

$ node _build/default/foo.bc.js

hello from js

hhugo

From a quick look, I think I prefer this approach better.

It seems you have not implemented the generation of a bc.js when js_of_ocaml is disabled, do you still want it ?

hhugo · 2024-11-04T13:21:33Z

src/dune_rules/exe_rules.ml

+          Js_of_ocaml.Mode.Set.inter jsoo_enabled_modes jsoo_is_whole_program
+        in
+        let bytecode_exe_needed =
+          L.Map.foldi


L.Map.exists

hhugo · 2024-11-04T13:33:32Z

src/dune_rules/inline_tests.ml

+              | Wasm -> Exe.Linkage.wasm
+            in
+            if Js_of_ocaml.Mode.Pair.select ~mode:jsoo_mode jsoo_is_whole_program
+            then [ Exe.Linkage.byte_for_jsoo; linkage ]


Don't we have an issue if byte_for_jsoo appears twice (once for js and once for wasm)

hhugo · 2024-11-04T13:36:27Z

src/dune_rules/jsoo/js_of_ocaml.ml

+    | JS
+    | Wasm
+
+  let equal = Poly.equal


Suggested change

let equal = Poly.equal

let equal (a : t) b = Poly.equal a b

hhugo · 2024-11-04T13:38:15Z

src/dune_rules/jsoo/js_of_ocaml.ml

+    let union = Pair.map2 ~f:( || )
+    let is_empty (x : t) = not (x.js || x.wasm)
+
+    let to_list (x : t) =


Suggested change

let to_list (x : t) =

let to_list ({ wasm; js}: t) =

hhugo · 2024-11-04T13:38:57Z

src/dune_rules/jsoo/js_of_ocaml.ml

+      match mode with
+      | Mode.JS -> "js"


Suggested change

match mode with

| Mode.JS -> "js"

match (mode : Mode.t) with

| JS -> "js"

hhugo · 2024-11-04T13:41:38Z

src/dune_rules/jsoo/js_of_ocaml.ml

@@ -242,6 +358,7 @@ module Env = struct
    ; sourcemap = None
    ; runtest_alias = None
    ; flags = Flags.default ~profile
+    ; enabled_if = Some Blang.true_


Why Some and not None ?

This is so that we don't need a case for None there:

dune/src/dune_rules/jsoo/jsoo_rules.ml

Lines 18 to 20 in ec321a5

(match parent.enabled_if with

| Some (Const default) -> default

| _ -> assert false)

dune/src/dune_rules/jsoo/jsoo_rules.ml

Lines 603 to 605 in ec321a5

(match js_of_ocaml.enabled_if with

| Some (Const default) -> Memo.return default

| _ -> assert false)

I think it would be either to read if there were less hidden invariants. What do you think of the following ?

diff --git a/src/dune_rules/jsoo/js_of_ocaml.ml b/src/dune_rules/jsoo/js_of_ocaml.ml index dc2fad30a..2c8c8e185 100644 --- a/src/dune_rules/jsoo/js_of_ocaml.ml +++ b/src/dune_rules/jsoo/js_of_ocaml.ml @@ -358,7 +358,7 @@ module Env = struct ; sourcemap = None ; runtest_alias = None ; flags = Flags.default ~profile - ; enabled_if = Some Blang.true_ + ; enabled_if = None } ;; end diff --git a/src/dune_rules/jsoo/jsoo_rules.ml b/src/dune_rules/jsoo/jsoo_rules.ml index b10ac8f27..98b29afeb 100644 --- a/src/dune_rules/jsoo/jsoo_rules.ml +++ b/src/dune_rules/jsoo/jsoo_rules.ml @@ -12,18 +12,16 @@ let compute_env ~mode = let local = Js_of_ocaml.Mode.select ~mode local.js_of_ocaml local.wasm_of_ocaml in let* parent = parent in let+ enabled_if = - match local.enabled_if with - | None -> - Memo.return - (match parent.enabled_if with - | Some (Const default) -> default - | _ -> assert false) - | Some enabled_if -> Expander.eval_blang expander enabled_if + match Option.first_some local.enabled_if parent.enabled_if with + | Some enabled_if -> + let+ v = Expander.eval_blang expander enabled_if in + Some (Blang.Const v) + | None -> Memo.return None in { Js_of_ocaml.Env.compilation_mode = Option.first_some local.compilation_mode parent.compilation_mode ; sourcemap = Option.first_some local.sourcemap parent.sourcemap - ; enabled_if = Some (Const enabled_if) + ; enabled_if ; runtest_alias = Option.first_some local.runtest_alias parent.runtest_alias ; flags = Js_of_ocaml.Flags.make @@ -601,8 +599,8 @@ let jsoo_enabled | None -> let* js_of_ocaml = jsoo_env ~dir ~mode in (match js_of_ocaml.enabled_if with - | Some (Const default) -> Memo.return default - | _ -> assert false) + | Some enabled_if -> eval enabled_if + | None -> Memo.return true) ;; let jsoo_enabled_modes ~expander ~dir ~in_context =

hhugo · 2024-11-04T13:49:07Z

src/dune_rules/jsoo/jsoo_rules.ml

    | Some x -> Memo.return x
  in
+  let runtime_files = javascript_files @ wasm_files in


This looks weird. Is it the case that wasm_files is always empty when mode is js ?

Yes, that's the case:

dune/src/dune_rules/jsoo/js_of_ocaml.ml

Lines 227 to 234 in ec321a5

and+ wasm_files =

match mode with

| Mode.JS -> return []

| Wasm ->

field

"wasm_files"

(Dune_lang.Syntax.since Stanza.syntax (3, 17) >>> repeat string)

~default:[]

Maybe we could add an assertion.

hhugo · 2024-11-04T14:04:30Z

@vouillon could you submit a version of ocsigen/js_of_ocaml#1724 using this PR ?

vouillon · 2024-11-04T15:23:07Z

It seems you have not implemented the generation of a bc.js when js_of_ocaml is disabled, do you still want it ?

It does not fit well. For the wasm_of_ocaml test suite, I have switched to using dune test stanzas and only had to add a few rules to copy a bc.wasm.js file to bc.js.

vouillon · 2024-11-04T15:42:06Z

See ocsigen/js_of_ocaml#1726

hhugo · 2024-11-05T10:35:41Z

This looks good, I left one last suggestion. The documentation still needs to be updated

Signed-off-by: Jérôme Vouillon <[email protected]>

Signed-off-by: Rudi Grinberg <[email protected]>

Signed-off-by: Jérôme Vouillon <[email protected]>

hhugo · 2024-11-05T15:03:32Z

src/dune_rules/install_rules.ml

@@ -110,11 +110,17 @@ end = struct
      (List.rev_concat_map
         ~f:(List.rev_map ~f:(fun f -> Section.Lib, f))
         (let { Mode.Dict.byte; native } = Lib_info.archives lib in
+          let jsoo_files =
+            (* A same runtime file can be used both for jsoo and wasmoo *)


Does that comment still make sense now that jsoo and wasm options are different ?

Yes, it still makes sense. In particular, one may want to pass a JavaScript file to wasm_of_ocaml to get some primitive purity information. Or one need to include some JavaScript code both when generating JavaScript and Wasm code. For instance, I have this for the virtual_dom package:

(js_of_ocaml (javascript_files ../lib/virtualdom.compiled.js ./thunk.js ./hooks.js)) (wasm_of_ocaml (javascript_files ../lib/virtualdom.compiled.js ./thunk.js ./hooks.js))

hhugo · 2024-11-05T15:07:31Z

src/dune_rules/jsoo/jsoo_rules.ml

+           let name = Path.basename fn in
+           let dir = in_build_dir build_context ~config [ lib_name ] in
+           let in_context =
+             { Js_of_ocaml.In_context.flags = Js_of_ocaml.Flags.standard


Should this be Js_of_ocaml.In_context.default ?

This is your code...
But yes, that should work.

This is your code...

I know

hhugo · 2024-11-05T15:09:49Z

src/dune_rules/stanzas/buildable.ml

+      (Js_of_ocaml.In_buildable.decode ~executable ~mode:JS)
+      ~default:Js_of_ocaml.In_buildable.default
+  and+ wasm_of_ocaml =
+    let executable =


this could be lifted up and shared with the one inside js_of_ocaml above

hhugo · 2024-11-05T15:25:40Z

@hhugo I'm not completely sure what to do about the ```(wasm_of_ocaml (sourcemap ))option. The distinction betweenfile` and `inline` does not really make sense: there is no inline source maps in Wasm and the source maps are going to be in the `.bc.wasm.asset` directory. I can keep the three options `no`, file or ``inline`` to stay in line with Js_of_ocaml. Or I can have just two options ``yes`` / ``no``.

I don't know either. We could also let dune complain if one tries to use inline with wasm.

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon · 2024-11-05T15:40:35Z

I don't know either. We could also let dune complain if one tries to use inline with wasm.

Or rather it should complain for file since one does not generate a separate .map target file?

hhugo · 2024-11-05T15:45:38Z

src/dune_rules/jsoo/js_of_ocaml.ml

-               "compilation_mode"
-               (Dune_lang.Syntax.since Stanza.syntax (3, 17) >>> Compilation_mode.decode)
-           else return None
+           only_in_library


This should be only_in_executable

Signed-off-by: Jérôme Vouillon <[email protected]>

hhugo · 2024-11-05T16:22:58Z

doc/wasmoo.rst

+=================
+
+Dune has full support for building wasm_of_ocaml libraries and executables transparently.
+There's no need to customise or enable anything to compile OCaml


Should the documentation mention the external dependencies that are required for the build ?

I now explicitly point to the github repository above.

hhugo

Approved. One will need to squash commits before merging

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon · 2024-11-05T17:09:23Z

I have made another clean-up pass.
I have also disallowed the enable_if option in libraries, like other build-related options.

hhugo · 2024-11-05T18:44:54Z

.github/workflows/workflow.yml

+        with:
+          ocaml-compiler: 4.14.x
+          opam-pin: false
+          opam-depext: false


Opam depext is unused in setup-ocaml v3

Signed-off-by: Jérôme Vouillon <[email protected]>

rgrinberg · 2024-11-05T23:01:01Z

LGTM. Feel free to send the docs in another PR

vouillon requested a review from christinerose as a code owner November 4, 2024 11:50

vouillon force-pushed the wasm_of_ocaml-alternative branch from 3c35724 to 8122af8 Compare November 4, 2024 11:53

samoht reviewed Nov 4, 2024

View reviewed changes

vouillon force-pushed the wasm_of_ocaml-alternative branch 2 times, most recently from 83731d5 to ec321a5 Compare November 4, 2024 13:20

hhugo reviewed Nov 4, 2024

View reviewed changes

vouillon mentioned this pull request Nov 4, 2024

Wasm_of_ocaml with alternative dune support ocsigen/js_of_ocaml#1726

Closed

vouillon force-pushed the wasm_of_ocaml-alternative branch from 9f3eb50 to 4a389d5 Compare November 4, 2024 20:27

maiste added the jsoo label Nov 5, 2024

vouillon mentioned this pull request Nov 5, 2024

Revised dune support ocaml-wasm/wasm_of_ocaml#126

Merged

vouillon force-pushed the wasm_of_ocaml-alternative branch from 04c28d2 to a837c52 Compare November 5, 2024 11:44

vouillon and others added 15 commits November 5, 2024 12:46

Add wasm_files option and wasmoo_runtime META file entry

f1d0369

Signed-off-by: Jérôme Vouillon <[email protected]>

Compilation to Wasm using Wasm_of_ocaml

774d877

Signed-off-by: Jérôme Vouillon <[email protected]>

Update documentation

5da9c12

Signed-off-by: Jérôme Vouillon <[email protected]>

Add tests

fae1f0c

Signed-off-by: Jérôme Vouillon <[email protected]>

Changes

7ba44dd

Signed-off-by: Jérôme Vouillon <[email protected]>

ci: enable wasm_of_ocaml tests

a1cbe1f

Signed-off-by: Jérôme Vouillon <[email protected]>

Wording changes

5c1c339

Signed-off-by: Jérôme Vouillon <[email protected]>

_

4ae8cc4

Signed-off-by: Rudi Grinberg <[email protected]>

_

e26ece4

Signed-off-by: Rudi Grinberg <[email protected]>

_

21ca0ca

Signed-off-by: Rudi Grinberg <[email protected]>

_

3fb2541

Signed-off-by: Rudi Grinberg <[email protected]>

_

583ba09

Signed-off-by: Rudi Grinberg <[email protected]>

_

0e10ece

Signed-off-by: Rudi Grinberg <[email protected]>

Wasm_of_ocaml support: alternative design

f4e3c59

Signed-off-by: Jérôme Vouillon <[email protected]>

CR

409a0f8

Signed-off-by: Jérôme Vouillon <[email protected]>

Updated documentation

fd5acfc

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon force-pushed the wasm_of_ocaml-alternative branch from afff2b0 to eba508f Compare November 5, 2024 12:54

Some clean-up

d31295f

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon force-pushed the wasm_of_ocaml-alternative branch from eba508f to d31295f Compare November 5, 2024 13:06

hhugo reviewed Nov 5, 2024

View reviewed changes

CR

be2bd4a

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon force-pushed the wasm_of_ocaml-alternative branch from daa107e to be2bd4a Compare November 5, 2024 15:38

hhugo reviewed Nov 5, 2024

View reviewed changes

vouillon added 2 commits November 5, 2024 16:49

CR

17e954a

Signed-off-by: Jérôme Vouillon <[email protected]>

Disallow (wasm_of_ocaml (sourcemap file))

aea189c

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon force-pushed the wasm_of_ocaml-alternative branch from 734967d to aea189c Compare November 5, 2024 15:51

hhugo reviewed Nov 5, 2024

View reviewed changes

hhugo approved these changes Nov 5, 2024

View reviewed changes

vouillon added 3 commits November 5, 2024 18:04

Clean-up

2c0c5f6

Signed-off-by: Jérôme Vouillon <[email protected]>

enabled_if not allowed in libraries

a4bc007

Signed-off-by: Jérôme Vouillon <[email protected]>

Doc: how to install wasm_of_ocaml

f76b05e

Signed-off-by: Jérôme Vouillon <[email protected]>

vouillon force-pushed the wasm_of_ocaml-alternative branch from 9c5c761 to f76b05e Compare November 5, 2024 17:04

vouillon changed the title ~~Wasm of ocaml support: alternative design~~ Wasm of ocaml support Nov 5, 2024

vouillon changed the title ~~Wasm of ocaml support~~ Wasm_of_ocaml support Nov 5, 2024

hhugo reviewed Nov 5, 2024

View reviewed changes

CR

906b5e0

Signed-off-by: Jérôme Vouillon <[email protected]>

rgrinberg merged commit 3995bb5 into ocaml:main Nov 5, 2024
28 checks passed

hhugo mentioned this pull request Nov 6, 2024

Merge wasm_of_ocaml ocsigen/js_of_ocaml#1724

Open

vouillon deleted the wasm_of_ocaml-alternative branch November 6, 2024 10:19

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Wasm_of_ocaml support #11093

Wasm_of_ocaml support #11093

vouillon commented Nov 4, 2024

samoht Nov 4, 2024

vouillon Nov 4, 2024

hhugo left a comment

hhugo Nov 4, 2024

hhugo Nov 4, 2024

hhugo Nov 4, 2024

hhugo Nov 4, 2024

hhugo Nov 4, 2024

hhugo Nov 4, 2024

vouillon Nov 4, 2024

hhugo Nov 5, 2024

hhugo Nov 4, 2024

vouillon Nov 4, 2024

hhugo Nov 4, 2024

hhugo commented Nov 4, 2024

vouillon commented Nov 4, 2024

vouillon commented Nov 4, 2024

hhugo commented Nov 5, 2024

hhugo Nov 5, 2024

vouillon Nov 5, 2024

hhugo Nov 5, 2024

vouillon Nov 5, 2024

hhugo Nov 5, 2024

hhugo Nov 5, 2024

hhugo commented Nov 5, 2024

vouillon commented Nov 5, 2024

hhugo Nov 5, 2024

hhugo Nov 5, 2024

vouillon Nov 5, 2024

hhugo left a comment

vouillon commented Nov 5, 2024

hhugo Nov 5, 2024

rgrinberg commented Nov 5, 2024

	And then request the ``.js`` target:

	.. code:: console

	$ dune build ./foo.bc.js
	$ node _build/default/foo.bc.js
	hello from js

	(match parent.enabled_if with
	\| Some (Const default) -> default
	\| _ -> assert false)

	(match js_of_ocaml.enabled_if with
	\| Some (Const default) -> Memo.return default
	\| _ -> assert false)

	and+ wasm_files =
	match mode with
	\| Mode.JS -> return []
	\| Wasm ->
	field
	"wasm_files"
	(Dune_lang.Syntax.since Stanza.syntax (3, 17) >>> repeat string)
	~default:[]

Wasm_of_ocaml support #11093

Wasm_of_ocaml support #11093

Conversation

vouillon commented Nov 4, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhugo left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhugo commented Nov 4, 2024

vouillon commented Nov 4, 2024

vouillon commented Nov 4, 2024

hhugo commented Nov 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhugo commented Nov 5, 2024

vouillon commented Nov 5, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhugo left a comment

Choose a reason for hiding this comment

vouillon commented Nov 5, 2024

Choose a reason for hiding this comment

rgrinberg commented Nov 5, 2024