use "use strict" when possible.

see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Functions_and_function_scope/Strict_mode

The array bounds check should be

var l2 = chan.data.array.length - chan.data.offset;

This makes Streams work properly.

mishoo/UglifyJS#179

js_of_ocaml generates JS with a lot of global caml_* variables. This could be avoided by putting the code which interact with these functions in an IIFE and exporting only the useful ones and/or using a namespace, e.g.:

var caml = (function(global, ns) {
  ns.foo = function caml_foo() { /* … */ };
  ns.bar = function caml_bar() { /* … */ };
  function caml_global_function() { /* … */ };

  // export global functions
  global.caml_global_function = caml_global_function;

  // return caml object
  return ns;
})(this, {});

Then use:

caml.foo();
caml.bar();

caml_global_function();

The psuedo-file system (from what I can gather) uses a call Js.register_file which in turn calls caml_ml_register_file (defined in js.ml).

The runtime defines a function joo_register_file (which is also mentioned in one place in the compiler). I believe joo_register_file and caml_ml_register_file should be the same thing.

To avoid re-compilation I have added

// missing?
function caml_ml_register_file(name,content) {
    joo_register_file(name,content)
}

to some custom javascript in my project and the filesystem stuff is somewhat working.

Because, why not :)

I'm experimenting with compiling a HTML parser using js_of_ocaml. I've looked at a few options, I'm currently working on Nethtml.

I managed to extract it from the Netsys / Unix dependencies, and aside from a few missing primitive warnings and the size of the JS file (200k) it works fairly well. In the 1.4 release.

Using the development branch (because -opt 3 fails for me in 1.4) I get the above TypeError. I tracked it down to code that first iterates over a list copying the values into a hashtable, and then runs List.assoc over the same list. It does more than that, but that seems to be the bulk of the cause.

I noticed in the dev branch, the JS generates long lists as separate arrays instead of using one big nested array. I confirmed this is related by shortening the list to 10 items which fits in a single JS array and doesn't trigger the issue.

I have simplified the replication case as follows:

let etable =
  [ "lt", 60;
  "gt", 62;
  "amp", 38;
  "quot", 34;
  "apos", 39;
  "nbsp", 160;
  "iexcl", 161;
  "cent", 162;
  "pound", 163;
  "curren", 164;

  (* comment this out to avoid the error *)
  "yen", 165;
  ]

let first_iter =
  let ht = Hashtbl.create 50 in
    List.iter (fun (name,value) -> (Hashtbl.add ht name value))
              etable

let second_assoc = List.assoc "lt" etable


(* just so the optimiser doesn't remove the above *)
let _ = (Js.Unsafe.coerce Dom_html.window)##table <- Js.wrap_callback (fun _ -> second_assoc)

(* this is here because I get "missing primitive caml_mul" if multiplication is never used *)
let _ = (Js.Unsafe.coerce Dom_html.window)##square <- Js.wrap_callback (fun x -> x * x)

I'll log the caml_mul thing as a separate issue.

In order to avoid repetitions, we now generate code like: d8=caml_ml_output_char.
This fails when caml_ml_output_char is not defined.

I can see several way to fix that:

• only share a primitive if the linkers knows about it;
• output d8=this.caml_ml_output_char (but we will get puzzling error messages telling that d8 is not defined);
• have the linker generate a short piece of code for each missing primitive.

The first option seems best. It should not be difficult to implement as the linker already has all the information during code generation.

I was wondering if you have any plans to make a new release compatible with 4.01 ? Thanks for the hard work!

It should simplify a lot the build system.
related : #25

float_of_string "0o123" works with joo runtime where it fails with regular runtime
float_of_string "0xffffp2" fails with joo runtime where it succeed with regular runtime

comparing 2 different Weak.t will succeed with js_of_ocaml where it should raise an exception.
We should either fix the runtime or document the deviation.

The current js_of_ocaml opam package seems extremely old.
Because of this, we cannot compile some codes in https://ocsigen.org/tutorial/tutowidgets, if we install js_of_ocaml via opam.
(* I'm sorry I cannot contribute because I don't know about opam packages at all. *)

The rest of this post is how I reached the conclution above.

I'm trying the tutorial https://ocsigen.org/tutorial/tutowidgets, and encountered an error Error: Unbound value Lwt_js_events.onload.
I installed js_of_ocaml by the command below.
$ opam update; opam install eliom
I found that the source codes here and those in my computer (~/.opam/3.12.1/lib/js_of_ocaml) are different, while the versions are the same (1.3.2).
The function Lwt_js_events.onload was added by the commit fb5e555, so I think the current js_of_ocaml opam package is built at least before May 03 2013.

thanks,

Regexp.global_replace and Regexp.replace_first share the same implementation, which cannot be correct.

One needs to see how these functions should interact with Regexp.regexp_with_flag, which currently allow to set the global flag.

When the -pretty option is set, we want to have variable names of the shape _x_originalname.

@hhugo fix attempt does not work...

But more generally, we should generate short variable names with an informative suffix, and this requires more work...

It would be nice to have support for the backtrace stuff.

If that's impossible/too hard please at least implement a stub for the new primitive of 4.01.0 (see ocaml/ocaml@2b6c0f0).

I don't know if it's related to js_of_ocaml, but I got this warning with the latest (69df95a) dev version of js_of_ocaml:

Use of getUserData() or setUserData() is deprecated.  Use WeakMap or element.dataset instead.

Constants are not compiled correctly.

This code:

    ["%int_mul", "caml_mul";
     "%int_div", "caml_div";
     "%int_mod", "caml_mod";
     "caml_int32_neg", "%int_neg";
     "caml_int32_add", "%int_add";
     "caml_int32_sub", "%int_sub";
     ....

is compiled into:

     _ayg_=
      [0,
       [0,_b_("caml_mul"),_b_("caml_mul")],
       [0,
        [0,_b_("caml_div"),_b_("caml_div")],
        [0,
         [0,_b_("caml_mod"),_b_("caml_mod")],
         [0,
          [0,_b_(_if_),_b_(_if_)],
          [0,
           [0,_b_(_g__),_b_(_g__)],
           [0,
            [0,_b_(_il_),_b_(_il_)],
            [0,
            ....

While compiling cumulus with the latest js_of_ocaml compiler (21ac186), I've been surprised by the following behavior:

$ js_of_ocaml -noruntime +js_of_ocaml/runtime.js +eliom.client/eliom_client.js +js_of_ocaml/weak.js +js_of_ocaml/classlist.js src/client/cumulus.byte -o src/client/cumulus.js
Missing primitives:
  __eliom_request_data
  …
$ js_of_ocaml -noruntime -pretty -debuginfo -noinline +js_of_ocaml/runtime.js +eliom.client/eliom_client.js +js_of_ocaml/weak.js +js_of_ocaml/classlist.js src/client/cumulus.byte -o src/client/cumulus.js
Missing primitives:
   …

Namely, the missing primitive __eliom_request_data appears only without the debug flags -pretty -debuginfo -noinline.

I see that printf's are automatically redirected to the browser console which is nice (printf is still my debugging tool of choice).

Is that behaviour documented somewhere ? And is it possible to capture that output to redirect it somewhere else (e.g. to make it visible on the webpage). It couldn't find anything about that in the docs.

and avoid duplicating the js parser/lexer

ocaml/opam-repository#1300

The bytecode interpreter uses caching to speed-up method look-up (instruction GETPUBMET). We should do the same.

• Our JavaScript syntax tree should support two kind of strings: sequences of bytes (which is what is currently available) and UTF8 strings.
• When outputting UTF8 strings, one does not need to escape bytes above 127.
• The JavaScript parser should unescape the UTF8 parsed strings.
• The JavaScript parser should unescape regular expressions.
• When generating a regular expression literal, one should check whether it contains only allowed characters.

(See the ECMAScript spec for details on the syntax of literal strings and regular expressions.)

If Hashtbl.add is used in a simple project there's an unrecognised dependency on caml_mul. If I add explicit multiplication to force the dependency everything works.

This does not happen in the 1.4 release.

Code:

let table =
    let ht = Hashtbl.create 0 in
        Hashtbl.add ht 0 0;
        ht


let _ = (Js.Unsafe.coerce Dom_html.window)##table <- Js.wrap_callback (fun _ -> table)

(* uncomment this to work around the issue *)
(* let _ = (Js.Unsafe.coerce Dom_html.window)##square <- Js.wrap_callback (fun x -> x * x) *)

In the Demos list, the second link (hyperbolic tree viewer) is broken.
link: http://ocsigen.org/js_of_ocaml/files/hyperbolic/index.html

Bruno

let ff a =
  let f c =
    c##prop1 <- true
  in
  f a;
  f a##prop2

This requires that a and a##prop2 have same type.

val ff : (< prop1 : < set : bool -> unit; .. > Js.gen_prop;
            prop2 : < get : 'a Js.t; .. > Js.gen_prop; .. >
            as 'a)
          Js.t -> unit

But this allows differet types.

let f1 c =
  c##prop1 <- true

let ff1 a =
  f1 a;
  f1 a##prop2


val ff1 : < prop1 : < set : bool -> unit; .. > Js.gen_prop;
            prop2 : < get : < prop1 : < set : bool -> unit; .. > Js.gen_prop; .. > Js.t;
                      .. > Js.gen_prop;
            .. >
          Js.t -> unit

https://developer.mozilla.org/en-US/docs/Web/Guide/Performance/Using_web_workers

https://github.com/besport/js_of_ocaml-old/issues/1

I got this missing primitive and I wanted to know where does it comes from. Because it should appears only if the unix package is used (directly or as a dependency) but I only use eliom.client.

$ js_of_ocaml -noruntime -pretty -debuginfo -noinline +js_of_ocaml/runtime.js +eliom.client/eliom_client.js +js_of_ocaml/weak.js +js_of_ocaml/classlist.js src/client/cumulus.byte -o src/client/cumulus.js
Missing primitives:
  …
  unix_inet_addr_of_string

The repo ocp-webedit needs 'js_of_ocaml_compiler-libs' findlib package to compile itself. To build it we need to use Fabrice's changes in opam repository
The most interesting file is js_of_ocaml.patch

Are this changes going to be integrated to upstream?

What are the ways to make Recursive modules to work ?
It seams to me that overriding part of CamlinternalMod could be a way ..
Does it seams feasible ? Are there any other ways ?

Why is that here ? Isn't it possible to put the json parser into the deriving lib ? I find it confusing (plus it seems not to work when I want to derive both json and show for example).

I would like to add support for topfind to iocamljs (with support from it's cousin iocamlserver).

This session is my first attempt at getting this working.

The .cmo file is loaded and parsed by the Topdirs.dir_load machinery, then fails in Parsebyte_code.compile while processing the SETGLOBAL instruction.

Any pointers on where to look next? I assume something along the lines of Parse_bytecode.from_channel needs to happen to set up a global environment?

Hi,

opam install eliom wants to upgrade js_of_caml from 1.4.0 to 9999. During installation the error bwlow occurs. To me it seems like one of the examples failed. The error message which asks to use --infer can be found in the menhir source.

/Jörgen

root@raspberrypi:~# opam install eliom
The following actions will be performed:
 - install deriving.9999 [required by eliom]
 - install ocsigenserver.9999 [required by eliom]
 - upgrade js_of_ocaml.1.4.0 to 9999 [required by eliom]
 - install eliom.9999
3 to install | 0 to reinstall | 1 to upgrade | 0 to downgrade | 0 to remove
Do you want to continue ? [Y/n] 

=-=-= Removing Packages =-=-=
Removing js_of_ocaml.1.4.0.
  ocamlfind remove js_of_ocaml

=-=-= Installing deriving.9999 =-=-=
deriving.9999 Fetching https://github.com/ocsigen/deriving.git
Copying ~/.opam/repo/ocsigendev/packages/deriving.9999/files/eliom.install to ~/.opam/system/build/deriving.9999/
Building deriving.9999:
  make
  make install
Installing deriving.9999.

=-=-= Installing ocsigenserver.9999 =-=-=
ocsigenserver.9999 Fetching https://github.com/ocsigen/ocsigenserver.git
Copying ~/.opam/repo/ocsigendev/packages/ocsigenserver.9999/files/ocsigenserver.install to ~/.opam/system/build/ocsigenserver.9999/
Building ocsigenserver.9999:
  sh configure --prefix /root/.opam/system --ocsigen-user root --ocsigen-group root --commandpipe /root/.opam/system/lib/ocsigenserver/var/run/ocsigenserver_command --logdir /root/.opam/system/lib/ocsigenserver/var/log/ocsigenserver --mandir /root/.opam/system/man/man1 --docdir /root/.opam/system/lib/ocsigenserver/share/doc/ocsigenserver --commandpipe /root/.opam/system/lib/ocsigenserver/var/run/ocsigenserver_command --staticpagesdir /root/.opam/system/lib/ocsigenserver/var/www --datadir /root/.opam/system/lib/ocsigenserver/var/lib/ocsigenserver --sysconfdir /root/.opam/system/lib/ocsigenserver/etc/ocsigenserver
  make
  make install
Installing ocsigenserver.9999.

=-=-= Installing js_of_ocaml.9999 =-=-=
js_of_ocaml.9999 Fetching https://github.com/ocsigen/js_of_ocaml.git
Copying ~/.opam/repo/ocsigendev/packages/js_of_ocaml.9999/files/js_of_ocaml.install to ~/.opam/system/build/js_of_ocaml.9999/
Building js_of_ocaml.9999:
  make
  make install BINDIR=/root/.opam/system/bin
[ERROR] The compilation of js_of_ocaml.9999 failed.
Removing js_of_ocaml.9999.
  ocamlfind remove js_of_ocaml
  ocamlfind remove js_of_ocaml_compiler

[ERROR] Due to some errors while processing js_of_ocaml.9999, the following actions will NOT proceed:
 - install eliom.9999

=-=-= ERROR RECOVERY [js_of_ocaml.9999] =-=-=

=-=-= Installing js_of_ocaml.1.4.0 =-=-=
js_of_ocaml.1.4.0 Downloading http://ocsigen.org/download/js_of_ocaml-1.4.tar.gz
Copying ~/.opam/packages/js_of_ocaml/js_of_ocaml.1.4.0/files/js_of_ocaml.install to ~/.opam/system/build/js_of_ocaml.1.4.0/
Building js_of_ocaml.1.4.0:
  make
  make install BINDIR=/root/.opam/system/bin
Installing js_of_ocaml.1.4.0.

=-=-= Installing eliom.9999 =-=-=
eliom.9999 Fetching https://github.com/ocsigen/eliom.git
Copying ~/.opam/repo/ocsigendev/packages/eliom.9999/files/eliom.install to ~/.opam/system/build/eliom.9999/
Building eliom.9999:
  sh configure --prefix /root/.opam/system --mandir /root/.opam/system/man --docdir /root/.opam/system/lib/eliom/doc --datadir /root/.opam/system/lib/eliom/share
  make
  make install
[ERROR] The compilation of eliom.9999 failed.
Removing eliom.9999.
  rm -rf /root/.opam/system/lib/eliom /root/.opam/system/man/man1/eliomc.1 /root/.opam/system/man/man1/eliomopt.1 /root/.opam/system/man/man1/eliomcp.1 /root/.opam/system/man/man1/js_of_eliom.1 /root/.opam/system/man/man1/eliomdep.1 /root/.opam/system/man/man1/eliom-distillery.1

===== ERROR while upgrading to js_of_ocaml.9999 =====
# opam-version 1.1.0
# os           linux
# command      make
# path         /root/.opam/system/build/js_of_ocaml.9999
# compiler     system (4.01.0)
# exit-code    2
# env-file     /root/.opam/system/build/js_of_ocaml.9999/js_of_ocaml-10506-335865.env
# stdout-file  /root/.opam/system/build/js_of_ocaml.9999/js_of_ocaml-10506-335865.out
# stderr-file  /root/.opam/system/build/js_of_ocaml.9999/js_of_ocaml-10506-335865.err
### stdout ###
# ...[truncated]
# make[1]: Leaving directory `/root/.opam/system/build/js_of_ocaml.9999/doc'
# make -C examples
# make[1]: Entering directory `/root/.opam/system/build/js_of_ocaml.9999/examples'
# make -C boulderdash/
# make[2]: Entering directory `/root/.opam/system/build/js_of_ocaml.9999/examples/boulderdash'
# ocamlfind ocamlc -package lwt -pp "camlp4o ../../lib/syntax/pa_js.cmo" -I ../../lib -g -c boulderdash.ml
# ocamlfind ocamlc -package lwt -pp "camlp4o ../../lib/syntax/pa_js.cmo" -I ../../lib -g -linkpkg -o boulderdash.byte js_of_ocaml.cma boulderdash.cmo
# ../../compiler/js_of_ocaml -noruntime ../../runtime/runtime.js boulderdash.byte 
# make[2]: Leaving directory `/root/.opam/system/build/js_of_ocaml.9999/examples/boulderdash'
# make[1]: Leaving directory `/root/.opam/system/build/js_of_ocaml.9999/examples'
### stderr ###
# Warning: you are using the standard library and/or the %inline keyword. We
# recommend switching on --infer in order to avoid obscure type error messages.
# Warning: 12 states have shift/reduce conflicts.
# Warning: 47 shift/reduce conflicts were arbitrarily resolved.
# Warning: Class or class type event not found
# Fatal error: exception Failure("int_of_string")
# make[2]: *** [boulderdash.js] Fel 2
# make[1]: *** [boulderdash/] Fel 2
# make: *** [examples] Fel 2

The former state can be restored with opam switch import -f "/root/.opam/system/backup/state-20140004135256.export"
'opam install eliom' failed.

I think this is the only patch left I have to the runtime to enable topfind functionality.

function caml_sys_file_exists (name) {
  return (caml_global_data.files && caml_global_data.files[name.toString()]) ? 1 : 
      iocaml_request_file(name);
}

where iocaml_request_file will do a XmlHttmlRequest to the server and call caml_register_file.

I wonder if we could add a non-invasive feature like;

function caml_sys_file_exists (name) {
  return (caml_global_data.files && caml_global_data.files[name.toString()]) ? 1 : 
      (caml_global_data.auto_register_file === undefined ? 0 : 
                 caml_global_data.auto_register_file(name));
}

to allow user code to implement this.

With ocaml 4.01rc1, I got the following missing primitives (when compiling ocsimore):
caml_ml_output_char
caml_sys_exit
caml_weak_blit
caml_weak_create
caml_weak_get
caml_weak_set

The two primary primitives are usual, but the next ones aren't.

While compiling with js_of_ocaml, I got this warning:

code providing caml_unwrap_value_from_string (~/.opam/4.01.0/lib/eliom/client/eliom_client.js:2) may miss dependencies: MlString

Some files have no copyright headers...

When we know a function is under-applied, we should create a local function that expects the additional arguments and performs the full application rather than going through caml_call_gen*, for instance, rewriting f(a,b) into function (x,y){f(a,b,x,y)} when the arity of f is 4.

Similarly, in case of over-application, we should split the application into multiple applications, turning for instance f(a,b,c,d) into f(a,b)(c,d) when the arity of f is 2.

.. so fake Json_something module (for parsing and printing) can be generated. It reduces the size of the generated javascript and should be faster.

type ty = { my_complex_type } deriving (Json)

could generate

module Json_ty = struct
  value to_string (a : ty) : string = Js.to_string (Json.output a);
  value from_string (s : string) : ty = Json.unsafe_input (Js.string s);
end

I was hoping I could register a file, write to it, then read it back. Something like.

let () = Js.register_file ~name:"file2" ~content:""

let fout = open_out "file2"
let () = output_string fout "line 1\n"
let () = output_string fout "line 2\n"
let () = close_out fout

let fin = open_in "file2"
let a = input_line fin
let b = input_line fin
let c = input_line fin
let () = close_in fin

This leads to an exception on open_in

val fin : in_channel = 
Exception: End_of_file.

(This stuff is running in my hacked up js_of_ocaml top level)

Hello js_of_ocaml team.

I used to have this "warning" when compiling my program with ocaml 4.00.0 :

Missing primitives:
  caml_ml_output_char

I upgraded to 4.01.0 via opam yesterday, and my program did not changed, but I now have those messages when compiling :

Missing primitives:
  caml_ml_output_char
  caml_sys_const_big_endian
  caml_sys_const_ostype_cygwin
  caml_sys_const_ostype_unix
  caml_sys_const_ostype_win32
  caml_sys_const_word_size

The problem is that caml_ml_output_char did not prevent my program from running well, but these new errors do.

Is it a known issue ? Did I forget to upgrade something ?

What should be in the next release ?

I would release the latest commits

• It will help to get more testing.
• We can easily release minor versions with bug fixes if needed.

What about PR:

• #88
• #93
• #96

I'm getting errors when using weak.js from js_of_ocaml. It seams to work with eliom's implementation (found in src/client/eliom_client.js).
The diff seams to come from this commit 87f958a.

The two implementations should be merged into js_of_ocaml.
we could add an option to js_of_ocaml compiler to link against weak.js as well.
we have to install weak.js.

I see that JavaScript's typed array have now been bound to. Are there any plan to support bigarrays by implementing them on top of typed arrays ?

see https://sympa.inria.fr/sympa/arc/ocsigen/2013-06/msg00035.html

If a .cmi file starts with a capital letter then it will not be found by js_of_ocaml.

I guess it's a bit unusual but Camlp4 does capitalize it's interface files. The ocaml toploop will search for interface files with both filenames.

The following change to parse_bytecode.ml seems to fix the problem.

          Tbl.iter
            (fun id num ->
               if id.Ident.flags = 1 then begin
                 let name = String.uncapitalize id.Ident.name ^ ".cmi" in
                 let file =
                   try
                     Util.find_in_paths paths name
                   with Not_found -> begin
                     let name = String.capitalize id.Ident.name ^ ".cmi" in
                     try
                       Util.find_in_paths paths name
                     with Not_found ->
                      Format.eprintf "%s: interface file '%s' not found@."
                        Sys.argv.(0) name;
                     exit 1
                   end
                 in
                 let s = Util.read_file file in
                 fs := (Pc (IString name),Pc (IString s)) :: !fs
               end) symb.num_tbl;

We should propagate more debugging information during the compilation process. In particular, there is currently no information attached to variable declarations.

The mapping is not very precise at the moment: a piece of code without debugging information will be mapped to the last piece of code with debugging information above.

Possible improvements:

• Quite often, trampolines are added when there is in fact no mutually recursive tail-calls (because a let rec was used when it was not in fact necessary). This could be improved by having Generate.group_closures perform a topological sort of the functions.
• Some wrappers are unused. We could detect that by counting how many times the corresponding function is called in tail position.