… obj sugar (reasonml#2152)

…asonml#2151)

Fixes reasonml#2150

…x child (reasonml#2159)

Turns out caret behaves differently for versions under 1.0.0. `^0.0.3` is just like specifying `0.0.3` The More You Know.

Not putting the correct location on the Pexp_ident expression
(the component name), results in type errors on line 0.

<div foo=bar />

div -> Pexp_ident, needs a correct location in order to have
correct type errors

Needed for Merlin's jump to location and type info in editor.

* Minimal changes to compile with OCaml 4.07

* Add Travis testing for 4.07.0+rc1

* Add CircleCI testing for 4.07.0

* Add missing m4 dependency to CircleCI

* Pin utop to its development repository

4.07 support has been merged into its master branch.

* Add 4.07.0-specific variant formatting test

Thank you @IwanKaramazow!

* Use final 4.07.0 release in Travis CI

* rebase on master

* specific test for 4.07

* FIX: Refmt link typo in Readme

* Update README.md

* Use updated Menhir in Circle CI

* Update config.yml

* Update esy workflow on ci

* Fix travis ci

fixes reasonml#2190

The `consolidateSeparator` function was recursing on `Whitespace`
nodes, however `preOrderWalk` would already do that for us. Therefore,
we were calling this costly function in a hot path many more times than
necessary, effectively making it take exponential time (I think).

)

Before:
reactClass
->setNavigationOptions(
    NavigationOptions.t(~title="Title", ~gesturesEnabled=false, ()),
  );

After:
reactClass->setNavigationOptions(
  NavigationOptions.t(~title="Title", ~gesturesEnabled=false, ()),
);

* keep `let not = ...` and other OCaml <-> Reason swappable attributes

fixes reasonml#2169

* Make it work for OCaml <= 4.04

also fixes reasonml#2134

* Make it work for exprs and sigs too, add tests

* Fix label punning + add tests

* revert exposing escape_stars_slashes

* Clean up / Add identifier_or_original_longident

* Now without all the unnecessary code

* Fix precedence issue with unary ops and labelled args

* Remove special tokens and delay parsing labelled unaries until recovery

* Don't install the oprintTest executable

* Build testOprint prior to trying to use it

Otherwise output from build warnings breaks the tests.

reasonml#2205)

args.

There's a bug in the grammar: the parser tries to parse tuples with named
arguments, resulting in an exception crashing refmt.

```reason
let testCallNamedArgs = (foo: (~a: int, ~b: int) => int), a, b) => foo;

(~a: int, ~b: int) => int /* interpreted as a Ptyp_tuple */
```

To unbreak master refmt, this patch forces parens around the Ptyp_arrow
constraint. With parens the correct ast is parsed.
```reason
let testCallNamedArgs = (foo: ((int, ~b: int) => int), a, b) => foo;
```

The parsing without parens will be addressed asap in a different PR.

* Supreme printing of anonymous functions as props in jsx.

BEFORE:
<div
  onClick={
    event => {
      Js.log(event);
      handleChange(event);
    }
  }
/>;

AFTER:
<div
  onClick={event => {             ----> inline the arguments
    Js.log(event);
    handleChange(event);
  }}                             ----> brace hugging
/>;

* Add extra tests with named args and take care of constraint expressions

* Format single arguments with a constraint on the Pexp_fun correct.

* Add more tests

* Drop paren requirement Ptyp_arrow pattern constraints

Fixes reasonml#2141
Fixes reasonml#1957

* parses without parens
* prints without parens
* small cleanup of the menhir grammar

Before:
```reason
let testCallNamedArgs = (foo: ((~a: int, ~b: int) => int), a, b) => foo(~a, ~b);
```

After:
```reason
let testCallNamedArgs = (foo: (~a: int, ~b: int) => int, a, b) => foo(~a, ~b);
```

* regenerate empty error message table

* Consistent with Prettier
* Wastes less whitespace

Before:
```reason
<div>
  {
    switch (color) {
    | Black => ReasonReact.string("black")
    | Red => ReasonReact.string("red")
    }
  }
</div>;
```

```reason
<div>
  {switch (color) {
   | Black => ReasonReact.string("black")
   | Red => ReasonReact.string("red")
   }}
</div>;
```

The referenced directly doesn't exist anymore after b9823e5

* Polished error message for reserved keywords in records fields

fixes reasonml#1643

* Alternative approach

* inline Pexp_apply where possible with brace hugging
* improve formatting of spread children with Pexp_apply & Pexp_fun

* Set up error messages properly in non-recoverable mode

* Unfortunately have to let go of these

* simple heuristic to recover keyword errors

* Upgrade to Dune 1.4 and add Menhir `--infer`

* install dune, not jbuilder

* try something

Also add docs about releasing manually to OPAM

* Cleanup esy files for esy 0.3.X

* code review

Summary: This puts an overly rigid constraint on Menhir (for Windows,
only certain versions are correctly patched etc) and we weren't even
specifying the correct constraint on menhir introduced by an `--infer`
flag.

Test Plan:

Reviewers:

CC:

* WIP: Recover from parsing two postfix ops

fixes reasonml#126

* Tests and refactor

…easonml#2255)

compiler

…l#2051)

* Syntax sugar `["literal_string"]` for SHARPOPs `##` and `#=`

* rebase on master after reasonml#2050, parse `#=` with higher precedence, parens
unnecessary on the right side

* array, string and bigarray get also get `#=` -> `=`

* rebase on fast pipe fix

* small refactor

* Fix printer after rebase (caused by earlier refactor)

* fix tests after rebase

* Fixes after rebase

Opam-repository requires it.

* Preserve function body braces

This is an alternative to reasonml#1826 implementing just the brace preservation
code, to make it easier to be reviewed.

I think this is now in a better place to get into the codebase given
that we have `Reason_attributes` outside of `Reason_pprint_ast` and that
we now remove our own attributes before printing to OCaml.

I think we can probably refactor this code further in future PRs, as
well as add other stylistic preservation changes.

* Rebase and fix

* address review comments

* Adding so many optional laballed to `unparseExprRecurse` can't scale

* Add support for single line comments of the form //

* Fix edge case where // inside a sequence wrapped in braces didn't break.

```reason
let x = {
  //
};
```

If the { }-sequence didn't break we would get parse error:

```reason
let x = { // };
```

…onml#1990)

* interleave whitespace between recursive let and module bindings

* Inline braces surrounding a Pexp_match as jsx child.

* Consistent with Prettier
* Wastes less whitespace

Before:
```reason
<div>
  {
    switch (color) {
    | Black => ReasonReact.string("black")
    | Red => ReasonReact.string("red")
    }
  }
</div>;
```

```reason
<div>
  {switch (color) {
   | Black => ReasonReact.string("black")
   | Red => ReasonReact.string("red")
   }}
</div>;
```

* Cleanup unused code

* fix esy build

…easonml#2051)" (reasonml#2267)

This reverts commit 2d2a5da.

)

* Install pkg-config explicitly

* Bring in ncurses library

* Remove npm install / npm run build while we're getting CI going

* Use open_out/open_out_bin conditionally depending on the print type and the platform

* Format changes

* Implement EOL detection and incorporate in refmt

* Remove temporary test cases

* Format & remove extra pint

* Fix formatting

* Remove unused functions in refmt_impl

* Fix casing of 'str' in dune file

* Detect CRLF w/o regex

* Remove Str from dune file

* Add CRLF test

* Add EOL conversion function, hook up to formatter

* Fix formatting of eol_convert module

* Add .gitattribute setting for actual_output, too

* Add CRLF tests to idempotent tests

* Update .gitattributes

* Add idempotent test category

* Add test output folder

…en hugging (reasonml#2288)

This is an alternative to reasonml#2289 that also renames `literalAttrs` to
`stylisticAttrs` as request in reasonml#2223's code review

* format closing brace in jsx with funAppl unit sugar

* Add tests

I noticed this missing article while reading the document.

* rename fast pipe to pipe first

* further pipe first cleanup

* Check in esy.lock files

* checksum the index.json

* Fix test command w esy x

* Update ternary formatting

* update typeCheckedTests for ternary formatting changes

* add test cases showing remaining ternary formatting issue

* Initial CI commit

* Remove unnecessary comment

* Remove old lockfile

* Add an OPAM config

* Fix up names

* Use apt install opam

* Check ocaml compiler version

* Add a project for opam+ocaml@4.06

* Fix version command

* Add trailing newline to reason.opam

* Log opam version

* Manually set switch for OCaml version

* Try installing OPAM via shell script

* Try downloading OPAM

* Add sudo

* Add flag to follow redirects

* Try running OPAM on Mac instead

* Add build for ocaml@4.07.1 via OPAM

* Remove ocamlv ersion check

* Add eval command prior to running tests

* Merge esy.json

* Relocate to existing azure-pipelines file

* Update refmt execution command for esy builds

* Fix esy build

* Port over improved caching

* Revert changes to lock / esy.json

* README: Add build badge

Fixes reasonml#2315