[ES6] Make distinction between * method and * operator #1229

avdg · 2016-07-22T10:25:49Z

No description provided.

rvanvelzen · 2016-07-22T11:28:57Z

This supersedes #1227, right?

avdg · 2016-07-22T11:44:22Z

Yes

kzc · 2016-07-22T12:33:26Z

Run with PR #1229:

$ cat percent.js 
        var foo = {
            "%": "foo",
            get "%"() {
                return "bar";
            },
            *"%"() {
                return "foobar";
            }
        }
        class bar {
            get "%"() {
                return "bar"
            }
            *"%"() {
                return "foobar";
            }
        }

$ bin/uglifyjs percent.js -b
var foo = {
    "%": "foo",
    get %() {
        return "bar";
    },
    *%() {
        return "foobar";
    }
};

class bar {
    get %() {
        return "bar";
    }
    *%() {
        return "foobar";
    }
}

$ bin/uglifyjs percent.js -b | bin/uglifyjs -b
Parse error at -:3,8
SyntaxError: Unexpected token: operator (%)

Edit:

furthermore:

$ node percent.js 
$ echo $?
0

$ bin/uglifyjs percent.js -b | node
[stdin]:3
    get %() {
        ^
SyntaxError: Unexpected token %

avdg · 2016-07-22T12:36:50Z

Errors too soon

avdg · 2016-07-22T12:37:34Z

Better to add an identifier check

kzc · 2016-07-22T12:42:09Z

I made that example because the PR was special casing "*" and suspected the logic would break for getters/setters with operator-like names.

avdg · 2016-07-22T12:43:29Z

Seems to be a recent breakage

avdg · 2016-07-22T12:43:52Z

Actually we shouldn't need this very special case at all (too many cases that ends up being invalid identifiers)

kzc · 2016-07-22T12:49:00Z

When you fix it, could you add test cases for % and few other operator-like names, as well as a setter/getter name with a space in it?

For bonus points, test setters/getters using some unprintable unicode names like "\x00\x01" with ascii_only=1 output mode.

avdg · 2016-07-22T12:50:31Z

This issue needs lots of test cases anyway :-)

avdg · 2016-07-22T12:57:37Z

Basically, all the logic we need was there in ObjectKeyVal

kzc · 2016-07-22T13:01:09Z

@avdg @rvanvelzen Here's an idea for test/run-tests.js - for all passing test/compress/*.js tests, the successful tests' JS output should be (re)parsed and output with the same test compress/mangle/output options to verify that uglify can simply (re)parse and (re)output its own output. It would catch inconsistancies between output and parsing. What do you think?

Edit: the re-parsed output need not necessarily match the original expected test output, which is fine - just as long as it it doesn't error out.

kzc · 2016-07-22T13:21:31Z

test/compress/object.js

+                return "bar";
+            },
+            *"%"() {
+                return "foobar";


You don't need any return statements here at all for the purpose of this test to make the expect_exact string shorter. Or you can just have 1; 2; 3; etc, if you want to easily visually cross reference the specific failing part of the test.

I'll replace them with numbers

avdg · 2016-07-22T13:22:58Z

Well, we shouldn't care about ie8 with method naming, but we still have getters that have modified behaviour and don't support ie8.

avdg · 2016-07-22T13:23:36Z

Failing test is due to timeout on node 0.12

kzc · 2016-07-22T13:26:28Z

  1) bin/uglifyjs should produce a functional build when using --self:
     Error: timeout of 5000ms exceeded. Ensure the done() callback is being called in this test.

Please increase the timeout of this specific mocha test to 15000 to compensate for occasionally slow test machines.

kzc · 2016-07-22T13:27:25Z

Just add the increased timeout for this unrelated test failure to this PR - don't bother making a new PR.

avdg · 2016-07-22T13:40:13Z

Done, any more test cases?

kzc · 2016-07-22T14:53:20Z

PR #1230 exposed 4 test failures on the harmony branch. Please test with it.

kzc · 2016-07-22T14:58:20Z

PR #1230 would have caught the % getter name error before it was corrected.

avdg · 2016-07-22T15:02:23Z

2 errors related to surrogated identifiers
2 errors related to export (which I don't work on yet)

kzc · 2016-07-22T15:07:00Z

Yeah, they are unrelated issues to this PR. But test was inspired by error in this PR.

kzc · 2016-07-22T16:45:11Z

property_with_operator_value and property_with_unprintable were only testing getters. To exercise the new function _print_getter_setter could you put a couple of setters in those tests?

kzc · 2016-07-22T19:27:01Z

test/compress/object.js

            *"\x00\x01"() {
                return "foobar";
            }
        }
    }
-    expect_exact: 'var foo={"\\0\x01":"foo",get"\\0\x01"(){return"bar"},*"\\0\x01"(){return"foobar"}};class bar{get"\\0\x01"(){return"bar"}*"\\0\x01"(){return"foobar"}}'
+    expect_exact: 'var foo={"\\0\x01":"foo",get"\\0\x01"(){return"bar"},set"\\0\x01"(foo){save(foo)},*"\\0\x01"(){return"foobar"}};class bar{get"\\0\x01"(){return"bar"}set"\\0\x01"(foo){save(foo)}*"\\0\x01"(){return"foobar"}}'
 }


Can you make a new test that's a copy of the test property_with_unprintable except using ascii_only=true?

avdg · 2016-07-26T12:16:48Z

Uh, we do have to store these quotes no? (keep quotes option)

kzc · 2016-07-28T17:55:53Z

Uh, we do have to store these quotes no? (keep quotes option)

Probably should store the quote but not emit them by default if not required. It should respect output.option("quote_keys")

avdg · 2016-07-28T21:26:40Z

Normal properties seems fine, just the getters and setters not

avdg · 2016-07-29T01:20:54Z

Current progress: property shortcut not quoted properly yet as I do have to find out a printing strategy yet when I have more time.

avdg/UglifyJS2@generator-symbol...avdg:generator-symbol-tmp

avdg · 2016-07-29T19:25:44Z

@kzc I better squash the commits for now since the fix for the star alone seems to be important. There are other broken stuff, but they aren't regressions just yet.

Also add quotes to properties when necessary, this might be the case if the name isn't a valid identifier

avdg · 2016-07-29T19:31:47Z

done

avdg · 2016-07-29T23:01:24Z

I've added more tests on the generator-symbol-tmp branch. ~~avdg/UglifyJS2@generator-symbol...avdg:generator-symbol-tmp~~

Edit: since the merge, this should now be avdg/UglifyJS2@harmony...avdg:generator-symbol-tmp

avdg · 2016-07-30T14:51:19Z

(note on additional optional patches) I might move the concise method thing to an AST_ObjectProperty, given the getters are functions as well and doing fine there.

Edit; AST move should be done there, I still have issues with computed properties as concise method, which seems to make the scope analyzer complain a lot. Still investigating cause, but it has something to do with things like [Symbol.iterator](){/* do stuff */} where doing some stuff on Symbol makes uglify complain.

Edit2: Seems that is solved by adapting the walker, maybe this can be generalized in such a way we don't need AST_ObjectComputedKeyVal anymore.

kzc · 2016-07-30T15:37:35Z

You may just want get this one merged to address the regression and wait to add features in other PRs.

avdg · 2016-07-30T16:00:01Z

Yes yes, that's what I'm doing right now. I want to prevent any more unnecessary blockages.

Although I can go a step further by opening a separate issue/pr.

Hypercubed · 2016-08-08T02:03:51Z

Just a note, jspm often generates output like:

System.registerDynamic("npm:core-js@2.4.1.json", [], false, function() {
  return {
    "main": "index.js",
    "format": "cjs",
    "meta": {
      "*": {
        "globals": {
          "process": "process"
        }
      },
      "*.json": {
        "format": "json"
      }
    }
  };
});

This PR prevents the harmony branch from crashing at the "*" key.

rvanvelzen · 2016-08-14T19:34:42Z

Awesome, thanks!

avdg mentioned this pull request Jul 22, 2016

Safely check for operators as object property key #1227

Closed

kzc reviewed Jul 22, 2016
View reviewed changes

avdg force-pushed the generator-symbol branch from 64bb13f to 14b1282 Compare July 22, 2016 13:29

kzc mentioned this pull request Jul 22, 2016

Verify output of uglify tests is parseable #1230

Closed

kzc mentioned this pull request Jul 22, 2016

harmony: Broken parsing of objects with asterisks as keys #1228

Closed

kzc reviewed Jul 22, 2016
View reviewed changes

avdg changed the title ~~Make distinction between * method and * operator~~ [ES6] Make distinction between * method and * operator Jul 26, 2016

kzc mentioned this pull request Jul 28, 2016

Harmony support #448

Closed

kzc mentioned this pull request Jul 29, 2016

[ES6] Objects with a key "*" trigger an error #1241

Closed

julienw mentioned this pull request Jul 29, 2016

Make tests work with Karma and Chrome (Closes #60) fxbox/calendar#67

Merged

Make distinction between * method and * operator

110a1ac

Also add quotes to properties when necessary, this might be the case if the name isn't a valid identifier

avdg force-pushed the generator-symbol branch from 4bfb634 to 110a1ac Compare July 29, 2016 19:31

julienw mentioned this pull request Aug 2, 2016

Update uglify once their bug is resolved fxbox/calendar#68

Open

rvanvelzen merged commit 110a1ac into mishoo:harmony Aug 14, 2016

avdg deleted the generator-symbol branch August 15, 2016 18:30

[ES6] Make distinction between * method and * operator #1229

[ES6] Make distinction between * method and * operator #1229

Conversation

avdg commented Jul 22, 2016

rvanvelzen commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016 • edited Loading

avdg commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016

avdg commented Jul 22, 2016

avdg commented Jul 22, 2016 • edited Loading

kzc commented Jul 22, 2016

avdg commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016 • edited Loading

kzc Jul 22, 2016

Choose a reason for hiding this comment

avdg Jul 22, 2016

Choose a reason for hiding this comment

avdg commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016

kzc commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016

kzc commented Jul 22, 2016

avdg commented Jul 22, 2016

kzc commented Jul 22, 2016

kzc commented Jul 22, 2016

kzc Jul 22, 2016

Choose a reason for hiding this comment

avdg Jul 26, 2016

Choose a reason for hiding this comment

avdg commented Jul 26, 2016 • edited Loading

kzc commented Jul 28, 2016

avdg commented Jul 28, 2016

avdg commented Jul 29, 2016

avdg commented Jul 29, 2016

avdg commented Jul 29, 2016

avdg commented Jul 29, 2016 • edited Loading

avdg commented Jul 30, 2016 • edited Loading

kzc commented Jul 30, 2016

avdg commented Jul 30, 2016

Hypercubed commented Aug 8, 2016

rvanvelzen commented Aug 14, 2016

kzc commented Jul 22, 2016 •

edited

Loading

avdg commented Jul 22, 2016 •

edited

Loading

kzc commented Jul 22, 2016 •

edited

Loading

avdg commented Jul 26, 2016 •

edited

Loading

avdg commented Jul 29, 2016 •

edited

Loading

avdg commented Jul 30, 2016 •

edited

Loading