[BUGFIX release-1-13] Fix string classify normalizer

[skeate]

Resolves #11326. If a piece of a name starts with dash (-), it will be changed into an underscore (_).

[rwjblue]

We can't use Array.prototype.map in release-1-13 (because it must support ES3 browsers), can you swap this to a for loop instead to allow this to be pulled into those branches?

[skeate]

Changed.

[rwjblue]

Can you also add tests for these scenarios:

• classify('_FooBar') === '_FooBar'
• classify('_Foo_Bar') === '_FooBar'
• classify('_Foo-Bar') === '_FooBar'
• classify('_foo/_bar') === '_Foo/_Bar'
• classify('-foo/_bar') === '_Foo/_Bar'
• classify('-foo/-bar') === '_Foo/_Bar'

[skeate]

Any issue with redoing all these tests as a sort of macro function? i.e. they're all like

QUnit.test('classify normal string', function() {
  deepEqual(classify('my favorite items'), 'MyFavoriteItems');
  if (Ember.EXTEND_PROTOTYPES) {
    deepEqual('my favorite items'.classify(), 'MyFavoriteItems');
  }
});

-- could write one function

function test(description, given, expected) {
  QUnit.test(description, function() {
    deepEqual(classify(given), expected);
    if (Ember.EXTEND_PROTOTYPES) {
      deepEqual(given.classify(), expected);
    }
  });
}

and call it for all test cases.

Would be easier to write them and easier to see all the different test cases.

[Edit: and if so, should I make that a separate commit or amend it?]

[rwjblue]

Ya, one commit with the fix and new tests (the old way), and a second commit that refactors to a saner structure sounds great.

[rwjblue]

Only the first commit would be [BUGFIX release-1-13] though. The other would be [CLEANUP canary] I think.

[skeate]

alright, have it done in a min

[rwjblue]

Looks great, thank you!