一种方法是有一个单独的服务，可以用作这些参数的“容器”，其中它们是公共数据成员。

这里有一个解决方案(基于Marcin Wyszynski的建议)，当你想要传递一个值到你的控制器，但你没有显式地在你的html中声明控制器(ng-init似乎需要)-例如，如果你用ng-view渲染你的模板，并通过routeProvider为相应的路由声明每个控制器。

JS

messageboard.directive('currentuser', ['CurrentUser', function(CurrentUser) {
  return function(scope, element, attrs) {
    CurrentUser.name = attrs.name;
  };
}]);

html

<div ng-app="app">
  <div class="view-container">
    <div ng-view currentuser name="testusername" class="view-frame animate-view"></div>
  </div>
</div>

在这个解决方案中，CurrentUser是一个服务，它可以被注入到任何控制器中，此时.name属性可用。

两个笔记:

a problem I've encountered is that .name gets set after the controller loads, so as a workaround I have a short timeout before rendering username on the controller's scope. Is there a neat way of waiting until .name has been set on the service? this feels like a very easy way to get a current user into your Angular App with all the authentication kept outside Angular. You could have a before_filter to prevent non-logged in users getting to the html where your Angular app is bootstrapped in, and within that html you could just interpolate the logged in user's name and even their ID if you wanted to interact with the user's details via http requests from your Angular app. You could allow non-logged in users to use the Angular App with a default 'guest user'. Any advice on why this approach would be bad would be welcome - it feels too easy to be sensible!)

我对此很晚了，我不知道这是否是一个好主意，但你可以在控制器函数中包括$attrs注入，允许控制器使用元素上提供的“参数”进行初始化，例如。

app.controller('modelController', function($scope, $attrs) {
    if (!$attrs.model) throw new Error("No model for modelController");

    // Initialize $scope using the value of the model attribute, e.g.,
    $scope.url = "http://example.com/fetch?model="+$attrs.model;
})

<div ng-controller="modelController" model="foobar">
  <a href="{{url}}">Click here</a>
</div>

同样，不知道这是否是一个好主意，但它似乎有效，是另一种选择。

视图不应该指定配置

In Angular, the template should never dictate configuration, which is inherently what people desire when they want to pass arguments to controllers from a template file. This becomes a slippery slope. If config settings are hard-coded in templates (such as by a directive or controller argument attribute), you can no longer re-use that template for anything but that single use. Soon you'll want to re-use that template, but with different config and now in order to do so you'll either be pre-processing the templates to inject variables before it gets passed to angular or using massive directives to spit out giant blocks of HTML so you re-use all of the controller HTML except for the wrapper div and it's arguments. For small projects it's no big deal. For something big (what angular excels at), it gets ugly quick.

替代方案:模块

This type of configuration is what modules were designed to handle. In many angular tutorials people have a single module for their entire application, but really the system is designed and fully supports many small modules each which wrap small pieces of the total application. Ideally, controllers, modules etc would be declared in separate files and stitched together in specific re-usable chunks. When your application is designed this way, you get a lot of re-use in addition to easy controller arguments.

The example below has 2 modules, re-using the same controller, but each with their own config settings. That config settings are passed in via dependency injection using module.value. This adheres to the angular way because we have the following: constructor dependency injection, reusable controller code, reusable controller templates (the controller div could easily be included with ng-include), easily unit-testable system without HTML, and lastly re-usable modules as the vehicle for stitching the pieces together.

这里有一个例子:

<!-- index.html -->
<div id="module1">
    <div ng-controller="MyCtrl">
        <div>{{foo}}</div>
    </div>
</div>
<div id="module2">
    <div ng-controller="MyCtrl">
        <div>{{foo}}</div>
    </div>
</div>
<script>
    // part of this template, or a JS file designed to be used with this template
    angular.element(document).ready(function() {
        angular.bootstrap(document.getElementById("module1"), ["module1"]);
        angular.bootstrap(document.getElementById("module2"), ["module2"]);
    });
</script>

<!-- scripts which will likely in be in their seperate files -->
<script>
    // MyCtrl.js
    var MyCtrl = function($scope, foo) {
    $scope.foo = foo;
    }

    MyCtrl.$inject = ["$scope", "foo"];

    // Module1.js
    var module1 = angular.module('module1', []);
    module1.value("foo", "fooValue1");
    module1.controller("MyCtrl", MyCtrl);

    // Module2.js file
    var module2 = angular.module('module2', []);
    module2.value("foo", "fooValue2");
    module2.controller("MyCtrl", MyCtrl);
</script>

查看它的实际运行情况:jsFiddle。

一种方法是有一个单独的服务，可以用作这些参数的“容器”，其中它们是公共数据成员。

如果ng-init不是用来将对象传递到$scope，你可以编写自己的指令。这就是我得到的结果:

http://jsfiddle.net/goliney/89bLj/

Javasript：

var app = angular.module('myApp', []);
app.directive('initData', function($parse) {
    return function(scope, element, attrs) {
        //modify scope
        var model = $parse(attrs.initData);
        model(scope);
    };
});

function Ctrl1($scope) {
    //should be defined
    $scope.inputdata = {foo:"east", bar:"west"};
}

Html:

<div ng-controller="Ctrl1">
    <div init-data="inputdata.foo=123; inputdata.bar=321"></div>
</div>

但是我的方法只能修改已经在控制器中定义的对象。