asp.net mvc源码分析EditorFor

在我们的mvc日常开发会经常遇到什么LabelFor、EditorFor、Editor等等，这个扩展方法有很多是相似的。这里我们以EditorFor来说说吧，我觉得这个相对要复杂一点。

首先我们来看看EditorFor的定义：

 public static MvcHtmlString EditorFor<TModel, TValue>(this HtmlHelper<TModel> html, Expression<Func<TModel, TValue>> expression, object additionalViewData) {
            return TemplateHelpers.TemplateFor(html, expression, null /* templateName */, null /* htmlFieldName */, DataBoundControlMode.Edit, additionalViewData);
        }

虽然EditorFor有很多定义，但是实际上都是调用 TemplateHelpers.TemplateFor方法。

  internal static MvcHtmlString TemplateFor<TContainer, TValue>(this HtmlHelper<TContainer> html, Expression<Func<TContainer, TValue>> expression,
                                                                      string templateName, string htmlFieldName, DataBoundControlMode mode,
                                                                      object additionalViewData) {
            return MvcHtmlString.Create(TemplateFor(html, expression, templateName, htmlFieldName, mode, additionalViewData, TemplateHelper));
        }
现在大家应该知道TemplateFor方法的主要参数都有哪些了吧，但是在实际开发中我们的templateName、htmlFieldName、additionalViewData通常都是null，mode是DataBoundControlMode.Edit
我们还是举一个例子来说说吧：

 public class UserInfo
    {
        [StringLength(100, MinimumLength = 10)]
        [Required]
        public string UserName { set; get; }

    }

  @Html.EditorFor(model => model.UserName)

这个代码是不是很简单。

现在我们来看看TemplateFor的实现

      return templateHelper(html,
                                  ModelMetadata.FromLambdaExpression(expression, html.ViewData),
                                  htmlFieldName ?? ExpressionHelper.GetExpressionText(expression),
                                  templateName,
                                  mode,
                                  additionalViewData);

首先我们来看看  ModelMetadata.FromLambdaExpression(expression, html.ViewData)这句是如何获取ModelMetadata的，具体实现如下：

public static ModelMetadata FromLambdaExpression<TParameter, TValue>(Expression<Func<TParameter, TValue>> expression,
                                                                             ViewDataDictionary<TParameter> viewData) {
            if (expression == null) {
                throw new ArgumentNullException("expression");
            }
            if (viewData == null) {
                throw new ArgumentNullException("viewData");
            }

            string propertyName = null;
            Type containerType = null;
            bool legalExpression = false;

            // Need to verify the expression is valid; it needs to at least end in something
            // that we can convert to a meaningful string for model binding purposes

            switch (expression.Body.NodeType) {
                // ArrayIndex always means a single-dimensional indexer; multi-dimensional indexer is a method call to Get()
                case ExpressionType.ArrayIndex:
                    legalExpression = true;
                    break;

                // Only legal method call is a single argument indexer/DefaultMember call
                case ExpressionType.Call:
                    legalExpression = ExpressionHelper.IsSingleArgumentIndexer(expression.Body);
                    break;

                // Property/field access is always legal
                case ExpressionType.MemberAccess:
                    MemberExpression memberExpression = (MemberExpression)expression.Body;
                    propertyName = memberExpression.Member is PropertyInfo ? memberExpression.Member.Name : null;
                    containerType = memberExpression.Expression.Type;
                    legalExpression = true;
                    break;

                // Parameter expression means "model => model", so we delegate to FromModel
                case ExpressionType.Parameter:
                    return FromModel(viewData);
            }

            if (!legalExpression) {
                throw new InvalidOperationException(MvcResources.TemplateHelpers_TemplateLimitations);
            }

            TParameter container = viewData.Model;
            Func<object> modelAccessor = () => {
                try {
                    return CachedExpressionCompiler.Process(expression)(container);
                }
                catch (NullReferenceException) {
                    return null;
                }
            };

            return GetMetadataFromProvider(modelAccessor, typeof(TValue), propertyName, containerType);
        }

 如果我们调用的是Editor那么之力调用就是FromStringExpression方法而不是FromLambdaExpression方法，这2个方法相差不大。我们还是来看看FromLambdaExpression这个方法吧：

看了这张图 那么 return GetMetadataFromProvider(modelAccessor, typeof(TValue), propertyName, containerType);这句代码里面的参数相信大家都应该是到了吧。

propertyName=“UserName” containerType=MvcApp.Controllers.UserInfo，modelAccessor就是创建一个实例，实例的创建是通过 model => model.UserName这句。至于GetMetadataFromProvider这个方法就没什么好讲的了，前面的文章已经讲过了，它实际是创建 了一个DataAnnotationsModelMetadata实例。

至于ExpressionHelper.GetExpressionText(expression)这句说白了默认就是返回一个属性名称，具体实现：

  public static string GetExpressionText(LambdaExpression expression) {
            // Split apart the expression string for property/field accessors to create its name
            Stack<string> nameParts = new Stack<string>();
            Expression part = expression.Body;

            while (part != null) {
                if (part.NodeType == ExpressionType.Call) {
                    MethodCallExpression methodExpression = (MethodCallExpression)part;

                    if (!IsSingleArgumentIndexer(methodExpression)) {
                        break;
                    }

                    nameParts.Push(
                        GetIndexerInvocation(
                            methodExpression.Arguments.Single(),
                            expression.Parameters.ToArray()
                        )
                    );

                    part = methodExpression.Object;
                }
                else if (part.NodeType == ExpressionType.ArrayIndex) {
                    BinaryExpression binaryExpression = (BinaryExpression)part;

                    nameParts.Push(
                        GetIndexerInvocation(
                            binaryExpression.Right,
                            expression.Parameters.ToArray()
                        )
                    );

                    part = binaryExpression.Left;
                }
                else if (part.NodeType == ExpressionType.MemberAccess) {
                    MemberExpression memberExpressionPart = (MemberExpression)part;
                    nameParts.Push("." + memberExpressionPart.Member.Name);
                    part = memberExpressionPart.Expression;
                }
                else if (part.NodeType == ExpressionType.Parameter) {
                    // Dev10 Bug #907611
                    // When the expression is parameter based (m => m.Something...), we'll push an empty
                    // string onto the stack and stop evaluating. The extra empty string makes sure that
                    // we don't accidentally cut off too much of m => m.Model.
                    nameParts.Push(String.Empty);
                    part = null;
                }
                else {
                    break;
                }
            }

            // If it starts with "model", then strip that away
            if (nameParts.Count > 0 && String.Equals(nameParts.Peek(), ".model", StringComparison.OrdinalIgnoreCase)) {
                nameParts.Pop();
            }

            if (nameParts.Count > 0) {
                return nameParts.Aggregate((left, right) => left + right).TrimStart('.');
            }

            return String.Empty;
        }

 这里的循环执行两次，第一次是执行

 else if (part.NodeType == ExpressionType.MemberAccess) {
                    MemberExpression memberExpressionPart = (MemberExpression)part;
                    nameParts.Push("." + memberExpressionPart.Member.Name);
                    part = memberExpressionPart.Expression;
                }

第二次执行

    else if (part.NodeType == ExpressionType.Parameter) {
                    // Dev10 Bug #907611
                    // When the expression is parameter based (m => m.Something...), we'll push an empty
                    // string onto the stack and stop evaluating. The extra empty string makes sure that
                    // we don't accidentally cut off too much of m => m.Model.
                    nameParts.Push(String.Empty);
                    part = null;
                }

当然这个方法默认的返回结果这里就是UserName了,它默认就是生成html是的id和name属性的值。

现在我们再来看看TemplateHelper方法了

 internal static string TemplateHelper(HtmlHelper html, ModelMetadata metadata, string htmlFieldName, string templateName, DataBoundControlMode mode, object additionalViewData, ExecuteTemplateDelegate executeTemplate) {
            // TODO: Convert Editor into Display if model.IsReadOnly is true? Need to be careful about this because
            // the Model property on the ViewPage/ViewUserControl is get-only, so the type descriptor automatically
            // decorates it with a [ReadOnly] attribute...

            if (metadata.ConvertEmptyStringToNull && String.Empty.Equals(metadata.Model)) {
                metadata.Model = null;
            }

            object formattedModelValue = metadata.Model;
            if (metadata.Model == null && mode == DataBoundControlMode.ReadOnly) {
                formattedModelValue = metadata.NullDisplayText;
            }

            string formatString = mode == DataBoundControlMode.ReadOnly ? metadata.DisplayFormatString : metadata.EditFormatString;
            if (metadata.Model != null && !String.IsNullOrEmpty(formatString)) {
                formattedModelValue = String.Format(CultureInfo.CurrentCulture, formatString, metadata.Model);
            }

            // Normally this shouldn't happen, unless someone writes their own custom Object templates which
            // don't check to make sure that the object hasn't already been displayed
            object visitedObjectsKey = metadata.Model ?? metadata.RealModelType;
            if (html.ViewDataContainer.ViewData.TemplateInfo.VisitedObjects.Contains(visitedObjectsKey)) {    // DDB #224750
                return String.Empty;
            }

            ViewDataDictionary viewData = new ViewDataDictionary(html.ViewDataContainer.ViewData) {
                Model = metadata.Model,
                ModelMetadata = metadata,
                TemplateInfo = new TemplateInfo {
                    FormattedModelValue = formattedModelValue,
                    HtmlFieldPrefix = html.ViewContext.ViewData.TemplateInfo.GetFullHtmlFieldName(htmlFieldName),
                    VisitedObjects = new HashSet<object>(html.ViewContext.ViewData.TemplateInfo.VisitedObjects),    // DDB #224750
                }
            };

            if (additionalViewData != null) {
                foreach (KeyValuePair<string, object> kvp in new RouteValueDictionary(additionalViewData)) {
                    viewData[kvp.Key] = kvp.Value;
                }
            }

            viewData.TemplateInfo.VisitedObjects.Add(visitedObjectsKey);    // DDB #224750

            return executeTemplate(html, viewData, templateName, mode, GetViewNames, GetDefaultActions);
        }

 这个方法其实也很简单，获取当前model的值，以及呈现html的format格式，最后这里从新创建了一个ViewDataDictionary实例 viewData，并且把参数中的additionalViewData也合并到这个viewData中来，把当前的值 （visitedObjectsKey也就是最后呈现给textbox的value）给添加到viewData的VisitedObjects属性中。最 后再调用
  return executeTemplate(html, viewData, templateName, mode, GetViewNames, GetDefaultActions);方法。

那么现在我们应该看看ExecuteTemplate方法了：

   internal static string ExecuteTemplate(HtmlHelper html, ViewDataDictionary viewData, string templateName, DataBoundControlMode mode, GetViewNamesDelegate getViewNames, GetDefaultActionsDelegate getDefaultActions) {
            Dictionary<string, ActionCacheItem> actionCache = GetActionCache(html);
            Dictionary<string, Func<HtmlHelper, string>> defaultActions = getDefaultActions(mode);
            string modeViewPath = modeViewPaths[mode];

            foreach (string viewName in getViewNames(viewData.ModelMetadata, templateName, viewData.ModelMetadata.TemplateHint, viewData.ModelMetadata.DataTypeName)) {
                string fullViewName = modeViewPath + "/" + viewName;
                ActionCacheItem cacheItem;

                if (actionCache.TryGetValue(fullViewName, out cacheItem)) {
                    if (cacheItem != null) {
                        return cacheItem.Execute(html, viewData);
                    }
                }
                else {
                    ViewEngineResult viewEngineResult = ViewEngines.Engines.FindPartialView(html.ViewContext, fullViewName);
                    if (viewEngineResult.View != null) {
                        actionCache[fullViewName] = new ActionCacheViewItem { ViewName = fullViewName };

                        using (StringWriter writer = new StringWriter(CultureInfo.InvariantCulture)) {
                            viewEngineResult.View.Render(new ViewContext(html.ViewContext, viewEngineResult.View, viewData, html.ViewContext.TempData, writer), writer);
                            return writer.ToString();
                        }
                    }

                    Func<HtmlHelper, string> defaultAction;
                    if (defaultActions.TryGetValue(viewName, out defaultAction)) {
                        actionCache[fullViewName] = new ActionCacheCodeItem { Action = defaultAction };
                        return defaultAction(MakeHtmlHelper(html, viewData));
                    }

                    actionCache[fullViewName] = null;
                }
            }

            throw new InvalidOperationException(
                String.Format(
                    CultureInfo.CurrentCulture,
                    MvcResources.TemplateHelpers_NoTemplate,
                    viewData.ModelMetadata.RealModelType.FullName
                )
            );
        }

 这里的GetActionCache方法很简单就是从当前的context.Items中获取一个字典数据，如果没有就实例化一个然后加入到context.Items中。GetDefaultActions方法也很简单
  internal static Dictionary<string, Func<HtmlHelper, string>> GetDefaultActions(DataBoundControlMode mode) {
            return mode == DataBoundControlMode.ReadOnly ? defaultDisplayActions : defaultEditorActions;
        }

其中GetViewNames就是获取view的名称的一个方法：

 internal static IEnumerable<string> GetViewNames(ModelMetadata metadata, params string[] templateHints) {
            foreach (string templateHint in templateHints.Where(s => !String.IsNullOrEmpty(s))) {
                yield return templateHint;
            }

            // We don't want to search for Nullable<T>, we want to search for T (which should handle both T and Nullable<T>)
            Type fieldType = Nullable.GetUnderlyingType(metadata.RealModelType) ?? metadata.RealModelType;

            // TODO: Make better string names for generic types
            yield return fieldType.Name;

            if (!metadata.IsComplexType) {
                yield return "String";
            }
            else if (fieldType.IsInterface) {
                if (typeof(IEnumerable).IsAssignableFrom(fieldType)) {
                    yield return "Collection";
                }

                yield return "Object";
            }
            else {
                bool isEnumerable = typeof(IEnumerable).IsAssignableFrom(fieldType);

                while (true) {
                    fieldType = fieldType.BaseType;
                    if (fieldType == null)
                        break;

                    if (isEnumerable && fieldType == typeof(Object)) {
                        yield return "Collection";
                    }

                    yield return fieldType.Name;
                }
            }
        }

 默认返回参数顺序是templateHints中的view，其次就是根据参数 数据类型返回相应的默认view。我这里返回的是String。

那么现在我们回到ExecuteTemplate方法中来，

 string fullViewName = modeViewPath + "/" + viewName;这句就是已经找到我们的view了，如果我们先前actionCache中包含该key就直接执行该view并返回，其次通过 ViewEngines.Engines.FindPartialView来找该view，如果找到则输出该view并返回。否则调用默认的处理方式

   if (defaultActions.TryGetValue(viewName, out defaultAction)) {
                        actionCache[fullViewName] = new ActionCacheCodeItem { Action = defaultAction };
                        return defaultAction(MakeHtmlHelper(html, viewData));
                    }

这里的defaultAction对应则DefaultDisplayTemplates.StringTemplate，因为我的 viewName是String，这里的MakeHtmlHelper方法是根据当前的ViewContext和viewData从新实例化一个 HtmlHelper。

DefaultEditorTemplates.StringTemplate方法非常简单：

  return html.TextBox(String.Empty,
                                html.ViewContext.ViewData.TemplateInfo.FormattedModelValue,
                                CreateHtmlAttributes("text-box single-line")).ToHtmlString();

它里面主要是调用TextBox方法：

    public static MvcHtmlString TextBox(this HtmlHelper htmlHelper, string name, object value, IDictionary<string, object> htmlAttributes) {
            return InputHelper(htmlHelper, InputType.Text, null, name, value, (value == null) /* useViewData */, false /* isChecked */, true /* setId */, true /* isExplicitValue */, htmlAttributes);
        }

    private static MvcHtmlString InputHelper(HtmlHelper htmlHelper, InputType inputType, ModelMetadata metadata, string name, object value, bool useViewData, bool isChecked, bool setId, bool isExplicitValue, IDictionary<string, object> htmlAttributes) {
            string fullName = htmlHelper.ViewContext.ViewData.TemplateInfo.GetFullHtmlFieldName(name);
            if (String.IsNullOrEmpty(fullName)) {
                throw new ArgumentException(MvcResources.Common_NullOrEmpty, "name");
            }

            TagBuilder tagBuilder = new TagBuilder("input");
            tagBuilder.MergeAttributes(htmlAttributes);
            tagBuilder.MergeAttribute("type", HtmlHelper.GetInputTypeString(inputType));
            tagBuilder.MergeAttribute("name", fullName, true);

            string valueParameter = Convert.ToString(value, CultureInfo.CurrentCulture);
            bool usedModelState = false;

            switch (inputType) {
                case InputType.CheckBox:
                    bool? modelStateWasChecked = htmlHelper.GetModelStateValue(fullName, typeof(bool)) as bool?;
                    if (modelStateWasChecked.HasValue) {
                        isChecked = modelStateWasChecked.Value;
                        usedModelState = true;
                    }
                    goto case InputType.Radio;
                case InputType.Radio:
                    if (!usedModelState) {
                        string modelStateValue = htmlHelper.GetModelStateValue(fullName, typeof(string)) as string;
                        if (modelStateValue != null) {
                            isChecked = String.Equals(modelStateValue, valueParameter, StringComparison.Ordinal);
                            usedModelState = true;
                        }
                    }
                    if (!usedModelState && useViewData) {
                        isChecked = htmlHelper.EvalBoolean(fullName);
                    }
                    if (isChecked) {
                        tagBuilder.MergeAttribute("checked", "checked");
                    }
                    tagBuilder.MergeAttribute("value", valueParameter, isExplicitValue);
                    break;
                case InputType.Password:
                    if (value != null) {
                        tagBuilder.MergeAttribute("value", valueParameter, isExplicitValue);
                    }
                    break;
                default:
                    string attemptedValue = (string)htmlHelper.GetModelStateValue(fullName, typeof(string));
                    tagBuilder.MergeAttribute("value", attemptedValue ?? ((useViewData) ? htmlHelper.EvalString(fullName) : valueParameter), isExplicitValue);
                    break;
            }

            if (setId) {
                tagBuilder.GenerateId(fullName);
            }

            // If there are any errors for a named field, we add the css attribute.
            ModelState modelState;
            if (htmlHelper.ViewData.ModelState.TryGetValue(fullName, out modelState)) {
                if (modelState.Errors.Count > 0) {
                    tagBuilder.AddCssClass(HtmlHelper.ValidationInputCssClassName);
                }
            }

            tagBuilder.MergeAttributes(htmlHelper.GetUnobtrusiveValidationAttributes(name, metadata));

            if (inputType == InputType.CheckBox) {
                // Render an additional <input type="hidden".../> for checkboxes. This
                // addresses scenarios where unchecked checkboxes are not sent in the request.
                // Sending a hidden input makes it possible to know that the checkbox was present
                // on the page when the request was submitted.
                StringBuilder inputItemBuilder = new StringBuilder();
                inputItemBuilder.Append(tagBuilder.ToString(TagRenderMode.SelfClosing));

                TagBuilder hiddenInput = new TagBuilder("input");
                hiddenInput.MergeAttribute("type", HtmlHelper.GetInputTypeString(InputType.Hidden));
                hiddenInput.MergeAttribute("name", fullName);
                hiddenInput.MergeAttribute("value", "false");
                inputItemBuilder.Append(hiddenInput.ToString(TagRenderMode.SelfClosing));
                return MvcHtmlString.Create(inputItemBuilder.ToString());
            }

            return tagBuilder.ToMvcHtmlString(TagRenderMode.SelfClosing);
        }

这里的InputHelper是真正生成html字符串的地方。这个方法整体比较好理解，不过要注意这个方法里面有这么一句

            tagBuilder.MergeAttributes(htmlHelper.GetUnobtrusiveValidationAttributes(name, metadata));

这句就是处理对象上面的那些验证属性，效果如图：

我想大家到这里应该对EditorFor方法有个了解了吧，感觉很是复杂。