Delphi通过MSHTML实现一个HTML解析类

最近经常会模拟网页提交返回网页源码，然后获得网页中相应的元素，于是需要常常解析Html中相应的各种元素，网络是个好东西，搜索一番，就找到了好几个Delphi版本的HtmlParser的类库，试着使用了几个，发现解析起来都不完整，或多或少的回出现一些问题！于是想到了如果界面上有一个浏览器，我们可以通过WebBrowser的Document接口对网页元素进行操作，很是方便！但是模拟网页提交，界面上是不一定要出现WebBrowser的，肯定有办法，不通过WebBrowser就直接解析HTML的，那便是我不要WebBrowser这个外壳，只要他里面的Document文档接口对象就能实现对Html的解析了，查找了一番MSDN，然后Google一下，果然可行，构建方法如下：

//创建IHTMLDocument2接口
CoCreateInstance(CLASS_HTMLDocument, nil, CLSCTX_INPROC_SERVER, IID_IHTMLDocument2, FHtmlDoc);

接口创建好了之后就能够对文档元素进行解析了，很是爽快！

结合了我自己的特有操作，我对Combobox,Table,Frame等一些网页元素做了相应的封装，实现了一个HTMLParser，大致代码如下：

这里只给出声明，代码请在最后下载

代码

(******************************************************)
(*                得闲工作室                          *)
(*              网页元素操作类库                      *)
(*                                                    *)
(*              DxHtmlElement Unit                    *)
(*    Copyright(c) 2008-2010  不得闲                  *)
(*    email:appleak46@yahoo.com.cn     QQ:75492895    *)
(******************************************************)
unit DxHtmlElement;

interface
uses Windows,sysUtils,Clipbrd,MSHTML,ActiveX,OleCtrls,Graphics,TypInfo;

{Get EleMent Type}
function IsSelectElement(eleElement: IHTMLElement): Boolean;
function IsPwdElement(eleElement: IHTMLElement): Boolean;
function IsTextElement(element: IHTMLElement): boolean;
function IsTableElement(element: IHTMLElement): Boolean;
function IsElementCollection(element: IHTMLElement): Boolean;
function IsChkElement(element: IHTMLElement): boolean;
function IsRadioBtnElement(element: IHTMLElement): boolean;
function IsMemoElement(element: IHTMLElement): boolean;
function IsFormElement(element: IHTMLElement): boolean;
function IsIMGElement(element: IHTMLElement): boolean;
function IsInIMGElement(element: IHTMLElement): boolean;
function IsLabelElement(element: IHTMLElement): boolean;
function IsLinkElement(element: IHTMLElement): boolean;
function IsListElement(element: IHTMLElement): boolean;
function IsControlElement(element: IHTMLElement): boolean;
function IsObjectElement(element: IHTMLElement): boolean;
function IsFrameElement(element: IHTMLElement): boolean;
function IsInPutBtnElement(element: IHTMLElement): boolean;
function IsInHiddenElement(element: IHTMLElement): boolean;
function IsSubmitElement(element: IHTMLElement): boolean;
{Get ImgElement Data}
function GetPicIndex(doc: IHTMLDocument2; Src: string; Alt: string): Integer;
function GetPicElement(doc: IHTMLDocument2;imgName: string;src: string;Alt: string): IHTMLImgElement;
function GetRegCodePic(doc: IHTMLDocument2;ImgName: string; Src: string; Alt: string): TPicture; overload;
function GetRegCodePic(doc: IHTMLDocument2;Index: integer): TPicture; overload;
function GetRegCodePic(doc: IHTMLDocument2;element: IHTMLIMGElement): TPicture;overload;

type
  TObjectFromLResult = function(LRESULT: lResult;const IID: TIID; WPARAM: wParam;out pObject): HRESULT; stdcall;
  TEleMentType = (ELE_UNKNOW,ELE_TEXT,ELE_PWD,ELE_SELECT,ELE_CHECKBOX,ELE_RADIOBTN,ELE_MEMO,ELE_FORM,ELE_IMAGE,
  ELE_LABEL,ELE_LINK,ELE_LIST,ELE_CONTROL,ELE_OBJECT,ELE_FRAME,ELE_INPUTBTN,ELE_INIMAGE,ELE_INHIDDEN);

function GetElementType(element: IHTMLELEMENT): TEleMentType;
function GetElementTypeName(element: IHTMLELEMENT): string;
function GetHtmlTableCell(aTable: IHTMLTable;aRow,aCol: Integer): IHTMLElement;
function GetHtmlTable(aDoc: IHTMLDocument2; aIndex: Integer): IHTMLTable;
function GetWebBrowserHtmlTableCellText(Doc: IHTMLDocument2;
         const TableIndex, RowIndex, ColIndex: Integer;var ResValue: string):   Boolean;
function GetHtmlTableRowHtml(aTable: IHTMLTable; aRow: Integer): IHTMLElement;

function GetWebBrowserHtmlTableCellHtml(Doc: IHTMLDocument2;
         const TableIndex,RowIndex,ColIndex: Integer;var ResValue: string):   Boolean;
function GeHtmlTableHtml(aTable: IHTMLTable; aRow: Integer): IHTMLElement;
function GetWebBrowserHtmlTableHtml(Doc: IHTMLDocument2;
         const TableIndex,RowIndex: Integer;var ResValue: string):   Boolean;

type
  TDxWebFrameCollection = class;
  TDxWebElementCollection = class;

  TLoadState = (Doc_Loading,Doc_Completed,Doc_Invalidate);

  TDxWebFrame = class
  private
    FFrame: IHTMLWINDOW2;
    FElementCollections: TDxWebElementCollection;
    FWebFrameCollections: TDxWebFrameCollection;
    function GetSrc: string;
    function GetElementCount: integer;
    function GetWebFrameCollections: TDxWebFrameCollection;
    function GetElementCollections: TDxWebElementCollection;
    function GetDocument: IHTMLDOCUMENT2;
    function GetReadState: TLoadState;
    function GetIsLoaded: boolean;
    procedure SetFrame(const Value: IHTMLWINDOW2);
    function GetName: string;
  public
    Constructor Create(IFrame: IHTMLWINDOW2);
    Destructor Destroy;override;
    property Frame: IHTMLWINDOW2 read FFrame write SetFrame;
    property Src: string read GetSrc;
    property Document: IHTMLDOCUMENT2 read GetDocument;
    property Name: string read GetName;
    property Frames: TDxWebFrameCollection read GetWebFrameCollections;
    property ElementCount: integer read GetElementCount;
    property ElementCollections: TDxWebElementCollection read GetElementCollections;
    property ReadyState: TLoadState read GetReadState;
    property IsLoaded: boolean read GetIsLoaded;
  end;

  TDxWebFrameCollection = Class
  private
    FFrameCollection: IHTMLFramesCollection2;
    Frame: TDxWebFrame;
    function GetCount: integer;
    function GetFrameInterfaceByIndex(index: integer): IHTMLWINDOW2;
    function GetFrameInterfaceByName(Name: string): IHTMLWINDOW2;
    function GetFrameByIndex(index: integer): TDxWebFrame;
    function GetFrameByName(Name: string): TDxWebFrame;
    procedure SetFrameCollection(const Value: IHTMLFramesCollection2);
  public
    Constructor Create(ACollection: IHTMLFramesCollection2);
    Destructor Destroy;override;
    property FrameCollection: IHTMLFramesCollection2 read FFrameCollection write SetFrameCollection;
    property Count: integer read GetCount;
    property FrameInterfaceByIndex[index: integer]: IHTMLWINDOW2 read GetFrameInterfaceByIndex;
    property FrameInterfaceByName[Name: string]: IHTMLWINDOW2 read GetFrameInterfaceByName;

    property FrameByIndex[index: integer]: TDxWebFrame read GetFrameByIndex;
    property FrameByName[Name: string]: TDxWebFrame read GetFrameByName;
  end;

  TDxWebElementCollection = class
  private
    FCollection: IHTMLElementCollection;
    FChildCollection:  TDxWebElementCollection;
    function GetCollection(index: String): TDxWebElementCollection;
    function GetCount: integer;
    function GetElement(itemName: string; index: integer): IHTMLElement;
    function GetElementByName(itemName: string): IHTMLELEMENT;
    function GetElementByIndex(index: integer): IHTMLELEMENT;
    procedure SetCollection(const Value: IHTMLElementCollection);
  public
    Constructor Create(ACollection: IHTMLElementCollection);
    Destructor Destroy;override;
    property Collection: IHTMLElementCollection read FCollection write SetCollection;
    property ChildElementCollection[index: String]: TDxWebElementCollection read GetCollection;
    property ElementCount: integer read GetCount;
    property Element[itemName: string;index: integer]: IHTMLElement read GetElement;
    property ElementByName[itemName: string]: IHTMLELEMENT read GetElementByName;
    property ElementByIndex[index: integer]: IHTMLELEMENT read GetElementByIndex;
  end;

  TLinkCollection = class(TDxWebElementCollection)

  end;
  TDxWebTable = class;

  TDxTableCollection = class
  private
    FTableCollection: IHTMLElementCollection;
    FDocument: IHTMLDOCUMENT2;
    FWebTable: TDxWebTable;
    function GetTableInterfaceByName(AName: string): IHTMLTABLE;
    procedure SetDocument(Value: IHTMLDOCUMENT2);
    function GetTableInterfaceByIndex(index: integer): IHTMLTABLE;
    function GetCount: integer;
    function GetTableByIndex(index: integer): TDxWebTable;
    function GetTableByName(AName: string): TDxWebTable;
  public
    Constructor Create(Doc: IHTMLDOCUMENT2);
    destructor Destroy;override;
    property TableInterfaceByName[AName: string]: IHTMLTABLE read GetTableInterfaceByName;
    property TableInterfaceByIndex[index: integer]: IHTMLTABLE read GetTableInterfaceByIndex;

    property TableByName[AName: string]: TDxWebTable read GetTableByName;
    property TableByIndex[index: integer]: TDxWebTable read GetTableByIndex;

    property Document: IHTMLDOCUMENT2 read FDocument write SetDocument;
    property Count: integer read GetCount;
  end;

  TDxWebTable = class
  private
    FTableInterface: IHTMLTABLE;
    function GetRowCount: integer;
    procedure SetTableInterface(const Value: IHTMLTABLE);
    function GetCell(ACol, ARow: integer): string;
    function GetRowColCount(RowIndex: integer): integer;
    function GetInnerHtml: string;
    function GetInnerText: string;
    function GetCellElement(ACol, ARow: Integer): IHTMLTableCell;
  public
    Constructor Create(ATable: IHTMLTABLE);
    property TableInterface: IHTMLTABLE read FTableInterface write SetTableInterface;
    property RowCount: integer read GetRowCount;
    property Cell[ACol: integer;ARow: integer]: string read GetCell;
    property CellElement[ACol: Integer;ARow: Integer]: IHTMLTableCell read GetCellElement;
    property RowColCount[RowIndex: integer]: integer read GetRowColCount;
    property InnerHtml: string read GetInnerHtml;
    property InnerText: string read GetInnerText;
  end;

  TDxWebCombobox = class
  private
    FHtmlSelect: IHTMLSelectElement;
    function GetCount: Integer;
    procedure SetItemIndex(const Value: Integer);
    function GetItemIndex: Integer;
    function GetName: string;
    procedure SetName(const Value: string);
    function GetValue: string;
    procedure SetValue(const Value: string);
    procedure SetCombInterface(const Value: IHTMLSelectElement);
    function GetItemByName(EleName: string): string;
    function GetItemByIndex(index: integer): string;
    function GetItemAttribute(index: Integer; AttribName: string): OleVariant;
  public
    constructor Create(AWebCombo: IHTMLSelectElement);
    procedure Add(Ele: IHTMLElement);
    procedure Insert(Ele: IHTMLElement;Index: Integer);
    procedure Remove(index: Integer);

    property CombInterface: IHTMLSelectElement read FHtmlSelect write SetCombInterface;
    property Count: Integer read GetCount;
    property ItemIndex: Integer read GetItemIndex write SetItemIndex;
    property ItemByIndex[index: integer]: string read GetItemByIndex;
    property ItemByName[EleName: string]: string read GetItemByName;
    property ItemAttribute[index: Integer;AttribName: string]: OleVariant read GetItemAttribute;
    property Name: string read GetName write SetName;
    property value: string read GetValue write SetValue;
  end;

implementation
end.

HTMLParser解析类的代码实现单元

代码

(******************************************************)
(*                得闲工作室                          *)
(*              HTML解析单元库                        *)
(*                                                    *)
(*              DxHtmlParser Unit                     *)
(*    Copyright(c) 2008-2010  不得闲                  *)
(*    email:appleak46@yahoo.com.cn     QQ:75492895    *)
(******************************************************)
unit DxHtmlParser;

interface
uses Windows,MSHTML,ActiveX,DxHtmlElement,Forms;

type
  TDxHtmlParser = class
  private
    FHtmlDoc: IHTMLDocument2;
    FHTML: string;
    FWebTables: TDxTableCollection;
    FWebElements: TDxWebElementCollection;
    FWebComb: TDxWebCombobox;
    procedure SetHTML(const Value: string);
    function GetWebCombobox(AName: string): TDxWebCombobox;
  public
    constructor Create;
    destructor Destroy;override;
    property HTML: string read FHTML write SetHTML;
    property WebTables: TDxTableCollection read FWebTables;
    property WebElements: TDxWebElementCollection read FWebElements;
    property WebCombobox[Name: string]: TDxWebCombobox read GetWebCombobox;
  end;
implementation

{ TDxHtmlParser }

constructor TDxHtmlParser.Create;
begin
  CoInitialize(nil);
  //创建IHTMLDocument2接口
  CoCreateInstance(CLASS_HTMLDocument, nil, CLSCTX_INPROC_SERVER, IID_IHTMLDocument2, FHtmlDoc);
  Assert(FHtmlDoc<>nil,'构建HTMLDocument接口失败');
  FHtmlDoc.Set_designMode('On'); //设置为设计模式，不执行脚本
  while not (FHtmlDoc.readyState = 'complete') do
  begin
    sleep(1);
    Application.ProcessMessages;
  end;
  FWebTables := TDxTableCollection.Create(FHtmlDoc);
  FWebElements := TDxWebElementCollection.Create(nil);
  FWebComb := TDxWebCombobox.Create(nil);
end;

destructor TDxHtmlParser.Destroy;
begin
  FWebTables.Free;
  FWebElements.Free;
  FWebComb.Free;
  CoUninitialize;
  inherited;
end;

function TDxHtmlParser.GetWebCombobox(AName: string): TDxWebCombobox;
begin
   if FWebElements.Collection <> nil then
   begin
     FWebComb.CombInterface := FWebElements.ElementByName[AName] as IHTMLSelectElement;
     Result := FWebComb;
   end
   else Result := nil;
end;

procedure TDxHtmlParser.SetHTML(const Value: string);
begin
  if FHTML <> Value then
  begin
    FHTML := Value;
    FHtmlDoc.body.innerHTML := FHTML;
    FWebElements.Collection := FHtmlDoc.all;
  end;
end;

end.

全部代码下载

posted on 2010-01-05 09:08 不得闲阅读(14849) 评论(13) 编辑收藏举报