public SmartCrawler()
{
Http = new HttpItem();
CrawlItems = new ObservableCollection<CrawlItem>();
Http.URL = "http://www.cnblogs.com/";
helper = new HttpHelper();
IsMultiData = ListType.List;
Documents = new ObservableCollection<HttpItem>();
}
/// <summary>
/// 采用https协议访问网络,根据传入的URl地址，得到响应的数据字符串。
/// </summary>
/// <param name="requestitem">参数列表</param>
/// <returns>String类型的数据</returns>
public string GetHtml(HttpItem requestitem, out WebHeaderCollection responseHeaders, out HttpStatusCode code, string url = null, string post = null)
{
try
{
var request = SetRequest(requestitem, url, post);
var r = GetHttpRequestData(request, requestitem,out responseHeaders, out code);
if (!IsSuccess(code))
return "HTTP错误，类型:" + code;
return r;
}
catch (Exception ex)
{
code = HttpStatusCode.NotFound;
responseHeaders = null;
return ex.Message;
}
}
/// <summary>
/// 为请求准备参数
/// </summary>
/// <param name="item">参数列表</param>
/// <param name="_Encoding">读取数据时的编码方式</param>
private HttpWebRequest SetRequest(HttpItem item, string desturl = null, string post = null)
{
var url = desturl ?? item.URL;
if (url == null)
return null;
if (url.Contains("http") == false)
{
url = "http://" + url;
}
// 验证证书
if (url.Contains("https"))
ServicePointManager.ServerCertificateValidationCallback =
(sender, certificate, chain, sslPolicyErrors) => true;
//初始化对像，并设置请求的URL地址
var request = (HttpWebRequest) WebRequest.Create(GetUrl(url));
SetRequest(item, request, desturl, post);
encoding = AttributeHelper.GetEncoding(item.Encoding);
return request;
}
public static void SetRequest(HttpItem item, HttpWebRequest request, string desturl = null, string post = null)
{
var docu = item.GetHeaderParameter();
// 设置代理
//SetProxy(item);
//请求方式Get或者Post
request.Method = item.Method.ToString();
request.Timeout = item.Timeout;
request.ReadWriteTimeout = item.ReadWriteTimeout;
//Accept
request.Headers = new WebHeaderCollection();
if (docu["Headers"].ToString() != "")
{
var str = docu["Headers"].ToString().Split('\n');
foreach (var s in str)
{
var ms = s.Split(':');
if (ms.Length != 2)
continue;
var key = ms[0].Trim();
var value = ms[1].Trim();
if (SetHeaderValue(request.Headers, key, value) == false)
{
request.Headers.Add(key, value);
}
}
}
request.Accept = docu["Accept"].ToString();
//ContentType返回类型
request.ContentType = docu["Content_Type"].ToString();
//UserAgent客户端的访问类型，包括浏览器版本和操作系统信息
request.UserAgent = docu["User-Agent"].ToString();
var host = docu["Host"].ToString();
//if (string.IsNullOrEmpty(host) == false)
// request.Host = host;
//设置Cookie
var cookie = docu["Cookie"].ToString();
if (!string.IsNullOrEmpty(cookie))
{
request.Headers[HttpRequestHeader.Cookie] = cookie;
}
//来源地址
request.Referer = docu["Referer"].ToString();
//是否执行跳转功能
request.AllowAutoRedirect = item.Allowautoredirect;
//设置Post数据
string postdata = null;
if (post == null)
{
postdata = item.Postdata;
}
else
{
postdata = post;
}
//验证在得到结果时是否有传入数据
if (!string.IsNullOrEmpty(postdata) && request.Method.Trim().ToLower().Contains("post"))
{
var buffer = Encoding.Default.GetBytes(postdata);
request.ContentLength = buffer.Length;
request.GetRequestStream().Write(buffer, 0, buffer.Length);
}
////设置最大连接
//if (item.Connectionlimit > 0)
//{
// request.ServicePoint.ConnectionLimit = item.Connectionlimit;
//}
}
/// <summary>
/// 根据相传入的数据，得到相应页面数据
/// </summary>
/// <param name="strPostdata">传入的数据Post方式,get方式传NUll或者空字符串都可以</param>
/// <param name="ContentType">返回的响应数据的类型</param>
/// <returns>string类型的响应数据</returns>
private string GetHttpRequestData(HttpWebRequest request, HttpItem objhttpitem,out WebHeaderCollection responseHeaders, out HttpStatusCode statusCode)
{
var result = "";
#region 得到请求的response
using (var response = (HttpWebResponse) request.GetResponse())
{
MemoryStream stream;
var docu = objhttpitem.GetHeaderParameter();
if (response.Headers["set-cookie"] != null)
docu["Cookie"] = MergeCookie(docu["Cookie"].ToString(), response.Headers["set-cookie"]);
responseHeaders= response.Headers;
statusCode = response.StatusCode;
objhttpitem.Parameters = HttpItem.HeaderToString(docu);
//GZIIP处理
if (response.ContentEncoding != null &&
response.ContentEncoding.Equals("gzip", StringComparison.InvariantCultureIgnoreCase))
{
stream =
GetMemoryStream(new GZipStream(response.GetResponseStream(), CompressionMode.Decompress));
}
else
{
stream = GetMemoryStream(response.GetResponseStream());
}
//获取Byte
var rawResponse = stream.ToArray();
//是否返回Byte类型数据
if (objhttpitem.Encoding == EncodingType.Unknown || encoding == null)
{
var temp = Encoding.Default.GetString(rawResponse, 0, rawResponse.Length);
//<meta(.*?)charset([\s]?)=[^>](.*?)>
var meta = Regex.Match(temp, "<meta([^<]*)charset=([^<]*)[\"']",
RegexOptions.IgnoreCase | RegexOptions.Multiline);
var charter = (meta.Groups.Count > 2) ? meta.Groups[2].Value : string.Empty;
charter = charter.Replace("\"", string.Empty)
.Replace("'", string.Empty)
.Replace(";", string.Empty);
if (charter.Length > 0)
{
charter = charter.ToLower().Replace("iso-8859-1", "gbk");
if (charter.Contains("utf-8") || charter.Contains("UTF-8"))
{
encoding = Encoding.UTF8;
}
else if (charter.Contains("gb"))
{
encoding = Encoding.GetEncoding("GB2312");
}
else
{
encoding = Encoding.GetEncoding(charter);
}
}
else
{
if (response.CharacterSet != null && response.CharacterSet.ToLower().Trim() == "iso-8859-1")
{
encoding = Encoding.GetEncoding("gbk");
}
}
}
//得到返回的HTML
result = encoding.GetString(rawResponse);
//最后释放流
stream.Close();
}
return result;
}
public override object TransformData(IFreeDocument datas)
{
var text = datas[Column];
if (text == null)
return null;
var pattern = "all";
var format = "plain";
switch (ResultType)
{
case ContentType.Json:
format = "json";
break;
case ContentType.Text:
format = "plain";
break;
case ContentType.XML:
format = "xml";
break;
case ContentType.Byte:
format = "conll";
break;
}
switch (Pattern)
{
case Pattern.分词:
pattern = "ws";
break;
case Pattern.词性标注:
pattern = "pos";
break;
case Pattern.依存句法分析:
pattern = "dp";
break;
case Pattern.语义依存分析:
pattern = "sdp";
break;
case Pattern.语义角色标注:
pattern = "srl";
break;
case Pattern.命名实体识别:
pattern = "ner";
break;
}
var param = ("api_key=" + apiKey +
"&pattern=" + pattern +
"&format=" + format+
"&text=" + HttpUtility.UrlEncode(text.ToString())
);
var docs = buffHelper.Get(param);
if (docs == null)
{
var item = new HttpItem();
item.URL = uriBase+'?'+ param;
item.Method = MethodType.GET;
item.Encoding = EncodingType.UTF8;
var helper = new HttpHelper();
HttpStatusCode code;
var result = helper.GetHtml(item,out code);
if(code==HttpStatusCode.OK)
buffHelper.Set(param, result);
return result;
}
return docs;
}
public static HtmlDocument GetHtmlDocument(string url)
{
var httpitem = new HttpItem {URL = url};
var helper = new HttpHelper();
HttpStatusCode statusCode;
var doc2 = helper.GetHtml(httpitem, out statusCode);
if (statusCode != HttpStatusCode.OK)
return null;
var htmldoc = new HtmlDocument();
htmldoc.LoadHtml(doc2);
return htmldoc;
}
/// <summary>
/// 根据相传入的数据，得到相应页面数据
/// </summary>
/// <param name="strPostdata">传入的数据Post方式,get方式传NUll或者空字符串都可以</param>
/// <param name="ContentType">返回的响应数据的类型</param>
/// <returns>string类型的响应数据</returns>
private byte[] GetHttpRequestFile(HttpWebRequest request, HttpItem objhttpitem, out HttpStatusCode statusCode)
{
byte[] result = null;
using (var response = (HttpWebResponse) request.GetResponse())
{
var _stream = new MemoryStream();
var docu = objhttpitem.GetHeaderParameter();
if (response.Headers["set-cookie"] != null)
docu["Cookie"] = MergeCookie(docu["Cookie"].ToString(), response.Headers["set-cookie"]);
statusCode = response.StatusCode;
objhttpitem.Parameters = HttpItem.HeaderToString(docu);
//GZIIP处理
if (response.ContentEncoding != null &&
response.ContentEncoding.Equals("gzip", StringComparison.InvariantCultureIgnoreCase))
{
//开始读取流并设置编码方式
//new GZipStream(response.GetResponseStream(), CompressionMode.Decompress).CopyTo(_stream, 10240);
//.net4.0以下写法
_stream =
GetMemoryStream(new GZipStream(response.GetResponseStream(), CompressionMode.Decompress));
}
else
{
//开始读取流并设置编码方式
//response.GetResponseStream().CopyTo(_stream, 10240);
//.net4.0以下写法
_stream = GetMemoryStream(response.GetResponseStream());
}
//获取Byte
result = _stream.ToArray();
//是否返回Byte类型数据
_stream.Close();
}
return result;
}
public override void DictDeserialize(IDictionary<string, object> dicts, Scenario scenario = Scenario.Database)
{
base.DictDeserialize(dicts, scenario);
URL = dicts.Set("URL", URL);
RootXPath = dicts.Set("RootXPath", RootXPath);
IsMultiData = dicts.Set("IsMultiData", IsMultiData);
URLFilter = dicts.Set("URLFilter", URLFilter);
Crawler = dicts.Set("Crawler", Crawler);
ContentFilter = dicts.Set("ContentFilter", ContentFilter);
if (dicts.ContainsKey("HttpSet"))
{
var doc2 = dicts["HttpSet"];
var p = doc2 as IDictionary<string, object>;
Http.UnsafeDictDeserialize(p);
}
if (dicts.ContainsKey("Login"))
{
var doc2 = dicts["Login"];
var p = doc2 as IDictionary<string, object>;
var item = new HttpItem();
item.DictDeserialize(p);
Documents.Add(item);
}
if (dicts.ContainsKey("Generator"))
{
var doc2 = dicts["Generator"];
var p = doc2 as IDictionary<string, object>;
}
var doc = dicts as FreeDocument;
if (doc?.Children != null)
{
foreach (var child in doc.Children)
{
var item = new CrawlItem();
item.DictDeserialize(child);
CrawlItems.Add(item);
}
}
}
private void FiddlerApplicationAfterSessionComplete(Session oSession)
{
if (string.IsNullOrEmpty(URLFilter) == false)
{
URLFilter = URLFilter.Replace("http://", "");
if (URLFilter.Split(' ').Any(item => oSession.url.Contains(item) == false))
{
return;
}
}
var httpitem = new HttpItem {Parameters = oSession.oRequest.headers.ToString()};
if ((oSession.BitFlags & SessionFlags.IsHTTPS) != 0)
{
httpitem.URL = "https://" + oSession.url;
}
else
{
httpitem.URL = "http://" + oSession.url;
}
httpitem.Postdata = Encoding.Default.GetString(oSession.RequestBody);
if (string.IsNullOrEmpty(httpitem.Postdata) == false)
{
httpitem.Method = MethodType.POST;
ControlExtended.UIInvoke(() => Documents.Add(httpitem));
}
if (string.IsNullOrEmpty(ContentFilter) == false)
{
if (ContentFilter.Split(' ').Any(item => oSession.GetResponseBodyAsString().Contains(item) == false))
{
return;
}
}
httpitem.DictCopyTo(Http);
XLogSys.Print.Info("已经成功获取嗅探字段" + oSession.url);
}
/// <summary>
/// 采用https协议访问网络,根据传入的URl地址，得到响应的数据字符串。
/// </summary>
/// <param name="requestitem">参数列表</param>
/// <returns>String类型的数据</returns>
public string GetHtml(HttpItem requestitem, out HttpStatusCode code, string url = null, string post = null)
{
WebHeaderCollection responseHeaders = null;
return GetHtml(requestitem, out responseHeaders,out code, url, post);
}
public static void SetRequest(HttpItem item, HttpWebRequest request, string desturl = null, string post = null)
{
var docu = item.GetHeaderParameter();
// 设置代理
if (item.ProxyPort == 0 || string.IsNullOrEmpty(item.ProxyIP))
{
//不需要设置
}
else
{
//设置代理服务器
var myProxy = new WebProxy(item.ProxyIP, item.ProxyPort);
//建议连接
myProxy.Credentials = new NetworkCredential(item.ProxyUserName, item.ProxyPassword);
//给当前请求对象
request.Proxy = myProxy;
//设置安全凭证
request.Credentials = CredentialCache.DefaultNetworkCredentials;
}
//请求方式Get或者Post
request.Method = item.Method.ToString();
request.Timeout = item.Timeout;
request.ReadWriteTimeout = item.ReadWriteTimeout;
//Accept
request.Headers = new WebHeaderCollection();
if (docu["Headers"].ToString() != "")
{
var str = docu["Headers"].ToString().Split('\n');
foreach (var s in str)
{
var ms = s.Split(':');
if (ms.Length != 2)
continue;
var key = ms[0].Trim();
var value = ms[1].Trim();
if (SetHeaderValue(request.Headers, key, value) == false)
{
request.Headers.Add(key, value);
}
}
}
request.Accept = docu["Accept"].ToString();
//ContentType返回类型
request.ContentType = docu["Content_Type"].ToString();
//UserAgent客户端的访问类型，包括浏览器版本和操作系统信息
request.UserAgent = docu["User-Agent"].ToString();
var host = docu["Host"].ToString();
//if (string.IsNullOrEmpty(host) == false)
// request.Host = host;
//设置Cookie
var cookie = docu["Cookie"].ToString();
if (!string.IsNullOrEmpty(cookie))
{
request.Headers[HttpRequestHeader.Cookie] = cookie;
}
//来源地址
request.Referer = docu["Referer"].ToString();
//是否执行跳转功能
request.AllowAutoRedirect = item.Allowautoredirect;
//设置Post数据
string postdata = null;
if (post == null)
{
postdata = item.Postdata;
}
else
{
postdata = post;
}
//验证在得到结果时是否有传入数据
if (!string.IsNullOrEmpty(postdata) && request.Method.Trim().ToLower().Contains("post"))
{
var buffer = Encoding.Default.GetBytes(postdata);
request.ContentLength = buffer.Length;
request.GetRequestStream().Write(buffer, 0, buffer.Length);
}
////设置最大连接
//if (item.Connectionlimit > 0)
//{
// request.ServicePoint.ConnectionLimit = item.Connectionlimit;
//}
}
/// <summary>
/// 根据相传入的数据，得到相应页面数据
/// </summary>
/// <param name="strPostdata">传入的数据Post方式,get方式传NUll或者空字符串都可以</param>
/// <param name="ContentType">返回的响应数据的类型</param>
/// <returns>string类型的响应数据</returns>
private string GetHttpRequestData(HttpWebRequest request, HttpItem objhttpitem,out WebHeaderCollection responseHeaders, out HttpStatusCode statusCode)
{
var result = "";
#region 得到请求的response
using (var response = (HttpWebResponse) request.GetResponse())
{
MemoryStream stream;
var docu = objhttpitem.GetHeaderParameter();
if (response.Headers["set-cookie"] != null)
docu["Cookie"] = MergeCookie(docu["Cookie"].ToString(), response.Headers["set-cookie"]);
responseHeaders= response.Headers;
statusCode = response.StatusCode;
objhttpitem.Parameters = HttpItem.HeaderToString(docu);
//GZIIP处理
if (response.ContentEncoding != null &&
response.ContentEncoding.Equals("gzip", StringComparison.InvariantCultureIgnoreCase))
{
stream =
GetMemoryStream(new GZipStream(response.GetResponseStream(), CompressionMode.Decompress));
}
else
{
stream = GetMemoryStream(response.GetResponseStream());
}
//获取Byte
var rawResponse = stream.ToArray();
//是否返回Byte类型数据
if (objhttpitem.Encoding == EncodingType.Unknown || encoding == null)
{
var temp = Encoding.Default.GetString(rawResponse, 0, rawResponse.Length);
//<meta(.*?)charset([\s]?)=[^>](.*?)>
var meta = Regex.Match(temp, "<meta([^<]*)charset=([^<]*)[\"']",
RegexOptions.IgnoreCase | RegexOptions.Multiline);
var charter = (meta.Groups.Count > 2) ? meta.Groups[2].Value : string.Empty;
charter = charter.Replace("\"", string.Empty)
.Replace("'", string.Empty)
.Replace(";", string.Empty);
if (charter.Length > 0)
{
charter = charter.ToLower().Replace("iso-8859-1", "gbk");
if (charter.Contains("utf-8") || charter.Contains("UTF-8"))
{
encoding = Encoding.UTF8;
}
else if (charter.Contains("gb"))
{
encoding = Encoding.GetEncoding("GB2312");
}
else
{
encoding = Encoding.GetEncoding(charter);
}
}
else
{
if (response.CharacterSet != null && response.CharacterSet.ToLower().Trim() == "iso-8859-1")
{
encoding = Encoding.GetEncoding("gbk");
}
}
}
//得到返回的HTML
result = encoding.GetString(rawResponse);
//最后释放流
stream.Close();
}
return result;
}
public static IEnumerable<List<FreeDocument>> GetMultiDataFromURL(string url)
{
var httpitem = new HttpItem {URL = url};
var helper = new HttpHelper();
HttpStatusCode statusCode;
var doc2 = helper.GetHtml(httpitem, out statusCode);
if (statusCode != HttpStatusCode.OK)
yield break;
if (doc2 == null)
yield return new List<FreeDocument>();
var htmldoc = new HtmlDocument();
htmldoc.LoadHtml(doc2);
foreach (var item in htmldoc.GetDataFromHtml())
{
yield return item;
}
}
/// <summary>
/// 采用https协议访问网络,根据传入的URl地址，得到响应的数据字符串。
/// </summary>
/// <param name="requestitem">参数列表</param>
/// <returns>String类型的数据</returns>
public string GetHtml(HttpItem requestitem, out HttpStatusCode code, string url = null, string post = null)
{
WebHeaderCollection responseHeaders = null;
return GetHtml(requestitem, out responseHeaders,out code, url, post);
}
/// <summary>
/// 采用https协议访问网络,根据传入的URl地址，得到响应的数据字符串。
/// </summary>
/// <param name="requestitem">参数列表</param>
/// <returns>String类型的数据</returns>
public byte[] GetFile(HttpItem requestitem, out HttpStatusCode code, string url = null, string post = null)
{
try
{
var request = SetRequest(requestitem, url, post);
var r = GetHttpRequestFile(request, requestitem, out code);
if (!IsSuccess(code))
XLogSys.Print.ErrorFormat("HTTP错误，URL:{0},类型:{1}", url, code.ToString());
return r;
}
catch (Exception ex)
{
code = HttpStatusCode.NotFound;
return new byte[0];
}
}
public CookieAwareWebClient(HttpItem item)
{
this.Encoding = Encoding.UTF8;
HttpItem = item;
}
public string Translate(string item)
{
var res = buffHelper.Get(item);
if (res != null)
return res;
if (string.IsNullOrWhiteSpace(item))
return item;
var httpitem = new HttpItem();
string url =
$"http://openapi.baidu.com/public/2.0/bmt/translate?client_id={ClientID}&q={item}&from={language[Source.SelectItem]}&to={language[Target.SelectItem]}";
httpitem.URL = url;
HttpStatusCode code;
string result = helper.GetHtml(httpitem,out code);
var r = JsonConvert.Import(result) as JsonObject;
if (r.Contains("error_code ") == false)
{
var sb = new StringBuilder();
var array = r["trans_result"] as JsonArray;
for (int i = 0; i < array.Length; i++)
{
var j = array[i] as JsonObject;
object r2 = j["dst"];
sb.AppendLine(r2.ToString());
}
string t = sb.ToString();
buffHelper.Set(item,t);
return t;
}
return "Error";
}
public static HtmlDocument GetDocumentFromURL(string url, EncodingType encoding = EncodingType.Unknown)
{
var httpitem = new HttpItem();
httpitem.URL = url;
httpitem.Encoding = encoding;
var helper = new HttpHelper();
HttpStatusCode code;
var doc = new HtmlDocument();
var result = helper.GetHtml(httpitem, out code);
if (!HttpHelper.IsSuccess(code))
return doc;
doc.LoadHtml(result);
return doc;
}
