/// <summary>
/// 清理值中一些垃圾，比如js等等
/// </summary>
/// <param name="item"></param>
/// <param name="value"></param>
public void ClearRubbish(SnifferItem item, ref string value)
{
if (!string.IsNullOrEmpty(item.ClearRegexString))
{
value = Regex.Replace(value, item.ClearRegexString, string.Empty, (RegexOptions)25);
}
value = Regex.Replace(value, "<script[^>]*?>(.*?)</script>", string.Empty, (RegexOptions)25);
value = Regex.Replace(value, "(<[^>]*? )(onclick=['\"].*?['\"])(( [^>]*?)?>.*?</[^>]*?>)", "$1$3", (RegexOptions)25);
value = Regex.Replace(value, "(<[^>]*?)=(['\"](java|vb)script:.*?['\"])(( [^>]*?)?>.*?</[^>]*?>)", "$1$4", (RegexOptions)25);
value = Regex.Replace(value, "<!--.*?-->", "", (RegexOptions)25);
}
/// <summary>
/// 清理A元素
/// </summary>
/// <param name="item"></param>
/// <param name="value"></param>
public void ClearAElement(SnifferItem item, ref string value)
{
value = Regex.Replace(value, "<a\\s[^>]*?>(.*?)</a>", "$1", (RegexOptions)25);
//MatchCollection matchs = Regex.Matches(value, "<a [^>]*?>(.*?)</a>", (RegexOptions)25);
//List<Match> matchList = new List<Match>();
//foreach (Match match in matchs)
//{
// bool bool1 = false;
// foreach (Match match1 in matchList)
// {
// if (string.Compare(match1.Value, match.Value, true) == 0)
// bool1 = true;
// }
// if (!bool1)
// {
// value = value.Replace(match.Value, match.Groups[1].Value);
// matchList.Add(match);
// }
//}
}
/// <summary>
/// 将链接替成绝对路径
/// </summary>
/// <param name="item"></param>
/// <param name="value"></param>
public void UrlToAbs(SnifferItem item, ref string value)
{
MatchCollection matchs = Regex.Matches(value, "((?<=<a[^>]*?href=[\"']?)(\\.|/|http)[^\"' >]*)|((?<=<img[^>]*?src=[\"']?)(\\.|/|http)[^\"' >]*)", (RegexOptions)25);
List <Match> matchList = new List <Match>();
foreach (Match match in matchs)
{
bool bool1 = false;
foreach (Match match1 in matchList)
{
if (string.Compare(match1.Value, match.Value, true) == 0)
{
bool1 = true;
}
}
if (!bool1)
{
string absUrl = FileUtil.GetAbsUrl(match.Value, this.SubPageBaseUrl);
value = value.Replace(match.Value, absUrl);
matchList.Add(match);
}
}
}
private static DetailPageConfiguration CreateDetailPageConfiguration(PageConfiguration parent, XmlNode parentConfNode, XmlNode pageConfNode)
{
XmlNode commonListConfigNode = null;
XmlNode commonConfigNode = null;
if (parentConfNode.Attributes["Config"] != null)
{
commonListConfigNode = parentConfNode.ParentNode.SelectSingleNode(string.Format("CommonConfig[@Name='{0}']", parentConfNode.Attributes["Config"].Value));
commonConfigNode = commonListConfigNode.SelectSingleNode("SnifferPage");
}
DetailPageConfiguration detailPageConf = new DetailPageConfiguration(parent);
detailPageConf.PageName = pageConfNode.Attributes["PageName"].Value;
if (pageConfNode.Attributes["PageUrl"] != null)
{
detailPageConf.PageUrl = pageConfNode.Attributes["PageUrl"].Value;
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageUrl"] != null)
{
detailPageConf.PageUrl = commonConfigNode.Attributes["PageUrl"].Value;
}
if (pageConfNode.Attributes["PageType"] != null)
{
detailPageConf.PageType = (PageType)Enum.Parse(typeof(PageType), pageConfNode.Attributes["PageType"].Value, false);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageType"] != null)
{
detailPageConf.PageType = (PageType)Enum.Parse(typeof(PageType), commonConfigNode.Attributes["PageType"].Value, false);
}
if (pageConfNode.Attributes["Encoding"] != null)
{
detailPageConf.Encoding = Encoding.GetEncoding(pageConfNode.Attributes["Encoding"].Value);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["Encoding"] != null)
{
detailPageConf.Encoding = Encoding.GetEncoding(commonConfigNode.Attributes["Encoding"].Value);
}
//翻页
if (pageConfNode.Attributes["PageQuery"] != null)
{
detailPageConf.PageQuery = pageConfNode.Attributes["PageQuery"].Value;
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageQuery"] != null)
{
detailPageConf.PageQuery = commonConfigNode.Attributes["PageQuery"].Value;
}
if (pageConfNode.Attributes["ReplacePageQuery"] != null)
{
detailPageConf.ReplacePageQuery = pageConfNode.Attributes["ReplacePageQuery"].Value;
}
else if (commonConfigNode != null && commonConfigNode.Attributes["ReplacePageQuery"] != null)
{
detailPageConf.ReplacePageQuery = commonConfigNode.Attributes["ReplacePageQuery"].Value;
}
if (pageConfNode.Attributes["PageIndexFormat"] != null)
{
detailPageConf.PageIndexFormat = pageConfNode.Attributes["PageIndexFormat"].Value;
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageIndexFormat"] != null)
{
detailPageConf.PageIndexFormat = commonConfigNode.Attributes["PageIndexFormat"].Value;
}
if (pageConfNode.Attributes["PageStartIndex"] != null)
{
detailPageConf.PageStartIndex = int.Parse(pageConfNode.Attributes["PageStartIndex"].Value);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageStartIndex"] != null)
{
detailPageConf.PageStartIndex = int.Parse(commonConfigNode.Attributes["PageStartIndex"].Value);
}
if (pageConfNode.Attributes["PageIndexSeed"] != null)
{
detailPageConf.PageIndexSeed = int.Parse(pageConfNode.Attributes["PageIndexSeed"].Value);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageIndexSeed"] != null)
{
detailPageConf.PageIndexSeed = int.Parse(commonConfigNode.Attributes["PageIndexSeed"].Value);
}
if (pageConfNode.Attributes["PageIndexStep"] != null)
{
detailPageConf.PageIndexStep = int.Parse(pageConfNode.Attributes["PageIndexStep"].Value);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageIndexStep"] != null)
{
detailPageConf.PageIndexStep = int.Parse(commonConfigNode.Attributes["PageIndexStep"].Value);
}
if (pageConfNode.Attributes["EndPageDetermineRegex"] != null)
{
detailPageConf.EndPageDetermineRegex = pageConfNode.Attributes["EndPageDetermineRegex"].Value;
}
else if (commonConfigNode != null && commonConfigNode.Attributes["EndPageDetermineRegex"] != null)
{
detailPageConf.EndPageDetermineRegex = commonConfigNode.Attributes["EndPageDetermineRegex"].Value;
}
if (pageConfNode.Attributes["PageMethod"] != null)
{
detailPageConf.PageMethod = (PageMethod)Enum.Parse(typeof(PageMethod), pageConfNode.Attributes["PageMethod"].Value);
}
else if (commonConfigNode != null && commonConfigNode.Attributes["PageMethod"] != null)
{
detailPageConf.PageMethod = (PageMethod)Enum.Parse(typeof(PageMethod), commonConfigNode.Attributes["PageMethod"].Value);
}
XmlNode snifferUrlItemNode = pageConfNode.SelectSingleNode("SnifferSubPageUrlItem");
if (snifferUrlItemNode == null && commonConfigNode != null && commonConfigNode.Attributes["SnifferSubPageUrlItem"] != null)
{
snifferUrlItemNode = commonConfigNode.SelectSingleNode("SnifferSubPageUrlItem");
}
if (snifferUrlItemNode != null)
{
detailPageConf.SnifferSubPageUrlItem = CreateSnifferUrlItem(snifferUrlItemNode);
}
List <XmlNodeList> lstItemLists = new List <XmlNodeList>();
XmlNodeList itemNodes = pageConfNode.SelectNodes("SnifferItem");
if (itemNodes != null)
{
lstItemLists.Add(itemNodes);
}
if (commonConfigNode != null)
{
XmlNodeList comItemNodes = commonConfigNode.SelectNodes("SnifferItem");
if (comItemNodes != null)
{
lstItemLists.Add(comItemNodes);
}
}
for (int i = 0; i < lstItemLists.Count; i++)
{
itemNodes = lstItemLists[i];
foreach (XmlNode itemNode in itemNodes)
{
string itemName = itemNode.Attributes["ItemName"].Value;
bool bool1 = false;
foreach (SnifferItem itm in detailPageConf.SnifferItems)
{
if (itm.ItemName == itemName)
{
bool1 = true;
}
}
if (!bool1)
{
SnifferItem item = new SnifferItem();
item.ItemName = itemNode.Attributes["ItemName"].Value;
if (itemNode.Attributes["SaveImage"] != null)
{
item.SaveImage = bool.Parse(itemNode.Attributes["SaveImage"].Value);
}
if (itemNode.Attributes["SaveImagesPath"] != null)
{
item.SaveImagesPath = itemNode.Attributes["SaveImagesPath"].Value;
}
if (itemNode.Attributes["ImageUrlPath"] != null)
{
item.ImageUrlPath = itemNode.Attributes["ImageUrlPath"].Value;
}
if (itemNode.Attributes["IsClearHTML"] != null)
{
item.IsClearHTML = bool.Parse(itemNode.Attributes["IsClearHTML"].Value);
}
if (itemNode.Attributes["IsUrl"] != null)
{
item.IsUrl = bool.Parse(itemNode.Attributes["IsUrl"].Value);
}
if (itemNode.Attributes["UrlToAbs"] != null)
{
item.UrlToAbs = bool.Parse(itemNode.Attributes["UrlToAbs"].Value);
}
if (itemNode.Attributes["Separator"] != null)
{
item.Separator = itemNode.Attributes["Separator"].Value;
}
if (itemNode.Attributes["ClearAElement"] != null)
{
item.ClearAElement = bool.Parse(itemNode.Attributes["ClearAElement"].Value);
}
if (itemNode.Attributes["MutiPage"] != null)
{
item.MutiPage = bool.Parse(itemNode.Attributes["MutiPage"].Value);
}
if (itemNode.Attributes["MutiPageSeparator"] != null)
{
item.MutiPageSeparator = itemNode.Attributes["MutiPageSeparator"].Value;
}
XmlNode ndClearRegexString = itemNode.SelectSingleNode("ClearRegexString");
if (ndClearRegexString != null)
{
item.ClearRegexString = ndClearRegexString.InnerText;
}
XmlNode ndDefaultValue = itemNode.SelectSingleNode("DefaultValue");
if (ndDefaultValue != null)
{
item.DefaultValue = ndDefaultValue.InnerText;
}
XmlNode ndRegexString = itemNode.SelectSingleNode("RegexString");
if (ndRegexString != null)
{
item.RegexString = CreateRegexString(ndRegexString);
}
detailPageConf.SnifferItems.Add(item);
}
}
}
XmlNodeList subPageNodes = pageConfNode.SelectNodes("SnifferPage");
foreach (XmlNode subPageNode in subPageNodes)
{
detailPageConf.SubPageConfigurations.Add(CreateDetailPageConfiguration(detailPageConf, pageConfNode, subPageNode));
}
if (commonConfigNode != null)
{
subPageNodes = commonConfigNode.SelectNodes("SnifferPage");
foreach (XmlNode subPageNode in subPageNodes)
{
detailPageConf.SubPageConfigurations.Add(CreateDetailPageConfiguration(detailPageConf, pageConfNode, subPageNode));
}
}
return(detailPageConf);
}
/// <summary>
/// 保存图片
/// </summary>
/// <param name="value"></param>
public void SaveImages(SnifferItem item, ref string value)
{
value = value.Replace("wallpaper:", "http:");
MatchCollection imgMatchs;
if (item.IsUrl)
{
imgMatchs = Regex.Matches(value, "^.*$", (RegexOptions)25);
}
else
{
imgMatchs = Regex.Matches(value, "(?<=<img[^>]*?src=[\"']?)(\\.|/|http)[^\"' >]*", (RegexOptions)25);
}
ListPage parentPage = this.Parent as ListPage;
if (parentPage == null)
{
parentPage = this.Parent.Parent as ListPage;
}
string saveImagesPath = string.Empty;
string dtPath = DateTime.Now.ToString("yyyy\\\\MM");
if (!string.IsNullOrEmpty(item.SaveImagesPath))
{
saveImagesPath = string.Format(item.SaveImagesPath, dtPath);
}
else
{
saveImagesPath = parentPage.SavePathAndFileName.Substring(0, parentPage.SavePathAndFileName.LastIndexOf(".")) + "\\";
}
if (!saveImagesPath.EndsWith("\\"))
{
saveImagesPath += "\\";
}
if (!Directory.Exists(saveImagesPath))
{
Directory.CreateDirectory(saveImagesPath);
}
foreach (Match match in imgMatchs)
{
int lastIndex = match.Value.LastIndexOf("/") + 1;
string fileName;
if (lastIndex >= 0)
{
fileName = match.Value.Substring(lastIndex);
}
else
{
fileName = match.Value;
}
if (fileName.Contains("?"))
{
fileName = fileName.Substring(0, fileName.IndexOf("?"));
}
fileName = HttpUtility.UrlDecode(fileName);
try
{
string absUrl = FileUtil.GetAbsUrl(match.Value, this.SubPageBaseUrl);
Stream stream = FileUtil.GetPageStream(absUrl, this.SubPageBaseUrl);
if (stream != null)
{
if (this.Upload)
{
Thread thread = new Thread(new ParameterizedThreadStart(UploadFile));
thread.Start(new object[] { stream, saveImagesPath, fileName, "image/jpeg" });
}
else
{
FileUtil.StreamSaveToFile(stream, saveImagesPath + fileName);
}
}
string text1 = string.Empty;
if (!string.IsNullOrEmpty(item.ImageUrlPath))
{
string imageUrlPath = string.Format(item.ImageUrlPath, dtPath.Replace("\\", "/"));
if (!imageUrlPath.EndsWith("/"))
{
imageUrlPath += "/";
}
text1 = value.Replace(match.Value, imageUrlPath + fileName);
}
else
{
text1 = value.Replace(match.Value, saveImagesPath + fileName);
}
value = text1;
}
catch (System.Exception e)
{
LogManager.WriteLog(string.Format("<saveimgerr><img>{0}</img><error>{1}</error></saveimgerr>", match.Value, e.Message));
//throw e;
};
}
}
private bool InlineSniffer()
{
if (!this.Succeed)
{
_done = true;
return(false);
}
//采集项
foreach (SnifferItem item in this.DetailPageConfiguration.SnifferItems)
{
//这里的意思应该是，如果不是多页的内容，则当然页如果大于开始页了，则不再采集了。
if (this.PageIndex > this.DetailPageConfiguration.PageStartIndex && !item.MutiPage)
{
continue;
}
if (item.RegexString != null)
{
MatchCollection matchs = Regex.Matches(this.PageBody, item.RegexString.Expression, (RegexOptions)25);
if (matchs.Count > 0)
{
System.Text.StringBuilder sb = new StringBuilder();
foreach (Match match in matchs)
{
if (matchs.Count > 1 && sb.Length > 0)
{
sb.Append(item.Separator);
}
string value = match.Groups[item.RegexString.ValueGroupIndex].Value;
if (string.IsNullOrEmpty(value))
{
value = item.DefaultValue;
}
else
{
//清理垃圾
ClearRubbish(item, ref value);
//将内容里的URL转成绝对路径
if (item.UrlToAbs)
{
UrlToAbs(item, ref value);
}
//如果采集的是Url则转换成绝对路径
if (item.IsUrl)
{
value = FileUtil.GetAbsUrl(value, this.SubPageBaseUrl);
}
//保存图片
if (item.SaveImage)
{
SaveImages(item, ref value);
}
//清除A元素
if (item.ClearAElement)
{
ClearAElement(item, ref value);
}
//清除HTML代码
if (item.IsClearHTML)
{
value = ClearHTML(value);
}
}
sb.Append(value);
}
this.ResultItems.Add(item.ItemName, sb.ToString().Trim());
}
else
{
this.ResultItems.Add(item.ItemName, item.DefaultValue);
}
}
else
{
this.ResultItems.Add(item.ItemName, item.DefaultValue);
}
}
//整合字段
string text1 = this.PageUrl;
DetailPage parentPage = this.Parent as DetailPage;
if (parentPage != null)
{
foreach (string key in this.ResultItems.Keys)
{
string value = (string)this.ResultItems[key];
if (this.ResultItems.Contains(key))
{
string parentPageValue = (string)parentPage.ResultItems[key];
if (parentPageValue != value)
{
SnifferItem item = null;
foreach (SnifferItem itm in parentPage.DetailPageConfiguration.SnifferItems)
{
if (itm.ItemName == key)
{
item = itm;
break;
}
}
parentPage.ResultItems[key] = parentPageValue + item.MutiPageSeparator + value;
}
}
else
{
parentPage.ResultItems.Add(key, value);
}
}
}
//采集子页
foreach (DetailPageConfiguration conf in this.DetailPageConfiguration.SubPageConfigurations)
{
MatchCollection matches = Regex.Matches(this.PageBody, conf.SnifferSubPageUrlItem.Expression, (RegexOptions)25);
if (matches != null && matches.Count > 0)
{
foreach (Match match in matches)
{
if (!string.IsNullOrEmpty(match.Value))
{
DetailPage detailPage = new DetailPage(this, conf);
detailPage.PageName = match.Groups[conf.SnifferSubPageUrlItem.TitleGroupIndex].Value;
string url = match.Groups[conf.SnifferSubPageUrlItem.UrlGroupIndex].Value;
if (!string.IsNullOrEmpty(conf.SnifferSubPageUrlItem.UrlFormat))
{
url = string.Format(conf.SnifferSubPageUrlItem.UrlFormat, url);
}
detailPage.PageUrl = FileUtil.GetAbsUrl(url, this.SubPageBaseUrl);
detailPage.Sniffer();
this.SubPages.Add(detailPage);
}
}
}
}
_done = true;
return(true);
}
