private static void GetSiteLinks(RssSource rs, string strContent)
{
string strUrlRule = rs.strArticleUrlPattern;
strUrlRule = strUrlRule.Replace(".", "\\.");
strUrlRule = strUrlRule.Replace("*", ".*?");
HtmlAgilityPack.HtmlDocument htmlDoc = GetHtmlDocument(strContent);
if (rs.strArticleUrlRangeCssPath != "")
{
IEnumerable <HtmlNode> NodesUrlContent = htmlDoc.DocumentNode.QuerySelectorAll(rs.strArticleUrlRangeCssPath);
if (NodesUrlContent.Count() > 0)
{
string strReturnPage = NodesUrlContent.ToArray()[0].InnerHtml;//进一步缩小范围
htmlDoc = GetHtmlDocument(strReturnPage);
}
}
string baseUrl = GetUrlLeftPart(rs.strSiteUrl);
DocumentWithLinks links = htmlDoc.GetLinks();
foreach (string link in links.Links.Union(links.References))
{
if (string.IsNullOrEmpty(link))
{
continue;
}
string decodedLink = link;
string normalizedLink = GetNormalizedLink(baseUrl, decodedLink);
if (string.IsNullOrEmpty(normalizedLink))
{
continue;
}
MatchCollection matchs = Regex.Matches(normalizedLink, strUrlRule, RegexOptions.Singleline);
if (matchs.Count > 0)
{
string strLinkText = "";
foreach (string strTemp in links.m_dicLink2Text.Keys)
{
if (strTemp.Contains(normalizedLink))
{
strLinkText = links.m_dicLink2Text[strTemp];
break;
}
}
if (strLinkText == "")
{
if (links.m_dicLink2Text.Keys.Contains(link))
{
strLinkText = links.m_dicLink2Text[link].TrimEnd().TrimStart();
}
if (links.m_dicLink2Text.Keys.Contains(link.ToLower()))
{
strLinkText = links.m_dicLink2Text[link.ToLower()].TrimEnd().TrimStart();
}
}
SaveUrlToDB(normalizedLink, strLinkText);
Console.WriteLine(normalizedLink);
}
}
return;
}
public void Process(Crawler crawler, PropertyBag propertyBag)
{
AspectF.Define.
NotNull(crawler, "crawler").
NotNull(propertyBag, "propertyBag");
string stepUri = Uri.UnescapeDataString(propertyBag.Step.Uri.AbsoluteUri);
if (stepUri.Length > 396)
{
stepUri = stepUri.Substring(0, 396);
}
var crawlHistory = AspectF.Define.
Return<CrawlHistory, NCrawlerEntitiesDbServices>(
e => e.CrawlHistory.Where(m => m.Key == stepUri).FirstOrDefault());
if (crawlHistory == null)
{
AspectF.Define.Do<NCrawlerEntitiesDbServices>(e =>
{
e.ExecuteStoreCommand("delete Crawlqueue where [key] ={0}", stepUri);
});
return;
}
try
{
if (propertyBag.StatusCode != HttpStatusCode.OK)
{
AspectF.Define.Do<NCrawlerEntitiesDbServices>(e =>
{
e.ExecuteStoreCommand("delete Crawlqueue where [key] ={0}", crawlHistory.Key);
//CrawlQueue result = e.CrawlQueue.FirstOrDefault(q => q.Key == crawlHistory.Key);
//if (!result.IsNull())
//{
// e.DeleteObject(result);
// e.SaveChanges();
//}
});
return;
}
if (!IsHtmlContent(propertyBag.ContentType))
{
AspectF.Define.Do<NCrawlerEntitiesDbServices>(e =>
{
e.ExecuteStoreCommand("delete Crawlqueue where [key] ={0}", crawlHistory.Key);
//CrawlQueue result = e.CrawlQueue.FirstOrDefault(q => q.Key == crawlHistory.Key);
//if (!result.IsNull())
//{
// e.DeleteObject(result);
// e.SaveChanges();
//}
});
return;
}
HtmlDocument htmlDoc = new HtmlDocument
{
OptionAddDebuggingAttributes = false,
OptionAutoCloseOnEnd = true,
OptionFixNestedTags = true,
OptionReadEncoding = true
};
using (Stream reader = propertyBag.GetResponse())
{
Encoding documentEncoding = htmlDoc.DetectEncoding(reader);
reader.Seek(0, SeekOrigin.Begin);
if (!documentEncoding.IsNull())
{
htmlDoc.Load(reader, documentEncoding, true);
}
else
{
htmlDoc.Load(reader, true);
}
//string content = reader.ReadToEnd();
//resultHtmlContent = content;
}
//string steplUri = propertyBag.ResponseUri.OriginalString;
string orginalHtmlContent = htmlDoc.DocumentNode.OuterHtml;
string baseUrl = propertyBag.ResponseUri.GetLeftPart(UriPartial.Path);
DocumentWithLinks links = htmlDoc.GetLinks();
//string urlRegex = @"^http://www.bbc.co.uk/food/recipes/[^#/]+$";
List<string> recipeRegex = null;
var jsonStr = cache.Get(AppDomain.CurrentDomain.BaseDirectory + "OriginalWebSite") as string;
if (jsonStr == null)
{
using (var stream = new StreamReader(AppDomain.CurrentDomain.BaseDirectory + "OriginalWebSite.txt", Encoding.UTF8))
{
jsonStr = stream.ReadToEnd();
var policy = new CacheItemPolicy();
policy.Priority = CacheItemPriority.NotRemovable;
policy.AbsoluteExpiration = DateTimeOffset.Now.AddDays(1);
cache.Set(AppDomain.CurrentDomain.BaseDirectory + "OriginalWebSite", jsonStr, policy);
Console.WriteLine("cache --" + AppDomain.CurrentDomain.BaseDirectory + " :" + cache.Get(AppDomain.CurrentDomain.BaseDirectory + "OriginalWebSite"));
}
}
var json = JsonConvert.DeserializeObject<OriginalWebSiteTxt>(jsonStr);
if (json.RecipeRegex != null && json.RecipeRegex.Count > 0)
{
recipeRegex = json.RecipeRegex;
}
bool needToStore = false;
if (recipeRegex != null)
{
foreach (var regex in recipeRegex)
{
if (Regex.IsMatch(propertyBag.Step.Uri.AbsoluteUri, regex, RegexOptions.IgnoreCase))
{
needToStore = true;
break;
}
}
}
else
{
needToStore = true;
}
if (needToStore)
{
//string folderPath = "D:/CrawlerManager/CrawlerData";
//string instanceFolderPath = folderPath + "/" + crawlHistory.GroupId;
//string path = folderPath + "/" + crawlHistory.GroupId + "/" + string.Format("{0}.txt", crawlHistory.Id);
//if (!Directory.Exists(folderPath))
//{
// Directory.CreateDirectory(folderPath);
//}
//if (!Directory.Exists(instanceFolderPath))
//{
// Directory.CreateDirectory(instanceFolderPath);
//}
//if (!File.Exists(path))
//{
// try
// {
// using (StreamWriter sw = File.CreateText(path))
// {
// sw.WriteLine(orginalHtmlContent);
// }
// }
// catch (Exception ex)
// {
// log4net.Config.XmlConfigurator.Configure();
// log4net.ILog log = log4net.LogManager.GetLogger("logger-name");
// log.Error(ex);
// }
//}
var folderHelper = new FolderHelper();
var path = folderHelper.GetFolderPathToStore(crawlHistory.GroupId) + "/" + string.Format("{0}.txt", crawlHistory.Id);
Console.Write(path);
if (!File.Exists(path))
{
try
{
using (StreamWriter sw = File.CreateText(path))
{
sw.WriteLine(orginalHtmlContent);
}
}
catch (Exception ex)
{
log4net.Config.XmlConfigurator.Configure();
log4net.ILog log = log4net.LogManager.GetLogger("logger-name");
log.Error(ex);
}
}
//}
}
AspectF.Define.Do<NCrawlerEntitiesDbServices>(e =>
{
e.ExecuteStoreCommand("delete Crawlqueue where [key] ={0}", crawlHistory.Key);
});
foreach (string link in links.Links.Union(links.References))
{
if (link.IsNullOrEmpty() || link.Length > 396)
{
continue;
}
string decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(link);
string normalizedLink = "";
try
{
normalizedLink = NormalizeLink(baseUrl, decodedLink);
}
catch (Exception ex)
{
continue;
}
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
if (link.Contains("page="))
{
var a = 1;
}
crawler.AddStep(new Uri(normalizedLink), propertyBag.Step.Depth + 1,
propertyBag.Step, new Dictionary<string, object>
{
{Resources.PropertyBagKeyOriginalUrl, link},
{Resources.PropertyBagKeyOriginalReferrerUrl, propertyBag.ResponseUri}
});
}
}
catch (Exception ex)
{
AspectF.Define.Do<NCrawlerEntitiesDbServices>(e =>
{
e.ExecuteStoreCommand("delete Crawlqueue where [key] ={0}", crawlHistory.Key);
});
log4net.Config.XmlConfigurator.Configure();
log4net.ILog log = log4net.LogManager.GetLogger("logger-name");
log.Error(ex);
}
}
public void Process(Crawler crawler, PropertyBag propertyBag)
{
AspectF.Define.
NotNull(crawler, "crawler").
NotNull(propertyBag, "propertyBag");
if (propertyBag.StatusCode != HttpStatusCode.OK)
{
return;
}
if (!IsHtmlContent(propertyBag.ContentType))
{
return;
}
HtmlDocument htmlDoc = new HtmlDocument
{
OptionAddDebuggingAttributes = false,
OptionAutoCloseOnEnd = true,
OptionFixNestedTags = true,
OptionReadEncoding = true
};
using (Stream reader = propertyBag.GetResponse())
{
Encoding documentEncoding = htmlDoc.DetectEncoding(reader);
reader.Seek(0, SeekOrigin.Begin);
if (!documentEncoding.IsNull())
{
htmlDoc.Load(reader, documentEncoding, true);
}
else
{
htmlDoc.Load(reader, true);
}
}
string originalContent = htmlDoc.DocumentNode.OuterHtml;
if (HasTextStripRules || HasSubstitutionRules)
{
string content = StripText(originalContent);
content = Substitute(content, propertyBag.Step);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
propertyBag["HtmlDoc"].Value = htmlDoc;
HtmlNodeCollection nodes = htmlDoc.DocumentNode.SelectNodes("//title");
// Extract Title
if (!nodes.IsNull())
{
propertyBag.Title = string.Join(";", nodes.
Select(n => n.InnerText).
ToArray()).Trim();
}
// Extract Meta Data
nodes = htmlDoc.DocumentNode.SelectNodes("//meta[@content and @name]");
if (!nodes.IsNull())
{
propertyBag["Meta"].Value = (
from entry in nodes
let name = entry.Attributes["name"]
let content = entry.Attributes["content"]
where !name.IsNull() && !name.Value.IsNullOrEmpty() && !content.IsNull() && !content.Value.IsNullOrEmpty()
select name.Value + ": " + content.Value).ToArray();
}
propertyBag.Text = htmlDoc.ExtractText().Trim();
if (HasLinkStripRules || HasTextStripRules)
{
string content = StripLinks(originalContent);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
// Extract Links
DocumentWithLinks links = htmlDoc.GetLinks();
foreach (string link in links.Links.Union(links.References))
{
if (link.IsNullOrEmpty())
{
continue;
}
string baseUrl = propertyBag.ResponseUri.GetLeftPart(UriPartial.Path);
string decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(link);
string normalizedLink = NormalizeLink(baseUrl, decodedLink);
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
crawler.AddStep(new Uri(normalizedLink), propertyBag.Step.Depth + 1,
propertyBag.Step, new Dictionary<string, object>
{
{Resources.PropertyBagKeyOriginalUrl, link},
{Resources.PropertyBagKeyOriginalReferrerUrl, propertyBag.ResponseUri}
});
}
}
public Task<bool> Process(ICrawler crawler, PropertyBag propertyBag)
{
AspectF.Define
.NotNull(crawler, nameof(crawler))
.NotNull(propertyBag, nameof(propertyBag));
if (propertyBag.StatusCode != HttpStatusCode.OK)
{
return Task.FromResult(true);
}
if (!IsHtmlContent(propertyBag.ContentType))
{
return Task.FromResult(true);
}
HtmlDocument htmlDoc = new HtmlDocument
{
OptionAddDebuggingAttributes = false,
OptionAutoCloseOnEnd = true,
OptionFixNestedTags = true,
OptionReadEncoding = true
};
using (MemoryStream ms = new MemoryStream(propertyBag.Response))
{
Encoding documentEncoding = htmlDoc.DetectEncoding(ms);
ms.Seek(0, SeekOrigin.Begin);
if (!documentEncoding.IsNull())
{
htmlDoc.Load(ms, documentEncoding, true);
}
else
{
htmlDoc.Load(ms, true);
}
}
string originalContent = htmlDoc.DocumentNode.OuterHtml;
if (HasTextStripRules || HasSubstitutionRules)
{
string content = StripText(originalContent);
content = Substitute(content, propertyBag.Step);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
propertyBag["HtmlDoc"].Value = htmlDoc;
HtmlNodeCollection nodes = htmlDoc.DocumentNode.SelectNodes("//title");
// Extract Title
if (!nodes.IsNull())
{
propertyBag.Title = string.Join(";", nodes.
Select(n => n.InnerText).
ToArray()).Trim();
}
// Extract Meta Data
nodes = htmlDoc.DocumentNode.SelectNodes("//meta[@content and @name]");
if (!nodes.IsNull())
{
propertyBag["Meta"].Value = (
from entry in nodes
let name = entry.Attributes["name"]
let content = entry.Attributes["content"]
where !name.IsNull() && !name.Value.IsNullOrEmpty() && !content.IsNull() && !content.Value.IsNullOrEmpty()
select $"{name.Value}: {content.Value}").ToArray();
}
// Extract text
propertyBag.Text = htmlDoc.ExtractText().Trim();
if (HasLinkStripRules || HasTextStripRules)
{
string content = StripLinks(originalContent);
using (TextReader tr = new StringReader(content))
{
htmlDoc.Load(tr);
}
}
string baseUrl = propertyBag.ResponseUri.GetLeftPart(UriPartial.Path);
// Extract Head Base
nodes = htmlDoc.DocumentNode.SelectNodes("//head/base[@href]");
if (!nodes.IsNull())
{
baseUrl = nodes
.Select(entry => new {entry, href = entry.Attributes["href"]})
.Where(arg => !arg.href.IsNull()
&& !arg.href.Value.IsNullOrEmpty()
&& Uri.IsWellFormedUriString(arg.href.Value, UriKind.RelativeOrAbsolute))
.Select(t =>
{
if (Uri.IsWellFormedUriString(t.href.Value, UriKind.Relative))
{
return propertyBag.ResponseUri.GetComponents(UriComponents.SchemeAndServer, UriFormat.Unescaped) + t.href.Value;
}
return t.href.Value;
})
.AddToEnd(baseUrl)
.FirstOrDefault();
}
// Extract Links
DocumentWithLinks links = htmlDoc.GetLinks();
foreach (string link in links.Links.Union(links.References))
{
if (link.IsNullOrEmpty())
{
continue;
}
string decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(link);
string normalizedLink = NormalizeLink(baseUrl, decodedLink);
if (normalizedLink.IsNullOrEmpty())
{
continue;
}
crawler.Crawl(new Uri(normalizedLink), propertyBag);
}
return Task.FromResult(true);
}
protected bool SaveUrlToDB(string strVisitUrl, string strReturnPage, DoWorkEventArgs e)
{
try
{
string strUrlFilterRule = GetUrlFilterRule();
HtmlAgilityPack.HtmlDocument htmlDoc = GetHtmlDocument(strReturnPage);
string strUrlCss = GetUrlCss();
if (strUrlCss != "")
{
IEnumerable <HtmlNode> NodesUrlContent = htmlDoc.DocumentNode.QuerySelectorAll(strUrlCss);
if (NodesUrlContent.Count() > 0)
{
strReturnPage = NodesUrlContent.ToArray()[0].InnerHtml;//进一步缩小范围
htmlDoc = GetHtmlDocument(strReturnPage);
}
}
string baseUrl = new Uri(strVisitUrl).GetLeftPart(UriPartial.Authority);
DocumentWithLinks links = htmlDoc.GetLinks();
foreach (string link in links.Links.Union(links.References))
{
if (string.IsNullOrEmpty(link))
{
continue;
}
string decodedLink = link;
string normalizedLink = GetNormalizedLink(baseUrl, decodedLink);
if (string.IsNullOrEmpty(normalizedLink))
{
continue;
}
MatchCollection matchs = Regex.Matches(normalizedLink, strUrlFilterRule, RegexOptions.Singleline);
if (matchs.Count > 0)
{
string strLinkText = "";
if (links.m_dicLink2Text.Keys.Contains(normalizedLink))
{
strLinkText = links.m_dicLink2Text[normalizedLink];
}
if (strLinkText == "")
{
if (links.m_dicLink2Text.Keys.Contains(link))
{
strLinkText = links.m_dicLink2Text[link].TrimEnd().TrimStart();
}
}
lstThisTimesUrls.Add(normalizedLink);
if (!m_dicUrl2Title.Keys.Contains(normalizedLink))
{
m_dicUrl2Title.Add(normalizedLink, strLinkText);
}
m_wd.AddUrlQueue(normalizedLink);
}
}
}
catch (Exception ex)
{
}
bool bNoArticle = CheckArticles(lstThisTimesUrls);
return(bNoArticle);
}
protected static void SaveUrlToDB(string strReturnPage, SpiderTemplate s, List <DetailLink> list)
{
Dictionary <string, string> m_dicLink2Text = new Dictionary <string, string>();
string strUrlFilterRule = s.ListPattern;
//strUrlFilterRule = ParseUrl(strUrlFilterRule);
HtmlAgilityPack.HtmlDocument htmlDoc = GetHtmlDocument(strReturnPage);
// string baseUrl = new Uri(strVisitUrl).GetLeftPart(UriPartial.Authority);
string baseUrl = GetUrlLeftPart(s.ListUrl);
DocumentWithLinks links = htmlDoc.GetLinks();
bool bNoArticle = true;
List <string> lstRevomeSame = new List <string>();
// int nCountPerPage = 0;
// bool bExistFind = false;
// List<string> lstNeedDownLoad = new List<string>();
foreach (string link in links.Links.Union(links.References))
{
if (string.IsNullOrEmpty(link))
{
continue;
}
//string decodedLink = ExtendedHtmlUtility.HtmlEntityDecode(link);
string decodedLink = link;
//if (decodedLink != link)
//{
// int a = 1;
//}
//Console.WriteLine(decodedLink);
string normalizedLink = GetNormalizedLink(baseUrl, decodedLink);
//Console.WriteLine(normalizedLink);
if (string.IsNullOrEmpty(normalizedLink))
{
continue;
}
MatchCollection matchs = Regex.Matches(normalizedLink, strUrlFilterRule, RegexOptions.Singleline);
if (matchs.Count > 0)
{
string strLinkText = "";
foreach (string strTemp in links.m_dicLink2Text.Keys)
{
if (strTemp.Contains(normalizedLink))
{
strLinkText = links.m_dicLink2Text[strTemp];
break;
}
}
//if (links.m_dicLink2Text.Keys.Contains(normalizedLink))
// strLinkText = links.m_dicLink2Text[normalizedLink];
if (strLinkText == "")
{
if (links.m_dicLink2Text.Keys.Contains(link))
{
strLinkText = links.m_dicLink2Text[link].TrimEnd().TrimStart();
}
if (links.m_dicLink2Text.Keys.Contains(link.ToLower()))
{
strLinkText = links.m_dicLink2Text[link.ToLower()].TrimEnd().TrimStart();
}
}
if (lstRevomeSame.Contains(normalizedLink))
{
continue;
}
else
{
lstRevomeSame.Add(normalizedLink);
}
//bool bRet = AddLayerNodeToSaveUrlToDB(m_strWholeDbName, normalizedLink, ref strLinkText);
DetailLink lnk = new DetailLink();
lnk.Template = s;
lnk.Url = normalizedLink;
lnk.Title = strLinkText;
list.Add(lnk);
}
//Console.WriteLine(" uri is " + normalizedLink.ToString());
}
return;
}
