private Post ScrapePost(HtmlNode contentNode, Post post)
{
post = post ?? new Post();
contentNode.NullCheck();
/// title
HtmlNode titleNode = contentNode.SelectSingleNode(PostPageXPath.Title);
if (titleNode != null)
{
post.Title = titleNode.InnerText.SafeTrimAndEscapeHtml();
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePost, Title node is null - URL: {0}, NODE: {1}", post.Url, contentNode.SerializeHtmlNode());
}
/// subtitle
HtmlNode subtitleNode = contentNode.SelectSingleNode(PostPageXPath.Subtitle);
if (subtitleNode != null)
{
post.Subtitle = subtitleNode.InnerText.SafeTrimAndEscapeHtml();
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePost, Subtitle node is null - URL: {0}, NODE: {1}", post.Url, contentNode.SerializeHtmlNode());
}
/// author
HtmlNode authorNode = contentNode.SelectSingleNode(PostPageXPath.Auhtor);
if (authorNode != null)
{
IList<HtmlNode> authorNameNodes = authorNode.ChildNodes.Where(x => x.Name == "b" && x.ChildNodes.Where(t => t.Name == "a").Count() == 0).ToList();
if (!authorNameNodes.IsEmpty())
{
foreach (HtmlNode author in authorNameNodes)
{
//TODO http://www.rtvslo.si/mmc-priporoca/dame-niso-sposobne-zmagati-na-dirki-formule-ena/306771
User authorUser = new User()
{
Name = author.InnerText.SafeTrim().Replace(",", string.Empty).Replace("foto:", string.Empty).SafeTrimAndEscapeHtml(),
Function = UserFunctionEnum.Journalist
};
post.Authors.Add(authorUser);
}
}
//HtmlNode authorName = authorNode.ChildNodes.FindFirst("b");
//if (authorName != null)
//{
// post.Authors = authorName.InnerText.SafeTrimAndEscapeHtml();
//}
}
if (post.Authors.IsEmpty())
{
//this._logger.WarnFormat("ScrapingService, ScrapePost, Author is empty - URL: {0}, NODE: {1}", post.Url, contentNode.SerializeHtmlNode());
this._logger.WarnFormat("ScrapingService, ScrapePost, Author is empty - URL: {0}", post.Url);
}
/// info
HtmlNode infoNode = contentNode.SelectSingleNode(PostPageXPath.InfoContent);
if (infoNode != null)
{
// <div class="info">16. februar 2013 ob 07:22,<br>zadnji poseg: 16. februar 2013 ob 15:16<br>Schladming - MMC RTV SLO</div>
IList<HtmlNode> textNodes = infoNode.ChildNodes.Where(x => x.Name == "#text").ToList();
if (textNodes != null && textNodes.Count > 1)
{
/// Created datetime
string createdDateTimeString = textNodes.First().InnerText.SafeTrim();
DateTime createdDate;
if (createdDateTimeString.TryParseExactLogging(ParsingHelper.LongDateTimeParseExactPattern, this.cultureInfo, DateTimeStyles.None, out createdDate))
{
post.DateCreated = createdDate.ToUniversalTime();
post.LastUpdated = createdDate.ToUniversalTime();
}
/// Location
string locationString = textNodes.Last().InnerText;
IList<string> locationList = locationString.Split(new string[]{"-"}, StringSplitOptions.RemoveEmptyEntries).ToList();
if (locationList != null && locationList.Count > 1)
{
post.Location = locationList.First().SafeTrim();
if (locationList.Last().SafeTrim() != "MMC RTV SLO")
{
this._logger.DebugFormat("ScrapingService, ScrapePost, InfoNode, Location - URL: {0}, LIST: {1}", post.Url, locationList.SerializeObject());
}
}
else
{
this._logger.WarnFormat("ScrapingService, ScrapePost, InfoNode, Location - URL: {0}, NODE: {1}", post.Url, infoNode.SerializeHtmlNode());
}
if (textNodes.Count == 3)
{
/// Updated datetime
string updatedDateTimeString = textNodes[1].InnerText.SafeTrim();
Regex dateTimeRegex = new Regex(@"(?<date>[0-9\.]+[\w+\s+]+[0-9\:]+)", RegexOptions.IgnoreCase);
//TODO fix regex
Match dateTimeMatch = dateTimeRegex.Match(updatedDateTimeString);
if (dateTimeMatch.Success)
{
updatedDateTimeString = dateTimeMatch.Groups["date"].Value;
DateTime updatedDate;
if (updatedDateTimeString.TryParseExactLogging(ParsingHelper.LongDateTimeParseExactPattern, this.cultureInfo, DateTimeStyles.None, out updatedDate))
{
post.DateCreated = updatedDate.ToUniversalTime();
}
}
}
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePost, InfoNode - URL: {0}, NODE: {1}", post.Url, infoNode.SerializeHtmlNode());
}
}
/// Main content
IList<HtmlNode> contentNodes = new List<HtmlNode>();
foreach (HtmlNode node in contentNode.ChildNodes)
{
/// ends with author
if (node.Name == "div" && node.Attributes.FirstOrDefault(x => x.Value == "author") != null)
{
break;
}
if ((node.Name == "p" || node.Name == "div") && node.FirstChild != null && node.FirstChild.Name != "div" && contentNodes.Count > 0)
{
contentNodes.Add(node);
}
/// starts with p tag
if (node.Name == "p" && node.FirstChild.Name != "div" && contentNodes.Count == 0)
{
contentNodes.Add(node);
}
}
//TODO remove
string sasas = post.Url;
if (!contentNodes.IsEmpty())
{
/// Abstract - text inside strong tag in first node
HtmlNode abstractNode = contentNodes.First();
HtmlNode strongAbstractNode = abstractNode.ChildNodes.First(x => x.Name == "strong");
post.Abstract = strongAbstractNode.InnerText.SafeTrimAndEscapeHtml();
/// remove abstract from main content
abstractNode.ChildNodes.Remove(strongAbstractNode);
/// Content
StringBuilder content = new StringBuilder();
foreach (HtmlNode node in contentNodes)
{
// to get white space after paragraph title
foreach (HtmlNode childNode in node.ChildNodes)
{
string text = childNode.InnerText.SafeTrimAndEscapeHtml();
if (text.Length > 0)
{
content.AppendFormat("{0} ", text);
}
}
}
post.Content = content.ToString().SafeTrim();
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePost - Post content is null - URL: {0}, NODE: {1}", post.Url, contentNode.SerializeHtmlNode());
}
return post;
}
/// <summary>
/// Scrape post statistics
/// Rating, Number of comments, Number of FB likes, Number of tweets
/// </summary>
/// <param name="rootNode"></param>
/// <param name="post"></param>
/// <returns></returns>
private Post ScrapePostStatistics(HtmlNode rootNode, Post post)
{
post = post ?? new Post();
/// rating
HtmlNode ratingNode = rootNode.SelectSingleNode(PostPageXPath.RatingContent);
if (ratingNode != null)
{
string ratingContent = ratingNode.InnerText;
Regex ratingRegex = new Regex(@"\w+\s+(?<rating>[0-9\,]+)\s+\w+\s+(?<numRatings>[0-9]+)", RegexOptions.IgnoreCase);
Match ratingMatch = ratingRegex.Match(ratingContent);
if (ratingMatch.Success)
{
decimal rating;
int numRatings;
if (ratingMatch.Groups["rating"].Value.TryParseLogging(out rating))
{
post.AvgRating = rating;
}
if (ratingMatch.Groups["numRatings"].Value.TryParseLogging(out numRatings))
{
post.NumOfRatings = numRatings;
}
}
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, Rating node is null - URL: {0}, NODE: {1}", post.Url, rootNode.SerializeHtmlNode());
}
/// num of comments
HtmlNode numOfCommentsNode = rootNode.SelectSingleNode(PostPageXPath.NumOfComments);
if (numOfCommentsNode != null &&
!string.IsNullOrEmpty(numOfCommentsNode.InnerText) &&
numOfCommentsNode.InnerText.StartsWith("(") &&
numOfCommentsNode.InnerText.EndsWith(")"))
{
int numOfComments;
string numOfCommentsString = numOfCommentsNode.InnerText.Replace("(", string.Empty).Replace(")", string.Empty);
if (int.TryParse(numOfCommentsString, out numOfComments))
{
post.NumOfComments = numOfComments;
}
else
{
this._logger.WarnFormat("ScrapingService, ScrapePostStatistics, NumOfComments parsing: {2} - URL: {0}, NODE: {1}", post.Url, rootNode.SerializeHtmlNode(), numOfCommentsString);
}
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, NumOfComments - URL: {0}, NODE: {1}", post.Url, rootNode.SerializeHtmlNode());
}
/// FB social plugin
// https://www.facebook.com/plugins/like.php?href=http%3A%2F%2Fwww.rtvslo.si%2Fsport%2Fodbojka%2Fvodebova-in-fabjanova-ubranili-naslov-drzavnih-prvakinj%2F314078&layout=button_count
string fbUrlPattern = "http://www.facebook.com/plugins/like.php?href={0}&layout=button_count";
string encodedUrl = HttpUtility.UrlEncode(post.Url);
string fbUrl = string.Format(fbUrlPattern, encodedUrl);
string fbPluginPage = this.CreateWebRequest(new Uri(fbUrl));
if (!string.IsNullOrEmpty(fbPluginPage))
{
HtmlNode fbRootNode = fbPluginPage.CreateRootNode();
if (fbRootNode != null)
{
int fbLikes;
HtmlNode fbLikesNode = fbRootNode.SelectSingleNode(PostPageXPath.FbLikes);
if (fbLikesNode != null && !string.IsNullOrEmpty(fbLikesNode.InnerText) && int.TryParse(fbLikesNode.InnerText, out fbLikes))
{
post.NumOfFbLikes = fbLikes;
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, FbLikes - POST URL: {0}, FB URL: {1}, NODE: {2}", post.Url, fbUrl, fbRootNode.SerializeHtmlNode());
}
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, FbLikes Node NULL - POST URL: {0}, FB URL: {1}, NODE: {2}", post.Url, fbUrl, fbRootNode.SerializeHtmlNode());
}
}
/// Tweeter social plugin
// http://platform.twitter.com/widgets/tweet_button.1375828408.html?url=http%3A%2F%2Fwww.rtvslo.si%2Fzabava%2Fiz-sveta-znanih%2Fboy-george-napadel-isinbajevo-zaradi-homofobnih-izjav%2F315495
// http://cdn.api.twitter.com/1/urls/count.json?url=http%3A%2F%2Fwww.rtvslo.si%2Fzabava%2Fiz-sveta-znanih%2Fboy-george-napadel-isinbajevo-zaradi-homofobnih-izjav%2F315495&callback=twttr.receiveCount
string twUrlPattern = "http://cdn.api.twitter.com/1/urls/count.json?url={0}";
string twUrl = string.Format(twUrlPattern, encodedUrl);
string twJsonPage = this.CreateWebRequest(new Uri(twUrl));
try
{
JObject twJson = JObject.Parse(twJsonPage);
string countString = (string)twJson["count"];
int numOfTweets;
if (!string.IsNullOrEmpty(countString) && int.TryParse(countString, out numOfTweets))
{
post.NumOfTweets = numOfTweets;
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, NumOfTweets - POST URL: {0}, TW URL: {1}, NODE: {2}", post.Url, twUrl, twJsonPage);
}
}
catch (JsonReaderException ex)
{
this._logger.ErrorFormat("ScrapingService, ScrapePostStatistics, NumOfTweets Parse EXCEPTION - POST URL: {0}, TW URL: {1}, NODE: {2}, EX:{3}", post.Url, twUrl, twJsonPage, ex.Message);
}
return post;
}
private Comment ScrapeComment(HtmlNode commentNode)
{
Comment comment = new Comment()
{
AccessedDate = DateTime.UtcNow.ToUniversalTime(),
};
commentNode.NullCheck();
HtmlNode headerNode = commentNode.SelectSingleNode(CommentsPageXPath.HeaderInfo);
IList<HtmlNode> innerHeaderNodes = headerNode.ChildNodes.Where(x => x.Name == "a").ToList();
/// userUrl, url, username, id
if (innerHeaderNodes != null && innerHeaderNodes.Count == 2)
{
if (innerHeaderNodes[0].Attributes["href"] != null)
{
comment.UserUrl = innerHeaderNodes[0].Attributes["href"].Value.SafeTrim().ToFullRtvSloUrl();
comment.UserId = this.GetIdStringFromUrl(comment.UserUrl);
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapeComment - User url is null - NODE: {0}", commentNode.SerializeHtmlNode());
}
comment.UserName = innerHeaderNodes[0].InnerText.SafeTrim();
if (innerHeaderNodes[1].Attributes["href"] != null)
{
comment.Url = innerHeaderNodes[1].Attributes["href"].Value.SafeTrim().ToFullRtvSloUrl();
comment.Id = this.GetIdFromUrl(comment.Url);
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapeComment - Comment url is null - NODE: {0}", commentNode.SerializeHtmlNode());
}
}
/// created date time
string dateCreatedString = headerNode.LastChild.InnerText.SafeTrim();
DateTime created;
if (dateCreatedString.TryParseExactLogging(ParsingHelper.ShortDateTimeParseExactPattern, this.cultureInfo, DateTimeStyles.None, out created))
{
comment.DateCreated = created.ToUniversalTime();
}
HtmlNode contentNode = commentNode.SelectSingleNode(CommentsPageXPath.Content);
if (contentNode != null)
{
string content = contentNode.InnerText.SafeTrimAndEscapeHtml();
comment.Content = content;
}
else
{
this._logger.ErrorFormat("ScrapingService, ScrapeComment - Comment content is null - URL: {0}", comment.Url);
}
/// rating
HtmlNode ratingNode = commentNode.SelectSingleNode(CommentsPageXPath.Rating);
string plusRatingString = ratingNode.SelectSingleNode(CommentsPageXPath.PlusRating).InnerText.SafeTrim();
string minusRatingString = ratingNode.SelectSingleNode(CommentsPageXPath.MinusRating).InnerText.SafeTrim();
int plusRating = this.ScrapeCommentRating(plusRatingString, comment.Url);
int minusRating = this.ScrapeCommentRating(minusRatingString, comment.Url);
comment.Rating = plusRating + minusRating;
return comment;
}
