public override IEnumerable<WarningOccurrenceDto> Analyze(ISqlRepository sqlRepository, UrlDto urlDto)
{
var warningOccurrencesList = new List<WarningOccurrenceDto>();
if (urlDto.Status != HttpStatusCode.OK)
{
return warningOccurrencesList;
}
foreach (var link in sqlRepository.GetWebPageLinks(urlDto.UrlId.Value))
{
if (link.Status == HttpStatusCode.NotFound)
{
IWarning warning = GetWarning(typeof(BrokenLinkWarning));
warningOccurrencesList.Add(new WarningOccurrenceDto()
{
UrlId = (long)urlDto.UrlId,
Message = String.Format(warning.Description, link.Url),
WarningType = warning
}
);
}
}
return warningOccurrencesList;
}
public override IEnumerable<WarningOccurrenceDto> Analyze(ISqlRepository sqlRepository, UrlDto urlDto)
{
var warningOccurrencesList = new List<WarningOccurrenceDto>();
if (urlDto.Contents == null)
{
_log.Debug("urlDto is empty for url:" + urlDto.Url);
return warningOccurrencesList;
}
if (urlDto.Status != HttpStatusCode.OK)
{
return warningOccurrencesList;
}
var title = urlDto.Contents.SingleOrDefault(element => element.ContentType == ContentTypesEnum.Title);
_log.Debug(String.Format("Analyze title '{0}' for url '{1}'",
(title == null || title.Element == null) ? "" : title.Element,
urlDto.Url));
if (title == null || String.IsNullOrEmpty(title.Element))
{
IWarning warning = GetWarning(typeof(EmptyTitleWarning));
warningOccurrencesList.Add(new WarningOccurrenceDto()
{
UrlId = (long) urlDto.UrlId,
Message = warning.Description,
WarningType = warning
}
);
}
else
{
var elements = sqlRepository.GetContentForElementText(ContentTypesEnum.Title, title.Element);
if (elements.Any())
{
ContentDto targetContentDto = elements.FirstOrDefault(x => x.UrlId != urlDto.UrlId);
if (targetContentDto != null)
{
UrlDto targetUrlDto;
sqlRepository.FindUrl(targetContentDto.UrlId, out targetUrlDto);
IWarning warning = GetWarning(typeof(DuplicatedTitleWarning));
warningOccurrencesList.Add(new WarningOccurrenceDto()
{
UrlId = (long)urlDto.UrlId,
Message = String.Format(warning.Description, title.Element, targetUrlDto.Url),
WarningType = warning
}
);
}
}
}
return warningOccurrencesList;
}
public void ProcessPage(ISqlRepository sqlRepository, UrlDto urlDto, Report report)
{
foreach (var rule in _rules)
{
var warningOccurrences = rule.Analyze(sqlRepository, urlDto);
report.WarningOccurrences.AddRange(warningOccurrences);
}
}
internal bool FindUrl(long urlId, out UrlDto urlDto)
{
urlDto = new UrlDto();
bool success = false;
string selectSql = "SELECT * FROM Urls where url_id='" + urlId + "'";
using (var selectCommand = new SQLiteCommand(selectSql, _sqliteCon))
{
using (var dataReader = selectCommand.ExecuteReader())
{
if (dataReader.Read())
{
MapDataReaderToUrlDto(dataReader, out urlDto);
success = true;
}
}
}
return success;
}
internal ContentDto[] GetContentsForUrlDto(UrlDto urlDto)
{
var contents = new List<ContentDto>();
var selectSql = "SELECT * FROM Contents where url_id='" + urlDto.UrlId + "'";
using (var selectCommand = new SQLiteCommand(selectSql, _sqliteCon))
{
using (var dataReader = selectCommand.ExecuteReader())
{
if (dataReader.Read())
{
ContentDto contentDto;
MapDataReaderToContentDto(dataReader, out contentDto);
contents.Add(contentDto);
}
}
}
return contents.ToArray();
}
static void AddNewPage(CrawledPage crawledPage)
{
if (crawledPage.Uri.AbsoluteUri.StartsWith(_uriToCrawl.AbsoluteUri) == false)
return;
lock (_sqlRepository)
{
UrlDto urlDto;
if (_sqlRepository.FindUrl(crawledPage.Uri.AbsoluteUri, out urlDto))
{
// For pages that do not repond is empty
if (crawledPage.HttpWebResponse != null)
{
urlDto.Status = crawledPage.HttpWebResponse.StatusCode;
// This can be text/html or text/html;charset=utf-8
if (crawledPage.HttpWebResponse.ContentType.Contains("text/html"))
{
urlDto.IsWebPage = true;
}
}
}
else
{
urlDto = new UrlDto
{
Url = crawledPage.Uri.AbsoluteUri,
Status = crawledPage.HttpWebResponse.StatusCode,
IsWebPage = true
};
}
var webPageParsing = new WebPageParsing(crawledPage);
var list = GetLinks(webPageParsing, urlDto);
urlDto.Links = list.ToArray();
urlDto.Contents = GetContent(webPageParsing);
_sqlRepository.InsertOrUpdateUrl(urlDto);
}
}
static UrlDto InsertUrlDtoWithOneStatusCodeOKLink(ISqlRepository repository, HttpStatusCode httpStatusCode)
{
var urlDtoLink = new UrlDto
{
Status = httpStatusCode
};
repository.InsertOrUpdateUrl(urlDtoLink);
var urlDto = new UrlDto
{
Status = HttpStatusCode.OK,
Links = new List<LinkDto>()
{
new LinkDto()
{
TargetUrlId = (long) urlDtoLink.UrlId,
}
}.ToArray()
};
return repository.InsertOrUpdateUrl(urlDto);
}
public abstract IEnumerable<WarningOccurrenceDto> Analyze(ISqlRepository sqlRepository, UrlDto urlDto);
private static void MapDataReaderToUrlDto(SQLiteDataReader dataReader, out UrlDto urlDto)
{
urlDto = new UrlDto();
urlDto.UrlId = dataReader.GetInt64((int)UrlsField.UrlId);
urlDto.Url = dataReader.GetString((int)UrlsField.Url);
urlDto.Url = HttpUtility.UrlDecode(urlDto.Url);
urlDto.Status = (HttpStatusCode) dataReader.GetInt32((int)UrlsField.Status);
urlDto.IsWebPage = dataReader.GetInt32((int)UrlsField.IsWebPage) > 0;
}
internal UrlDto InsertOrUpdateUrl(UrlDto urlDto)
{
string sqlStatement;
string urlEncoded = HttpUtility.UrlEncode(urlDto.Url);
if (urlDto.UrlId.HasValue)
{
sqlStatement = String.Format("INSERT OR REPLACE INTO Urls (url_id, url, status, is_webpage) VALUES ('{0}', '{1}', '{2}', '{3}')",
urlDto.UrlId, urlEncoded, urlDto.StatusRaw, urlDto.IsWebPage ? 1 : 0);
ExecuteSqlCommand(sqlStatement);
}
else
{
sqlStatement = String.Format("INSERT INTO Urls (url, status, is_webpage) VALUES ('{0}', '{1}', '{2}')",
urlEncoded, urlDto.StatusRaw, urlDto.IsWebPage ? 1 : 0);
ExecuteSqlCommand(sqlStatement);
urlDto.UrlId = GetLastInsertedRowId();
}
if (urlDto.Contents != null)
{
foreach (var content in urlDto.Contents)
{
content.UrlId = urlDto.UrlId.HasValue ? urlDto.UrlId.Value : 0;
_contentDao.InsertOrUpdateContent(content);
}
}
if (urlDto.Links != null)
{
foreach (var link in urlDto.Links)
{
link.SourceUrlId = urlDto.UrlId.HasValue ? urlDto.UrlId.Value : 0;
_linkDao.AddLink(link);
}
}
return urlDto;
}
private static List<LinkDto> GetLinks(WebPageParsing webPageParsing, UrlDto urlDto)
{
var links = webPageParsing.GetLinks();
var list = new List<LinkDto>();
foreach (Uri link in links)
{
UrlDto linkUrlDto;
if (_sqlRepository.FindUrl(link.AbsoluteUri, out linkUrlDto) == false)
{
linkUrlDto = new UrlDto {Url = link.AbsoluteUri};
linkUrlDto = _sqlRepository.InsertOrUpdateUrl(linkUrlDto);
}
list.Add(new LinkDto()
{
TargetUrlId = linkUrlDto.UrlId.HasValue ? linkUrlDto.UrlId.Value : 0
});
}
return list;
}
public bool FindUrl(long urlId, out UrlDto urlDto)
{
return _urlDao.FindUrl(urlId, out urlDto);
}
public bool FindUrl(string url, out UrlDto urlDto)
{
return _urlDao.FindUrl(url, out urlDto);
}
public UrlDto InsertOrUpdateUrl(UrlDto urlDto)
{
return _urlDao.InsertOrUpdateUrl(urlDto);
}
static UrlDto GetValidTitle(string title)
{
var urlDto = new UrlDto
{
Status = HttpStatusCode.OK,
Contents = new List<ContentDto>()
{
new ContentDto()
{
ContentType = ContentTypesEnum.Title,
Element = title
}
}.ToArray()
};
return urlDto;
}