private readonly ConcurrentDictionary<string, CrawlUrl> working = new ConcurrentDictionary<string, CrawlUrl>(); // {hash, url}
#endregion Fields
#region Methods
public void Done(string key, CrawlUrl crawlUrl)
{
if (done.TryAdd(key, ""))
{
scheduled.TryRemove(key, out crawlUrl);
}
}
public void Done(string key, CrawlUrl crawlUrl)
{
if (done.TryAdd(key, ""))
{
scheduled.TryRemove(key, out crawlUrl);
}
}
public void Done(string key, CrawlUrl crawlUrl)
{
var typedClient = Client.As<CrawlUrl>();
if (typedClient.RemoveEntryFromHash(typedClient.GetHash<string>(WorkingHashId), crawlUrl.Hash))
{
typedClient.SetEntryInHashIfNotExists(typedClient.GetHash<string>(DoneHashId), crawlUrl.Hash, crawlUrl);
}
}
private void RaisePageProcessing(CrawlUrl crawlUrl)
{
var handler = PageProcessing;
if (handler != null)
{
handler.Invoke(crawlUrl);
}
}
private void RaisePageScheduled(CrawlUrl crawlUrl)
{
var handler = PageScheduled;
if (handler != null)
{
handler.Invoke(crawlUrl);
}
}
public CrawlUrl PeekNext()
{
CrawlUrl next = null;
if (scheduledQueue.TryPeek(out next))
{
return(next);
}
LoadMore();
scheduledQueue.TryPeek(out next);
return(next);
}
public int Schedule(IEnumerable <string> crawlUrls)
{
var hashes = crawlUrls.Select(x => x.Split('#')[0].TrimEnd('/')).Where(x => !string.IsNullOrWhiteSpace(x)).Distinct().ToDictionary(urlHasher.CalculateHashAsString);
var scheduledLinksCount = 0;
foreach (var hash in hashes)
{
if (crawlUrlRepository.IsKnown(hash.Key))
{
continue;
}
try
{
var crawlUrl = new CrawlUrl
{
Hash = hash.Key,
Url = hash.Value,
};
var websiteDefinition = websiteDefinitions.FirstOrDefault(x => x.Website.IsRelativeUrl(crawlUrl));
if (websiteDefinition != null)
{
crawlUrl.WebsiteDefinition = websiteDefinition;
if (crawlUrlRepository.TryAdd(hash.Key, crawlUrl))
{
Interlocked.Increment(ref websiteDefinition.UrlsToProcessCount);
Interlocked.Increment(ref scheduledLinksCount);
/*
* var inputCount = websiteProcessingDefinitions[websiteDefinition].Post(crawlUrl);
* log.DebugFormat("Process block has {0} pending messages", inputCount);
*
*/
RaisePageScheduled(crawlUrl);
}
}
}
catch (Exception ex)
{
log.Error(ex);
}
}
return(scheduledLinksCount);
}
public bool TryAdd(string key, CrawlUrl crawlUrl)
{
return(scheduled.TryAdd(key, crawlUrl));
}
public int Post(CrawlUrl crawlUrl)
{
ProcessingBlock.Post(crawlUrl);
return(ProcessingBlock.InputCount);
}
public bool TryAdd(string key, CrawlUrl crawlUrl)
{
return scheduled.TryAdd(key, crawlUrl);
}
private void RaisePageScheduled(CrawlUrl crawlUrl)
{
var handler = PageScheduled;
if (handler != null)
handler.Invoke(crawlUrl);
}
public int Schedule(IEnumerable<string> crawlUrls)
{
var hashes = crawlUrls.Select(x => x.Split('#')[0].TrimEnd('/')).Where(x => !string.IsNullOrWhiteSpace(x)).Distinct().ToDictionary(urlHasher.CalculateHashAsString);
var scheduledLinksCount = 0;
foreach (var hash in hashes)
{
if (crawlUrlRepository.IsKnown(hash.Key))
continue;
try
{
var crawlUrl = new CrawlUrl
{
Hash = hash.Key,
Url = hash.Value,
};
var websiteDefinition = websiteDefinitions.FirstOrDefault(x => x.Website.IsRelativeUrl(crawlUrl));
if (websiteDefinition != null)
{
crawlUrl.WebsiteDefinition = websiteDefinition;
if (crawlUrlRepository.TryAdd(hash.Key, crawlUrl))
{
Interlocked.Increment(ref websiteDefinition.UrlsToProcessCount);
Interlocked.Increment(ref scheduledLinksCount);
/*
var inputCount = websiteProcessingDefinitions[websiteDefinition].Post(crawlUrl);
log.DebugFormat("Process block has {0} pending messages", inputCount);
*/
RaisePageScheduled(crawlUrl);
}
}
}
catch (Exception ex)
{
log.Error(ex);
}
}
return scheduledLinksCount;
}
public bool TryAdd(string key, CrawlUrl crawlUrl)
{
Client.As<string>().Lists[ScheduledListId].Add(key);
var typedClient = Client.As<CrawlUrl>();
typedClient.SetEntryInHashIfNotExists(typedClient.GetHash<string>(ScheduledHashId), key, crawlUrl);
return true;
}
public bool IsRelativeUrl(CrawlUrl crawlUrl)
{
return crawlUrl.Uri.AbsoluteUri.StartsWith(rootUri.AbsoluteUri);
// return rootUri.IsBaseOf(crawlUrl.Uri);
}
private void RaisePageProcessing(CrawlUrl crawlUrl)
{
var handler = PageProcessing;
if (handler != null)
handler.Invoke(crawlUrl);
}
public int Post(CrawlUrl crawlUrl)
{
ProcessingBlock.Post(crawlUrl);
return ProcessingBlock.InputCount;
}
public bool IsRelativeUrl(CrawlUrl crawlUrl)
{
return(crawlUrl.Uri.AbsoluteUri.StartsWith(rootUri.AbsoluteUri));
// return rootUri.IsBaseOf(crawlUrl.Uri);
}