public async Task <CrawlResult> Crawl(Uri siteUri, CrawlSettings settings)
{
var result = new CrawlResult
{
CrawlStart = DateTime.UtcNow
};
var overallCrawlStopwatch = new Stopwatch();
overallCrawlStopwatch.Start();
var baseUri = new Uri(siteUri.GetLeftPart(UriPartial.Authority));
var robotsFile = await new RobotsFileParser(HttpClient).FromUriAsync(baseUri);
UpdateCrawlDelay(robotsFile, settings.UserAgent, settings.RequestProcessorOptions);
var crawlRunner = new CrawlRunner(baseUri, robotsFile, HttpClient, settings, Logger);
//Use any links referred to by the sitemap as a starting point
var urisFromSitemap = (await new SitemapQuery(HttpClient)
.GetAllSitemapsForDomainAsync(siteUri.Host))
.SelectMany(s => s.Urls.Select(u => u.Location).Distinct());
foreach (var uri in urisFromSitemap)
{
crawlRunner.AddRequest(uri);
}
result.CrawledUris = await crawlRunner.ProcessAsync(async (requestResult, crawlState) =>
{
using (requestResult.Content)
{
var headers = new CrawlHeaders(requestResult.ResponseHeaders, requestResult.ContentHeaders);
var content = settings.ContentProcessor.Parse(crawlState.Location, headers, requestResult.Content);
requestResult.Content.Seek(0, SeekOrigin.Begin);
content.RawContent = await new StreamReader(requestResult.Content).ReadToEndAsync();
crawlRunner.AddResult(crawlState.Location, content);
}
});
overallCrawlStopwatch.Stop();
result.ElapsedTime = overallCrawlStopwatch.Elapsed;
return(result);
}
public async Task <CrawlResult> Crawl(Uri siteUri, CrawlSettings settings)
{
var result = new CrawlResult
{
CrawlStart = DateTime.UtcNow
};
var overallCrawlStopwatch = new Stopwatch();
overallCrawlStopwatch.Start();
var baseUri = new Uri(siteUri.GetLeftPart(UriPartial.Authority));
var robotsFile = await new RobotsFileParser(HttpClient).FromUriAsync(baseUri);
UpdateCrawlDelay(robotsFile, settings.UserAgent, settings.RequestProcessorOptions);
var crawlRunner = new CrawlRunner(baseUri, robotsFile, HttpClient, settings, Logger);
//Use any links referred to by the sitemap as a starting point
var urisFromSitemap = (await new SitemapQuery(HttpClient)
.GetAllSitemapsForDomain(siteUri.Host))
.SelectMany(s => s.Urls.Select(u => u.Location).Distinct());
foreach (var uri in urisFromSitemap)
{
crawlRunner.AddRequest(uri);
}
result.CrawledUris = await crawlRunner.ProcessAsync(async (requestResult, crawlState) =>
{
var response = requestResult.ResponseMessage;
var crawlRequest = new CrawlRequest
{
RequestStart = requestResult.RequestStart,
ElapsedTime = requestResult.ElapsedTime,
StatusCode = response.StatusCode,
IsSuccessfulStatus = response.IsSuccessStatusCode
};
crawlState.Requests.Add(crawlRequest);
var redirectStatusCodes = new[]
{
HttpStatusCode.MovedPermanently,
HttpStatusCode.Redirect,
HttpStatusCode.TemporaryRedirect
};
if (redirectStatusCodes.Contains(crawlRequest.StatusCode.Value))
{
crawlRunner.AddRedirect(crawlState.Location, response.Headers.Location);
}
else if (crawlRequest.IsSuccessfulStatus)
{
using (var contentStream = await response.Content.ReadAsStreamAsync())
{
var headers = new CrawlHeaders(response.Headers, response.Content.Headers);
var content = settings.ContentProcessor.Parse(crawlState.Location, headers, contentStream);
contentStream.Seek(0, SeekOrigin.Begin);
content.RawContent = await new StreamReader(contentStream).ReadToEndAsync();
crawlRunner.AddResult(crawlState.Location, content);
}
}
else if ((int)crawlRequest.StatusCode >= 500 && (int)crawlRequest.StatusCode <= 599)
{
//On server errors, try to crawl the page again later
crawlRunner.AddRequest(crawlState.Location);
}
else
{
//On any other error, just save what we have seen and move on
//Consider the content of the request irrelevant
crawlRunner.AddResult(crawlState.Location, null);
}
});
overallCrawlStopwatch.Stop();
result.ElapsedTime = overallCrawlStopwatch.Elapsed;
return(result);
}
public async Task <CrawlResult> Crawl(Uri siteUri, CrawlSettings settings)
{
var result = new CrawlResult
{
CrawlStart = DateTime.UtcNow
};
var stopwatch = new Stopwatch();
stopwatch.Start();
var baseUri = new Uri(siteUri.GetLeftPart(UriPartial.Authority));
var robotsFile = await new RobotsParser(HttpClient).FromUriAsync(baseUri);
//Apply Robots.txt crawl-delay (if defined)
var userAgentEntry = robotsFile.GetEntryForUserAgent(settings.UserAgent);
var minimumCrawlDelay = userAgentEntry?.CrawlDelay ?? 0;
var taskDelay = Math.Max(minimumCrawlDelay * 1000, settings.TaskHandlerOptions.DelayBetweenTaskStart.TotalMilliseconds);
settings.TaskHandlerOptions.DelayBetweenTaskStart = new TimeSpan(0, 0, 0, 0, (int)taskDelay);
var seedUris = new List <UriCrawlState>
{
new UriCrawlState {
Location = baseUri
}
};
//Use any links referred to by the sitemap as a starting point
seedUris.AddRange((await new SitemapQuery(HttpClient)
.GetAllSitemapsForDomain(siteUri.Host))
.SelectMany(s => s.Urls.Select(u => new UriCrawlState {
Location = u.Location
}))
.Distinct()
);
var crawlContext = new CrawlContext
{
Settings = settings
};
await TaskHandler.For(seedUris.Distinct().ToArray(), async (crawlState, pagesToCrawl) =>
{
if (!CheckUriValidity(crawlState.Location, baseUri, crawlContext))
{
return;
}
if (crawlContext.CrawledUris.ContainsKey(crawlState.Location))
{
return;
}
crawlContext.SeenUris.TryAdd(crawlState.Location, 0);
var lastRequest = crawlState.Requests.LastOrDefault();
if (lastRequest != null && lastRequest.IsSuccessfulStatus)
{
return;
}
else if (crawlState.Requests.Count() == settings.NumberOfRetries)
{
crawlContext.CrawledUris.TryAdd(crawlState.Location, new CrawledUri
{
Location = crawlState.Location,
Status = CrawlStatus.MaxRetries,
Requests = crawlState.Requests,
RedirectChain = crawlState.Redirects
});
}
else if (robotsFile.IsAllowedAccess(crawlState.Location, settings.UserAgent))
{
var crawledUri = await PerformRequest(crawlState, pagesToCrawl, crawlContext);
if (crawledUri != null)
{
crawlContext.CrawledUris.TryAdd(crawlState.Location, crawledUri);
if (crawledUri.Content?.Links?.Any() == true)
{
foreach (var crawlLink in crawledUri.Content.Links)
{
if (CheckUriValidity(crawlLink.Location, baseUri, crawlContext))
{
if (crawlContext.SeenUris.ContainsKey(crawlLink.Location))
{
continue;
}
crawlContext.SeenUris.TryAdd(crawlLink.Location, 0);
pagesToCrawl.Enqueue(new UriCrawlState
{
Location = crawlLink.Location
});
}
}
}
}
}
else
{
crawlContext.CrawledUris.TryAdd(crawlState.Location, new CrawledUri
{
Location = crawlState.Location,
Status = CrawlStatus.RobotsBlocked
});
}
}, settings.TaskHandlerOptions);
stopwatch.Stop();
result.ElapsedTime = stopwatch.Elapsed;
result.CrawledUris = crawlContext.CrawledUris.Values;
return(result);
}
