public void StartWorkers(
PauseToken pauseToken,
WorkerRelevantJobData jobData,
ConcurrentQueue <Uri> queue,
ConcurrentQueue <string> htmlQueue,
ConcurrentDictionary <Uri, CrawlData> crawled)
{
foreach (var worker in _workers)
{
worker.Start(
_browser,
_cancellationTokenSource.Token,
pauseToken,
jobData,
queue,
htmlQueue,
crawled);
}
}
public void Start(Browser browser,
CancellationToken cancellationToken,
PauseToken pauseToken,
WorkerRelevantJobData jobData,
ConcurrentQueue <Uri> queue,
ConcurrentQueue <string> htmlQueue,
ConcurrentDictionary <Uri, CrawlData> crawled)
{
Thread = new Thread(
async() => await WorkAction(
browser,
cancellationToken,
pauseToken,
jobData,
queue,
crawled));
Thread.Start();
_tab = browser.NewPageAsync().ConfigureAwait(false).GetAwaiter().GetResult();
}
private async Task WorkAction(
Browser browser,
CancellationToken cancellationToken,
PauseToken pauseToken,
WorkerRelevantJobData jobData,
ConcurrentQueue <Uri> queue,
ConcurrentDictionary <Uri, CrawlData> crawled)
{
// create a new tab on the browser for this thread
var nextUri = DequeueOrRetry(queue, cancellationToken);
if (cancellationToken.IsCancellationRequested)
{
return;
}
// go to page
var response = await _tab.GoToAsync(nextUri.AbsolutePath);
if (response.Ok)
{
// perform the jobs actions
// each action could yield a return value, such as extracted data
// the url should be added to the crawl collection
}
else
{
// indicate in the crawled collection this was a failure + reason
crawled.TryAdd(nextUri, CrawlData.CreateFailureData(response.Status, response.StatusText));
}
// if we should look for some links on the page
if (jobData.LinkEnqueueType != LinkEnqueueType.None)
{
// get the page content and just put it in a collection
// parser group will sort through and add the links
}
}
public Task <ProgressResult> Start(
JobDTO job,
CancellationToken cancellationToken,
PauseToken pauseToken,
ProgressToken progressToken)
{
if (IsCrawling)
{
throw new CrawlerAlreadyRunningException();
}
return(Task.Run(() =>
{
var workerRelevantJobData = WorkerRelevantJobData.FromJobDTO(job);
// create worker groups (max 64 threads, could give them each a browser, can give them each a set of domains to work on)
var group = new WorkerGroup(
_browser,
WorkersPerGroup);
var uriQueue = new ConcurrentQueue <Uri>(job.Seeds);
var htmlQueue = new ConcurrentQueue <string>();
var crawled = new ConcurrentDictionary <Uri, CrawlData>();
group.StartWorkers(
pauseToken,
workerRelevantJobData,
uriQueue,
crawled);
IsCrawling = true;
// main thread will spin here, checking for cancellation, or progress requests
// quit if the queue is empty and no one seems to be adding to it?
// don't want to just quit in case the queue is empty because there might be something about to be added
// but it's just taking a long time
while (!cancellationToken.IsCancellationRequested /* TODO add other stop conditions such as max crawl time */)
{
// cancellation and pauses are passed down to worker groups
// progress is checked via the collections and then updated
if (progressToken.ProgressIsRequested)
{
progressToken.State = GetState(uriQueue, crawled);
}
Thread.Sleep(10);
}
// tell the worker group to stop
group.Cancel();
var start = DateTime.Now;
// after cancellation is requested we fall into another loop, this has a time limit, so if not all worker groups are done
while (!group.AllDone && (DateTime.Now - start) < AbortWorkersTimeout)
{
// notify of which threads are holding us up
Thread.Sleep(10);
}
// attempt to properly join the threads
if (group.AllDone)
{
group.DiposeWorkers();
}
// abort them if not done
else
{
group.AbortWorkers();
}
IsCrawling = false;
return GetResult(uriQueue, crawled);
}));
}
