private async Task ScanPageAsync(
Uri currentUrl,
int currentDepth,
Uri baseUrl,
ConcurrentDictionary <Uri, byte> urlsNotToScan,
ConcurrentDictionary <Uri, int> urlsToScan,
ConcurrentDictionary <Uri, byte> scannedUrls,
int maxLinksOnPageCount)
{
var page = await _downloader.DownloadPageAsync(currentUrl);
var parsedHtml = HtmlParser.ParseHtml(page, baseUrl);
var filteredLinks = _urlFilter.Filter(parsedHtml.Links, baseUrl);
filteredLinks
.Where(url => !scannedUrls.ContainsKey(url))
.Where(url => !urlsNotToScan.ContainsKey(url))
.Take(maxLinksOnPageCount)
.ForEach(uri => urlsToScan.TryAdd(uri, currentDepth + 1));
#pragma warning disable 4014
_database.InsertAsync(new ScannedPage(currentUrl, parsedHtml.Text));
#pragma warning restore 4014
scannedUrls.TryAdd(currentUrl, default);
}
public async Task <ScannedPage[]> SearchAsync(Uri domainUrl)
{
var pages = await _database.GetAllAsync();
var urls = pages.Select(page => page.Url);
var filteredUrls = _urlFilter.Filter(urls, domainUrl, true);
return(pages
.Where(page => filteredUrls.Contains(page.Url))
.ToArray());
}
public void SanitisesUrls()
{
List <string> inputUrls = new List <string>()
{
"/first/",
"/SECOND",
"http://MoNZO.com/third",
"https://MoNZO.com/fourth",
"http://WwW.MoNZO.com/fifth",
"https://WwW.MoNZO.com/sixth"
};
var filteredUrls = _urlFilter.Filter("monzo.com", inputUrls);
filteredUrls.Count.ShouldBe(6);
filteredUrls[0].ShouldBe("/first");
filteredUrls[1].ShouldBe("/second");
filteredUrls[2].ShouldBe("/third");
filteredUrls[3].ShouldBe("/fourth");
filteredUrls[4].ShouldBe("/fifth");
filteredUrls[5].ShouldBe("/sixth");
}
