public async Task <URLMetadata> ScrapeMetadataAsync(Uri uri)
{
var searchListRequest = youtubeService.Search.List("snippet");
searchListRequest.Q = uri.ToString();
searchListRequest.MaxResults = 1;
searchListRequest.Type = "video";
var searchListResponse = await searchListRequest.ExecuteAsync();
var videoID = searchListResponse.Items.First().Id.VideoId;
var videosListRequest = youtubeService.Videos.List("snippet");
videosListRequest.Id = videoID;
videosListRequest.MaxResults = 1;
var videoListResponse = await videosListRequest.ExecuteAsync();
URLMetadata metadata = new URLMetadata()
{
CanonicalURL = "https://youtu.be/" + videoID,
FaviconURL = "https://www.youtube.com/favicon.ico",
Title = videoListResponse.Items.First().Snippet.Title,
Summary = videoListResponse.Items.First().Snippet.Description,
Tags = videoListResponse.Items.First().Snippet.Tags?.ToList(),
Images = new System.Collections.Generic.List <string> {
videoListResponse.Items.First().Snippet.Thumbnails.Maxres.Url
}
};
return(metadata);
}
public async Task <URLMetadata> ScrapeMetadataAsync(Uri uri)
{
var searchListRequest = youtubeService.Search.List("snippet");
searchListRequest.Q = uri.ToString();
searchListRequest.MaxResults = 1;
searchListRequest.Type = "video";
var searchListResponse = await searchListRequest.ExecuteAsync();
string videoID;
if (searchListResponse.Items.Count > 0)
{
videoID = searchListResponse.Items.First().Id.VideoId;
}
else
{
videoID = System.Web.HttpUtility.ParseQueryString(uri.Query).Get("v");
}
var videosListRequest = youtubeService.Videos.List("snippet");
videosListRequest.Id = videoID;
videosListRequest.MaxResults = 1;
var videoListResponse = await videosListRequest.ExecuteAsync();
var images = new List <string>();
if (videoListResponse.Items.First().Snippet.Thumbnails.Maxres != null)
{
images.Add(videoListResponse.Items.First().Snippet.Thumbnails.Maxres.Url);
}
if (videoListResponse.Items.First().Snippet.Thumbnails.High != null)
{
images.Add(videoListResponse.Items.First().Snippet.Thumbnails.High.Url);
}
if (videoListResponse.Items.First().Snippet.Thumbnails.Medium != null)
{
images.Add(videoListResponse.Items.First().Snippet.Thumbnails.Medium.Url);
}
if (videoListResponse.Items.First().Snippet.Thumbnails.Standard != null)
{
images.Add(videoListResponse.Items.First().Snippet.Thumbnails.Standard.Url);
}
URLMetadata metadata = new URLMetadata()
{
CanonicalURL = "https://youtu.be/" + videoID,
FaviconURL = "https://www.youtube.com/favicon.ico",
Title = videoListResponse.Items.First().Snippet.Title,
Summary = videoListResponse.Items.First().Snippet.Description,
Tags = videoListResponse.Items.First().Snippet.Tags?.ToList(),
Images = images
};
return(metadata);
}
public async Task <URLMetadata> ScrapeMetadataAsync(Uri uri)
{
var responseString = await httpClient.GetStringAsync(uri.GetLeftPart(UriPartial.Query));
htmlDoc.LoadHtml(responseString);
var canonicalUrl = htmlDoc.DocumentNode.SelectSingleNode("//link[@rel='canonical']")?.GetAttributeValue("href", null) ?? htmlDoc.DocumentNode.SelectSingleNode("//meta[@property='og:url']")?.GetAttributeValue("content", null);
try
{
if (canonicalUrl != null)
{
if (!canonicalUrl.StartsWith("http"))
{
canonicalUrl = new Uri(uri, canonicalUrl).ToString();
}
if (canonicalUrl != null && !canonicalUrl.Equals(uri.ToString(), StringComparison.InvariantCultureIgnoreCase))
{
responseString = await httpClient.GetStringAsync(canonicalUrl);
htmlDoc.LoadHtml(responseString);
}
}
}
catch (Exception ex) when(ex is InvalidOperationException || ex is HttpRequestException)
{
logger.LogError(ex, "Error trying to load canonical URL. Using current URL instead");
canonicalUrl = null;
}
var urlMetadata = new URLMetadata
{
CanonicalURL = canonicalUrl ?? uri.ToString(),
Title = htmlDoc.DocumentNode.SelectSingleNode("//meta[@property='og:title']")?.GetAttributeValue("content", null) ?? htmlDoc.DocumentNode.SelectSingleNode("//title")?.InnerText,
Summary = htmlDoc.DocumentNode.SelectSingleNode("//meta[@property='og:description']")?.GetAttributeValue("content", null) ?? htmlDoc.DocumentNode.SelectSingleNode("//meta[@name='description']")?.GetAttributeValue("content", null)
};
if (string.IsNullOrEmpty(urlMetadata.Summary))
{
var paragraphs = htmlDoc.DocumentNode.SelectNodes("//p");
if (paragraphs != null)
{
var longSummary = Regex.Replace(string.Join(' ', paragraphs.Select(p => p.InnerText)), @"\s+", " ");
urlMetadata.Summary = new string(longSummary.Take(300).ToArray());
}
}
var ogImage = htmlDoc.DocumentNode.SelectSingleNode("//meta[@property='og:image' or @name='twitter:image']")?.GetAttributeValue("content", null);
if (!string.IsNullOrEmpty(ogImage))
{
urlMetadata.Images.Add(ogImage);
}
var articleImages = htmlDoc.DocumentNode.SelectNodes("//img[not(@src='') and @src and not(starts-with(@src,'data:'))]");
if (articleImages != null)
{
urlMetadata.Images.AddRange(articleImages.Select(n =>
{
var imageSrc = n.GetAttributeValue("src", null);
if (!imageSrc.StartsWith("http", StringComparison.InvariantCultureIgnoreCase))
{
try
{
imageSrc = new Uri(uri, imageSrc).ToString();
}
catch
{
imageSrc = "";
}
}
return(imageSrc);
}).Distinct());
}
var tags = htmlDoc.DocumentNode.SelectNodes("//meta[@property='article:tag']");
if (tags != null)
{
urlMetadata.Tags.AddRange(tags.Select(t => t.GetAttributeValue("content", null)).ToList());
}
var keywords = htmlDoc.DocumentNode.SelectNodes("//meta[@name='news_keywords' or @name='keywords']");
if (keywords != null)
{
foreach (var newsKeyword in keywords)
{
urlMetadata.Tags.AddRange(newsKeyword.GetAttributeValue("content", null).Split(",").Select(t => t.Trim()));
}
}
urlMetadata.Tags = urlMetadata.Tags.Distinct().ToList();
var faviconTag = htmlDoc.DocumentNode.SelectNodes("//link[contains(@rel, 'icon') and not(contains(@rel, '-icon'))]")?.FirstOrDefault();
if (faviconTag != null)
{
urlMetadata.FaviconURL = faviconTag.GetAttributeValue("href", null);
if (!urlMetadata.FaviconURL.StartsWith("http", StringComparison.InvariantCultureIgnoreCase))
{
urlMetadata.FaviconURL = new Uri(uri, urlMetadata.FaviconURL).ToString();
}
}
if (String.IsNullOrEmpty(urlMetadata.FaviconURL))
{
logger.LogInformation("No favicons found in metadata. Trying to get default favicon from: " + new Uri(uri, "/favicon.ico").ToString());
try
{
HttpClient faviconClient = new HttpClient();
var faviconResponse = await faviconClient.GetAsync(new Uri(uri, "/favicon.ico"), HttpCompletionOption.ResponseHeadersRead);
if (faviconResponse.IsSuccessStatusCode)
{
urlMetadata.FaviconURL = new Uri(uri, "/favicon.ico").ToString();
}
}
catch (Exception ex)
{
//Leave without favicon if something's wrong
logger.LogError(ex, "Error trying to retrieve favicon. Ignoring favicons");
}
}
urlMetadata.Title = HtmlEntity.DeEntitize(urlMetadata.Title);
urlMetadata.Summary = HtmlEntity.DeEntitize(urlMetadata.Summary);
return(urlMetadata);
}