public static string ModifyUrls(string sourceBlog, string body, PhotoIndexTableAdapter photoIndexTableAdapter, List <Photo> sitePhotos, out List <Model.Tumblr.Photo> extractedPhotos) { extractedPhotos = null; if (string.IsNullOrEmpty(body)) { return(null); } string decodedBody = body; if (body.StartsWith("\"")) { decodedBody = JsonConvert.DeserializeObject <string>(body); } HtmlDocument htmlDoc = new HtmlDocument(); htmlDoc.LoadHtml(decodedBody); List <HtmlNode> imgNodes = htmlDoc.DocumentNode.Descendants("img").ToList(); bool hasPhotosNotFound = false; foreach (HtmlNode imgNode in imgNodes) { string url = imgNode.Attributes["src"].Value; string mappedUrl = TryToGetMappedUrl(url, sitePhotos, sourceBlog, photoIndexTableAdapter); if (mappedUrl != null) { imgNode.Attributes["src"].Value = mappedUrl; } else { hasPhotosNotFound = true; } } if (hasPhotosNotFound) { extractedPhotos = PostProcessor.ExctractPhotosFromHtml(htmlDoc); } StringWriter sw = new StringWriter(); htmlDoc.Save(sw); string result = sw.ToString(); return(result); }
public static async Task Run([QueueTrigger(Constants.BlogToIndexQueueName, Connection = "AzureWebJobsStorage")] string myQueueItem, TraceWriter log) { Startup.Init(); BlogToIndex blogToIndex = JsonConvert.DeserializeObject <BlogToIndex>(myQueueItem); PhotoIndexTableAdapter photoIndexTableAdapter = new PhotoIndexTableAdapter(); photoIndexTableAdapter.Init(); PostsTableAdapter postsTableAdapter = new PostsTableAdapter(); postsTableAdapter.Init(log); ReversePostsTableAdapter reversePostsTableAdapter = new ReversePostsTableAdapter(); reversePostsTableAdapter.Init(log); PostToGetQueueAdapter postToGetQueueAdapter = new PostToGetQueueAdapter(); postToGetQueueAdapter.Init(); BlogInfoTableAdapter blogInfoTableAdapter = new BlogInfoTableAdapter(); blogInfoTableAdapter.Init(); MediaToDownloadQueueAdapter mediaToDownloadQueueAdapter = new MediaToDownloadQueueAdapter(); mediaToDownloadQueueAdapter.Init(log); List <PhotoIndexEntity> photoIndexEntities = photoIndexTableAdapter.GetAll(blogToIndex.Blogname); log.Info("Loaded " + photoIndexEntities.Count + " photo index entities"); BlogEntity blogEntity = await blogInfoTableAdapter.GetBlog(blogToIndex.Blogname); Dictionary <string, List <Model.Site.Photo> > photosByBlogById = CreatePhotosByBlogById(photoIndexEntities); BlogStats blogStats = CreateBlogStatsFromPhotos(photoIndexEntities, blogToIndex.Blogname); blogStats.UpdateFromBlogEntity(blogEntity); List <PostEntity> postEntities = postsTableAdapter.GetAll(blogToIndex.Blogname); UpdateBlogStatsFromPosts(blogStats, postEntities); UpdateMonthIndex(blogToIndex.Blogname, postEntities, blogInfoTableAdapter); log.Info("Loaded " + postEntities.Count + " post entities"); foreach (PostEntity postEntity in postEntities) { if (!string.IsNullOrEmpty(postEntity.PhotoBlobUrls)) { try { Model.Site.Photo[] photos = JsonConvert.DeserializeObject <Model.Site.Photo[]>(postEntity.PhotoBlobUrls); if (photos.Any(x => !x.Name.Contains("_"))) { SendToReprocessing(postEntity.PartitionKey, mediaToDownloadQueueAdapter, log, postEntity); } } catch (Exception e) { log.Error("Error: " + e.Message); throw; } } } blogStats.DisplayablePosts = InsertReversePosts(blogToIndex.Blogname, photosByBlogById, postEntities, reversePostsTableAdapter, postsTableAdapter, photoIndexTableAdapter, mediaToDownloadQueueAdapter, log); blogInfoTableAdapter.InsertBlobStats(blogStats); }
private static int InsertReversePosts(string blogname, Dictionary <string, List <Model.Site.Photo> > photosByBlogById, List <PostEntity> postEntities, ReversePostsTableAdapter reversePostsTableAdapter, PostsTableAdapter postsTableAdapter, PhotoIndexTableAdapter photoIndexTableAdapter, MediaToDownloadQueueAdapter mediaToDownloadQueueAdapter, TraceWriter log) { int index = 0; List <ReversePostEntity> reverseEntities = new List <ReversePostEntity>(100); foreach (PostEntity entity in postEntities) { ReversePostEntity reversePost = new ReversePostEntity(entity.PartitionKey, entity.RowKey, entity.Type, entity.Date, entity.ModifiedBody, entity.Title); if (photosByBlogById.TryGetValue(entity.RowKey, out List <Model.Site.Photo> photos)) { reversePost.Photos = JsonConvert.SerializeObject(photos, JsonUtils.JsonSerializerSettings); } else if (!string.IsNullOrEmpty(entity.VideoBlobUrls) && entity.VideoBlobUrls.StartsWith("[{")) { reversePost.Videos = entity.VideoBlobUrls; } if (string.IsNullOrEmpty(entity.ModifiedBody) && !string.IsNullOrEmpty(entity.Body)) { string sourceBlog = string.IsNullOrEmpty(entity.SourceTitle) ? blogname : SanityHelper.SanitizeSourceBlog(entity.SourceTitle); string modifiedBody = BodyUrlModifier.ModifyUrls(sourceBlog, entity.Body, photoIndexTableAdapter, photos, out List <Photo> extractedPhotos); if (extractedPhotos != null && extractedPhotos.Count > 0) { PhotosToDownload photosToDownload = new PhotosToDownload(entity) { Photos = extractedPhotos.ToArray() }; mediaToDownloadQueueAdapter.SendPhotosToDownload(photosToDownload); log.Warning("Could not modify body successfully, sending PhotosToDownload to get missing photos"); } else { entity.ModifiedBody = modifiedBody; postsTableAdapter.InsertPost(entity); log.Info($"ModifiedBody updated on post {entity.PartitionKey}/{entity.RowKey}"); } } if (!string.IsNullOrEmpty(reversePost.Photos) || !string.IsNullOrEmpty(reversePost.Videos) || !string.IsNullOrEmpty(reversePost.Body)) { reverseEntities.Add(reversePost); index++; if (index % 100 == 0) { reversePostsTableAdapter.InsertBatch(reverseEntities); reverseEntities.Clear(); log.Info("Inserted " + index + " reverse posts for " + entity.PartitionKey); } } } reversePostsTableAdapter.InsertBatch(reverseEntities); log.Info("Inserted " + index + " reverse posts for " + blogname); return(index); }
private static string TryToGetMappedUrl(string origUrl, List <Photo> sitePhotos, string sourceBlog, PhotoIndexTableAdapter photoIndexTableAdapter) { PhotoUrlHelper helper = PhotoUrlHelper.ParseTumblr(origUrl); if (helper != null) { if (sitePhotos != null) { foreach (Photo sitePhoto in sitePhotos) { int underscoreIndex = sitePhoto.Name.IndexOf("_", StringComparison.Ordinal); if (underscoreIndex >= 0) { string containerPart = sitePhoto.Name.Substring(0, underscoreIndex); string namePart = sitePhoto.Name.Substring(underscoreIndex + 1, sitePhoto.Name.Length - underscoreIndex - 1); if (namePart.Equals(helper.Name) && containerPart.Equals(helper.Container)) { PhotoSize photoSize = sitePhoto.Sizes.OrderBy(x => x.Nominal).Last(); string blobBaseUrl = ConfigurationManager.AppSettings["BlobBaseUrl"]; string newUrl = blobBaseUrl + "/" + photoSize.Container + "/" + sitePhoto.Name + "_" + photoSize.Nominal + "." + sitePhoto.Extension; return(newUrl); } } } } PhotoUrlIndexEntity photoIndex = photoIndexTableAdapter.GetPhotoUrlndex(sourceBlog, origUrl); if (photoIndex != null) { return(photoIndex.BlobUrl); } string url = "https://" + helper.Server + ".media.tumblr.com/" + (helper.Container != null ? helper.Container + "/" : "") + "tumblr_" + helper.Name + "_" + 640 + "." + helper.Extension; photoIndex = photoIndexTableAdapter.GetPhotoUrlndex(sourceBlog, url); if (photoIndex != null) { return(photoIndex.BlobUrl); } } return(null); }
public static async Task Run([QueueTrigger(Constants.PhotosToDownloadQueueName, Connection = "AzureWebJobsStorage")] string myQueueItem, TraceWriter log) { Startup.Init(); string requestUrl = null; try { PhotosToDownload photosToDownload = JsonConvert.DeserializeObject <PhotosToDownload>(myQueueItem); BlobAdapter blobAdapter = new BlobAdapter(); blobAdapter.Init(); PhotoIndexTableAdapter photoIndexTableAdapter = new PhotoIndexTableAdapter(); photoIndexTableAdapter.Init(); PostsTableAdapter postsTableAdapter = new PostsTableAdapter(); postsTableAdapter.Init(log); ReversePostsTableAdapter reversePostsTableAdapter = new ReversePostsTableAdapter(); reversePostsTableAdapter.Init(log); List <Photo> sitePhotos = new List <Photo>(); string blogname = photosToDownload.IndexInfo.BlogName; string id = photosToDownload.IndexInfo.PostId; DateTime date = photosToDownload.IndexInfo.PostDate; string sourceBlog = string.IsNullOrEmpty(photosToDownload.SourceBlog) ? photosToDownload.IndexInfo.BlogName : photosToDownload.SourceBlog; sourceBlog = SanityHelper.SanitizeSourceBlog(sourceBlog); using (HttpClient httpClient = new HttpClient()) { httpClient.DefaultRequestHeaders.Accept.Add(new MediaTypeWithQualityHeaderValue("image/*")); foreach (Model.Tumblr.Photo photo in photosToDownload.Photos) { bool isOriginal = true; Photo sitePhoto = null; foreach (AltSize altSize in photo.Alt_sizes) { PhotoUrlHelper urlHelper = PhotoUrlHelper.ParseTumblr(altSize.Url); if (isOriginal || urlHelper != null && DownloadSizes.Contains(urlHelper.Size)) { if (sitePhoto == null) { sitePhoto = new Photo { Name = urlHelper.Container + "_" + urlHelper.Name, Extension = urlHelper.Extension, Sizes = new PhotoSize[0] } } ; PhotoUrlIndexEntity urlIndexEntity = photoIndexTableAdapter.GetPhotoUrlndex(sourceBlog, altSize.Url); if (urlIndexEntity != null) // photo already downloaded { AddSizeToSitePhoto(sitePhoto, urlIndexEntity.BlobUrl, altSize); // need this to produce correct sitePhotos isOriginal = false; } else // photo not downloaded { requestUrl = altSize.Url; byte[] photoBytes = await httpClient.GetByteArrayAsync(altSize.Url); if (photoBytes.Length > 0) { Uri blobUri = await blobAdapter.UploadPhotoBlob(urlHelper, photoBytes, isOriginal); AddSizeToSitePhoto(sitePhoto, blobUri.ToString(), altSize); photoIndexTableAdapter.InsertPhotoIndex(blogname, id, date, SanityHelper.SanitizeSourceBlog(photosToDownload.SourceBlog), blobUri.ToString(), urlHelper.Name, urlHelper.Size, altSize.Width, altSize.Height, altSize.Url); isOriginal = false; log.Info("Downloaded photo from: " + altSize.Url); } } } } if (sitePhoto?.Sizes.Length > 0) { sitePhotos.Add(sitePhoto); } } } string modifiedBody = BodyUrlModifier.ModifyUrls(sourceBlog, photosToDownload.Body, photoIndexTableAdapter, sitePhotos, out List <Model.Tumblr.Photo> extractedPhotos); if (extractedPhotos != null) { log.Warning("Trying to modify body in ProcessPhotosToDownload but some images were not possible to replace"); } postsTableAdapter.MarkPhotosAsDownloaded(photosToDownload.IndexInfo.BlogName, photosToDownload.IndexInfo.PostId, sitePhotos, modifiedBody); ReversePostEntity reversePost = new ReversePostEntity(photosToDownload.IndexInfo.BlogName, photosToDownload.IndexInfo.PostId, photosToDownload.PostType, photosToDownload.IndexInfo.PostDate, modifiedBody, photosToDownload.Title) { Photos = JsonConvert.SerializeObject(sitePhotos) }; reversePostsTableAdapter.InsertPost(reversePost); } catch (Exception ex) { if (ex is HttpRequestException httpRequestException && httpRequestException.Message.Contains("403") && httpRequestException.Message.Contains("Forbidden")) { log.Warning("HTTP request was forbidden to URL: " + requestUrl); }