|
| 1 | +namespace DataScraper |
| 2 | +{ |
| 3 | + using System; |
| 4 | + using System.Collections.Concurrent; |
| 5 | + using System.Collections.Generic; |
| 6 | + using System.IO; |
| 7 | + using System.Linq; |
| 8 | + using System.Net.Http; |
| 9 | + using System.Threading.Tasks; |
| 10 | + |
| 11 | + using HtmlAgilityPack; |
| 12 | + |
| 13 | + using Newtonsoft.Json; |
| 14 | + |
| 15 | + public class Link |
| 16 | + { |
| 17 | + public string Icon { get; set; } |
| 18 | + |
| 19 | + public string Url { get; set; } |
| 20 | + } |
| 21 | + |
| 22 | + public class Session |
| 23 | + { |
| 24 | + public int Id { get; set; } |
| 25 | + |
| 26 | + public string Title { get; set; } |
| 27 | + } |
| 28 | + |
| 29 | + public class Speaker |
| 30 | + { |
| 31 | + public int Id { get; set; } |
| 32 | + |
| 33 | + public string Name { get; set; } |
| 34 | + |
| 35 | + public string Bio { get; set; } |
| 36 | + |
| 37 | + public IEnumerable<Link> Links { get; set; } |
| 38 | + |
| 39 | + public IEnumerable<Session> Sessions { get; set; } |
| 40 | + } |
| 41 | + |
| 42 | + public class Program |
| 43 | + { |
| 44 | + public static void Main(string[] args) |
| 45 | + { |
| 46 | + const string AgendaUrl = "/Agenda"; |
| 47 | + const string OutputDir = |
| 48 | + "D:\\Code\\BeyondResponsiveDesign\\BeyondResponsiveDesign.Menus\\images\\speakers"; |
| 49 | + |
| 50 | + var imageClient = new HttpClient(); |
| 51 | + var htmlClient = new HttpClient { BaseAddress = new Uri("http://www.dddeastanglia.com") }; |
| 52 | + |
| 53 | + var htmlDoc = new HtmlDocument(); |
| 54 | + htmlDoc.LoadHtml(htmlClient.GetStringAsync(AgendaUrl).Result); |
| 55 | + |
| 56 | + var speakerLinks = |
| 57 | + htmlDoc.DocumentNode.Descendants("a") |
| 58 | + .Where(o => o.Attributes.Contains("class") && o.Attributes["class"].Value == "speakerName"); |
| 59 | + |
| 60 | + var speakers = new ConcurrentQueue<Speaker>(); |
| 61 | + Task.WaitAll(speakerLinks.Select(link => Task.Run(async () => |
| 62 | + { |
| 63 | + var href = link.Attributes["href"].Value; |
| 64 | + var speakerPage = new HtmlDocument(); |
| 65 | + speakerPage.LoadHtml(await htmlClient.GetStringAsync(href)); |
| 66 | + |
| 67 | + var container = |
| 68 | + speakerPage.DocumentNode.Descendants("div") |
| 69 | + .First( |
| 70 | + o => o.Attributes.Contains("class") && o.Attributes["class"].Value.Contains("speaker")); |
| 71 | + |
| 72 | + var titleElement = container.Descendants("h3").First(); |
| 73 | + var linkElements = container.Descendants("section").First().Descendants("p"); |
| 74 | + var paragraphElements = container.Elements("p"); |
| 75 | + var sessionLinkElements = container.Element("ul").Descendants("a"); |
| 76 | + |
| 77 | + var id = int.Parse(href.Split('/').Last()); |
| 78 | + var name = titleElement.Element("#text").InnerHtml; |
| 79 | + var bio = string.Join(string.Empty, paragraphElements.Select(o => o.InnerHtml)); |
| 80 | + var links = (from linkElement in linkElements |
| 81 | + let icon = linkElement.Descendants("i").First().Attributes["class"].Value |
| 82 | + let url = linkElement.Descendants("a").First().Attributes["href"].Value |
| 83 | + select new Link { Icon = icon, Url = url }).ToList(); |
| 84 | + var sessions = (from sessionLinkElement in sessionLinkElements |
| 85 | + let sessionId = |
| 86 | + int.Parse(sessionLinkElement.Attributes["href"].Value.Split('/').Last()) |
| 87 | + let title = sessionLinkElement.InnerText |
| 88 | + select new Session { Id = sessionId, Title = title }).ToList(); |
| 89 | + |
| 90 | + speakers.Enqueue( |
| 91 | + new Speaker { Id = id, Name = name, Links = links, Bio = bio, Sessions = sessions }); |
| 92 | + |
| 93 | + var imageTag = titleElement.Descendants("img").First(); |
| 94 | + var imageUrl = imageTag.Attributes["src"].Value.Replace("s=50", "s=300"); |
| 95 | + |
| 96 | + var stream = await imageClient.GetStreamAsync(imageUrl); |
| 97 | + var file = File.OpenWrite(string.Format("{0}\\{1}.jpg", OutputDir, id)); |
| 98 | + await stream.CopyToAsync(file); |
| 99 | + |
| 100 | + file.Close(); |
| 101 | + })).ToArray()); |
| 102 | + |
| 103 | + var streamWriter = new StreamWriter(string.Format("{0}\\speakers.json", OutputDir)); |
| 104 | + streamWriter.Write(JsonConvert.SerializeObject(speakers.OrderBy(o => o.Name))); |
| 105 | + streamWriter.Close(); |
| 106 | + |
| 107 | + Console.WriteLine("Process completed, press any key to exit"); |
| 108 | + Console.ReadKey(); |
| 109 | + } |
| 110 | + } |
| 111 | +} |
0 commit comments