war/src/main/java/pnews/servlet/ArticleProvider.java

   1 package pnews.servlet;
   2
   3 import java.io.IOException;
   4 import java.net.MalformedURLException;
   5 import java.net.URL;
   6 import java.time.Instant;
   7 import java.util.ArrayList;
   8 import java.util.Collections;
   9 import java.util.Comparator;
  10 import java.util.Date;
  11 import java.util.HashMap;
  12 import java.util.List;
  13 import java.util.Map;
  14 import java.util.concurrent.Executors;
  15 import java.util.concurrent.ScheduledExecutorService;
  16 import java.util.concurrent.TimeUnit;
  17 import java.util.logging.Level;
  18 import java.util.logging.Logger;
  19
  20 import org.jsoup.Jsoup;
  21
  22 import com.rometools.rome.feed.synd.SyndEnclosure;
  23 import com.rometools.rome.feed.synd.SyndEntry;
  24 import com.rometools.rome.feed.synd.SyndFeed;
  25 import com.rometools.rome.io.FeedException;
  26 import com.rometools.rome.io.SyndFeedInput;
  27 import com.rometools.rome.io.XmlReader;
  28
  29 import pnews.Article;
  30 import pnews.Category;
  31 import pnews.EntityStat;
  32 import pnews.Feed;
  33 import pnews.NER;
  34
  35 public class ArticleProvider {
  36         private static final String CLASS_NAME = ArticleProvider.class.getName();
  37         private static final Logger LOG = Logger.getLogger(CLASS_NAME);
  38         private final Map<Category, List<Article>> articlesByCategory = new HashMap<>();
  39         private final ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(Runtime.getRuntime().availableProcessors());
  40         private final Config config;
  41
  42         public ArticleProvider(Config config) {
  43                 this.config = config;
  44                 for (Category cat: config.getCategories())
  45                         scheduler.scheduleAtFixedRate(new Refresher(cat), 2, 600, TimeUnit.SECONDS);
  46         }
  47
  48         private static SyndFeed getSyndFeed(String u) throws IllegalArgumentException, FeedException, MalformedURLException, IOException {
  49                 XmlReader r;
  50
  51                 r = new XmlReader(new URL(u));
  52
  53                 return new SyndFeedInput().build(r);
  54         }
  55
  56         private List<Article> getArticlesForUpdate(Category cat) {
  57                 List<Article> result;
  58
  59                 synchronized (articlesByCategory) {
  60                         result = articlesByCategory.get(cat);
  61                         if (result == null) {
  62                                 result = new ArrayList<>();
  63                                 articlesByCategory.put(cat, result);
  64                         }
  65                         return result;
  66                 }
  67         }
  68
  69         private boolean exists(String articleLink, List<Article> articles) {
  70                 synchronized (articles) {
  71                         for (Article a: articles)
  72                                 if (a.link.equals(articleLink))
  73                                         return true;
  74                 }
  75                 return false;
  76         }
  77
  78         private Instant getArticleInstant(SyndEntry entry) {
  79                 Date date;
  80
  81                 date = entry.getUpdatedDate();
  82                 if (date == null)
  83                         date = entry.getPublishedDate();
  84
  85                 if (date == null)
  86                         return Instant.now();
  87
  88                 return date.toInstant();
  89         }
  90
  91         private Article toArticle(String link, SyndEntry entry, SyndFeed feed, String lang, Instant instant) {
  92                 String desc, title, thumbnail, feedTitle, str;
  93                 Date date;
  94                 List<String> entities;
  95
  96                 feedTitle = feed.getTitle();
  97                 if (feedTitle != null) {
  98                         feedTitle = feedTitle.trim();
  99                 }
 100
 101                 thumbnail = null;
 102                 for (SyndEnclosure e: entry.getEnclosures()) {
 103                         if (e.getType().startsWith("image/"))
 104                                 thumbnail = e.getUrl();
 105                         break;
 106                 }
 107
 108                 title = entry.getTitle().trim();
 109
 110                 if (entry.getDescription() != null) {
 111                         str = entry.getDescription().getValue();
 112                         desc = Jsoup.parse(str).text();
 113                 } else {
 114                         desc = null;
 115                         LOG.severe("No description for " + feedTitle + " - " + title);
 116                 }
 117
 118                 entities = new ArrayList<>();
 119                 if (desc != null && lang.equals("en"))
 120                         try {
 121                                 NER.classify(title, entities, config);
 122                                 NER.classify(desc, entities, config);
 123                         } catch (ClassCastException | ClassNotFoundException | IOException e1) {
 124                                 LOG.log(Level.SEVERE, "Cannot classify " + feedTitle, e1);
 125                         }
 126
 127                 return new Article(link, title, desc, thumbnail, instant, feedTitle, entities.toArray(new String[0]));
 128         }
 129
 130         private void addArticles(Category cat, SyndFeed feed) {
 131                 String feedTitle;
 132                 List<Article> articles;
 133                 Article a;
 134
 135                 feedTitle = feed.getTitle().trim();
 136
 137                 LOG.info("addArticles " + cat.getLabel() + " " + feedTitle + " number of articles: " + feed.getEntries().size());
 138
 139                 for (SyndEntry entry: feed.getEntries()) {
 140                         String link = entry.getLink().trim();
 141                         articles = getArticlesForUpdate(cat);
 142                         if (exists(link, articles)) {
 143                                 LOG.fine("addArticles " + link + " is already present");
 144                                 continue ;
 145                         }
 146
 147                         final Instant instant = getArticleInstant(entry);
 148
 149                         if (config.isObsolete(instant))
 150                                 continue ;
 151
 152                         a = ArticleStore.singleton.getArticle(link, ()->toArticle(link, entry, feed, cat.getLanguage(), instant));
 153
 154                         synchronized (articles) {
 155                                 articles.add(a);
 156
 157                                 Collections.sort(articles, new Comparator<Article>() {
 158                                         @Override
 159                                         public int compare(Article o1, Article o2) {
 160                                                 if (o1.publicationDate == o2.publicationDate)
 161                                                         return 0;
 162                                                 if (o1.publicationDate == null)
 163                                                         return 1;
 164                                                 if (o2.publicationDate == null)
 165                                                         return -1;
 166                                                 return o2.publicationDate.compareTo(o1.publicationDate);
 167                                         }
 168                                 });
 169                         }
 170                 }
 171
 172                 LOG.info("addArticles done " + cat.getLabel());
 173         }
 174
 175         private void retrieveArticles(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 176                 List<Feed> feeds;
 177
 178                 feeds = config.getFeedsByCategory().get(cat);
 179
 180                 if (feeds != null)
 181                         for (Feed f: feeds)
 182                                 try {
 183                                         addArticles(cat, getSyndFeed(f.getURL()));
 184                                 } catch (Throwable e) {
 185                                         LOG.log(Level.SEVERE,
 186                                                 "retrieveArticles failure " + cat.getLabel() + " " + f.toString(),
 187                                                 e);
 188                                 }
 189                 else
 190                         LOG.severe("No feed for category " + cat);
 191         }
 192
 193         /**
 194          * Returns a copy.
 195          */
 196         public List<Article> getArticles(Category cat, String entity)
 197                         throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 198                 List<Article> articles, result;
 199
 200                 synchronized (articlesByCategory) {
 201                         articles = getArticlesForUpdate(cat);
 202                 }
 203
 204                 synchronized (articles) {
 205                         if (entity == null)
 206                                 return new ArrayList<>(articles);
 207
 208                         result = new ArrayList<>(articles.size());
 209                         for (Article a: articles)
 210                                 if (a.hasEntity(entity))
 211                                         result.add(a);
 212
 213                         return result;
 214                 }
 215         }
 216
 217         public List<EntityStat> getEntityStats(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 218                 List<Article> articles;
 219                 Map<String, EntityStat> entities;
 220                 final String FUNCTION_NAME = "getEntities";
 221                 EntityStat s;
 222                 List<EntityStat> stats;
 223
 224                 LOG.entering(CLASS_NAME, FUNCTION_NAME, cat);
 225
 226                 articles = getArticles(cat, null);
 227
 228                 entities = new HashMap<>();
 229                 for (Article a: articles)
 230                         if (a.getEntities() != null) {
 231                                 for (String e: a.getEntities()) {
 232                                         s = entities.get(e);
 233                                         if (s == null) {
 234                                                 s = new EntityStat(e);
 235                                                 entities.put(e,  s);
 236                                         }
 237                                         s.increment();
 238                                 }
 239                         }
 240
 241                 stats = new ArrayList<>(entities.values());
 242                 stats.sort(new Comparator<EntityStat>() {
 243
 244                         @Override
 245                         public int compare(EntityStat o1, EntityStat o2) {
 246                                 return Integer.compare(o2.getCount(), o1.getCount());
 247                         }
 248
 249                 });
 250
 251                 LOG.exiting(CLASS_NAME, FUNCTION_NAME, stats);
 252
 253                 return stats;
 254         }
 255
 256         private class Refresher implements Runnable {
 257                 private final Category category;
 258
 259                 public Refresher(Category category) {
 260                         this.category = category;
 261                 }
 262
 263                 @Override
 264                 public void run() {
 265                         LOG.info("refresher "+ category.getLabel());
 266
 267                         try {
 268                                 retrieveArticles(category);
 269                         } catch (IllegalArgumentException | FeedException | IOException e) {
 270                                 LOG.log(Level.SEVERE, "refresher failure", e);
 271                         }
 272
 273                         LOG.info("refresher "+ category.getLabel() + " done");
 274                 }
 275         }
 276 }