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.time.temporal.ChronoUnit;
   8 import java.util.ArrayList;
   9 import java.util.Collections;
  10 import java.util.Comparator;
  11 import java.util.Date;
  12 import java.util.HashMap;
  13 import java.util.List;
  14 import java.util.Map;
  15 import java.util.concurrent.Executors;
  16 import java.util.concurrent.ScheduledExecutorService;
  17 import java.util.concurrent.TimeUnit;
  18 import java.util.logging.Level;
  19 import java.util.logging.Logger;
  20
  21 import org.jsoup.Jsoup;
  22
  23 import com.rometools.rome.feed.synd.SyndEnclosure;
  24 import com.rometools.rome.feed.synd.SyndEntry;
  25 import com.rometools.rome.feed.synd.SyndFeed;
  26 import com.rometools.rome.io.FeedException;
  27 import com.rometools.rome.io.SyndFeedInput;
  28 import com.rometools.rome.io.XmlReader;
  29
  30 import pnews.Article;
  31 import pnews.Category;
  32 import pnews.EntityStat;
  33 import pnews.Feed;
  34 import pnews.NER;
  35
  36 public class ArticleProvider {
  37         private static final String CLASS_NAME = ArticleProvider.class.getName();
  38         private static final Logger LOG = Logger.getLogger(CLASS_NAME);
  39         private final Map<Category, List<Article>> articlesByCategory = new HashMap<>();
  40         private final ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(Runtime.getRuntime().availableProcessors());
  41         private final Config config;
  42
  43         public ArticleProvider(Config config) {
  44                 this.config = config;
  45                 for (Category cat: config.getCategories())
  46                         scheduler.scheduleAtFixedRate(new Refresher(cat), 2, 600, TimeUnit.SECONDS);
  47         }
  48
  49         private static SyndFeed getSyndFeed(String u) throws IllegalArgumentException, FeedException, MalformedURLException, IOException {
  50                 XmlReader r;
  51
  52                 r = new XmlReader(new URL(u));
  53
  54                 return new SyndFeedInput().build(r);
  55         }
  56
  57         private List<Article> getArticlesForUpdate(Category cat) {
  58                 List<Article> result;
  59
  60                 synchronized (articlesByCategory) {
  61                         result = articlesByCategory.get(cat);
  62                         if (result == null) {
  63                                 result = new ArrayList<>();
  64                                 articlesByCategory.put(cat, result);
  65                         }
  66                         return result;
  67                 }
  68         }
  69
  70         private boolean exists(String articleLink, List<Article> articles) {
  71                 synchronized (articles) {
  72                         for (Article a: articles)
  73                                 if (a.link.equals(articleLink))
  74                                         return true;
  75                 }
  76                 return false;
  77         }
  78
  79         private Instant getArticleInstant(SyndEntry entry) {
  80                 Date date;
  81
  82                 date = entry.getUpdatedDate();
  83                 if (date == null)
  84                         date = entry.getPublishedDate();
  85
  86                 if (date == null)
  87                         return Instant.now();
  88
  89                 return date.toInstant();
  90         }
  91
  92         private Article toArticle(String link, SyndEntry entry, SyndFeed feed, String lang, Instant instant) {
  93                 String desc, title, thumbnail, feedTitle, str;
  94                 Date date;
  95                 List<String> entities;
  96
  97                 feedTitle = feed.getTitle();
  98                 if (feedTitle != null) {
  99                         feedTitle = feedTitle.trim();
 100                 }
 101
 102                 thumbnail = null;
 103                 for (SyndEnclosure e: entry.getEnclosures()) {
 104                         if (e.getType().startsWith("image/"))
 105                                 thumbnail = e.getUrl();
 106                         break;
 107                 }
 108
 109                 title = entry.getTitle().trim();
 110
 111                 if (entry.getDescription() != null) {
 112                         str = entry.getDescription().getValue();
 113                         desc = Jsoup.parse(str).text();
 114                 } else {
 115                         desc = null;
 116                         LOG.severe("No description for " + feedTitle + " - " + title);
 117                 }
 118
 119                 entities = new ArrayList<>();
 120                 if (desc != null && lang.equals("en"))
 121                         try {
 122                                 NER.classify(title, entities, config);
 123                                 NER.classify(desc, entities, config);
 124                         } catch (ClassCastException | ClassNotFoundException | IOException e1) {
 125                                 LOG.log(Level.SEVERE, "Cannot classify " + feedTitle, e1);
 126                         }
 127
 128                 return new Article(link, title, desc, thumbnail, instant, feedTitle, entities.toArray(new String[0]));
 129         }
 130
 131         private void addArticles(Category cat, SyndFeed feed) {
 132                 String feedTitle;
 133                 List<Article> articles;
 134                 Article a;
 135
 136                 feedTitle = feed.getTitle().trim();
 137
 138                 LOG.info("addArticles " + cat.getLabel() + " " + feedTitle + " number of articles: " + feed.getEntries().size());
 139
 140                 for (SyndEntry entry: feed.getEntries()) {
 141                         String link = entry.getLink().trim();
 142                         articles = getArticlesForUpdate(cat);
 143                         if (exists(link, articles)) {
 144                                 LOG.fine("addArticles " + link + " is already present");
 145                                 continue ;
 146                         }
 147
 148                         final Instant instant = getArticleInstant(entry);
 149
 150                         if (config.isObsolete(instant))
 151                                 continue ;
 152
 153                         a = ArticleStore.singleton.getArticle(link, ()->toArticle(link, entry, feed, cat.getLanguage(), instant));
 154
 155                         synchronized (articles) {
 156                                 articles.add(a);
 157
 158                                 Collections.sort(articles, new Comparator<Article>() {
 159                                         @Override
 160                                         public int compare(Article o1, Article o2) {
 161                                                 if (o1.publicationDate == o2.publicationDate)
 162                                                         return 0;
 163                                                 if (o1.publicationDate == null)
 164                                                         return 1;
 165                                                 if (o2.publicationDate == null)
 166                                                         return -1;
 167                                                 return o2.publicationDate.compareTo(o1.publicationDate);
 168                                         }
 169                                 });
 170                         }
 171                 }
 172
 173                 LOG.info("addArticles done " + cat.getLabel());
 174         }
 175
 176         private void retrieveArticles(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 177                 List<Feed> feeds;
 178
 179                 feeds = config.getFeedsByCategory().get(cat);
 180
 181                 if (feeds != null)
 182                         for (Feed f: feeds)
 183                                 try {
 184                                         addArticles(cat, getSyndFeed(f.getURL()));
 185                                 } catch (Throwable e) {
 186                                         LOG.log(Level.SEVERE,
 187                                                 "retrieveArticles failure " + cat.getLabel() + " " + f.toString(),
 188                                                 e);
 189                                 }
 190                 else
 191                         LOG.severe("No feed for category " + cat);
 192         }
 193
 194         /**
 195          * Returns a copy.
 196          */
 197         public List<Article> getArticles(Category cat, String entity)
 198                         throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 199                 List<Article> articles, result;
 200
 201                 synchronized (articlesByCategory) {
 202                         articles = getArticlesForUpdate(cat);
 203                 }
 204
 205                 synchronized (articles) {
 206                         if (entity == null)
 207                                 return new ArrayList<>(articles);
 208
 209                         result = new ArrayList<>(articles.size());
 210                         for (Article a: articles)
 211                                 if (a.hasEntity(entity))
 212                                         result.add(a);
 213
 214                         return result;
 215                 }
 216         }
 217
 218         public List<EntityStat> getEntityStats(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 219                 List<Article> articles;
 220                 Map<String, EntityStat> entities;
 221                 final String FUNCTION_NAME = "getEntities";
 222                 EntityStat s;
 223                 List<EntityStat> stats;
 224                 Instant minInstant;
 225
 226                 LOG.entering(CLASS_NAME, FUNCTION_NAME, cat);
 227
 228                 articles = getArticles(cat, null);
 229
 230                 minInstant = Instant.now().minus(15, ChronoUnit.DAYS);
 231
 232                 entities = new HashMap<>();
 233                 for (Article a: articles)
 234                         if (a.getPublicationDate().isAfter(minInstant) && a.getEntities() != null)
 235                                 for (String e: a.getEntities()) {
 236                                         s = entities.get(e);
 237                                         if (s == null) {
 238                                                 s = new EntityStat(e);
 239                                                 entities.put(e,  s);
 240                                         }
 241                                         s.increment();
 242                                 }
 243
 244                 stats = new ArrayList<>(entities.values());
 245                 stats.sort(new Comparator<EntityStat>() {
 246
 247                         @Override
 248                         public int compare(EntityStat o1, EntityStat o2) {
 249                                 return Integer.compare(o2.getCount(), o1.getCount());
 250                         }
 251
 252                 });
 253
 254                 LOG.exiting(CLASS_NAME, FUNCTION_NAME, stats);
 255
 256                 return stats;
 257         }
 258
 259         private class Refresher implements Runnable {
 260                 private final Category category;
 261
 262                 public Refresher(Category category) {
 263                         this.category = category;
 264                 }
 265
 266                 @Override
 267                 public void run() {
 268                         LOG.info("refresher "+ category.getLabel());
 269
 270                         try {
 271                                 retrieveArticles(category);
 272                         } catch (IllegalArgumentException | FeedException | IOException e) {
 273                                 LOG.log(Level.SEVERE, "refresher failure", e);
 274                         }
 275
 276                         LOG.info("refresher "+ category.getLabel() + " done");
 277                 }
 278         }
 279 }