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

   1 package net.wpitchoune.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 net.wpitchoune.pnews.Article;
  31 import net.wpitchoune.pnews.ArticleStore;
  32 import net.wpitchoune.pnews.Category;
  33 import net.wpitchoune.pnews.Config;
  34 import net.wpitchoune.pnews.EntityStat;
  35 import net.wpitchoune.pnews.Feed;
  36 import net.wpitchoune.pnews.classifier.NamedEntityRecognizer;
  37
  38 public class ArticleProvider {
  39         private static final String CLASS_NAME = ArticleProvider.class.getName();
  40         private static final Logger LOG = Logger.getLogger(CLASS_NAME);
  41         private final Map<Category, List<Article>> articlesByCategory = new HashMap<>();
  42         private final ScheduledExecutorService scheduler = Executors.newScheduledThreadPool(Runtime.getRuntime().availableProcessors());
  43         private final Config config;
  44
  45         public ArticleProvider(Config config) {
  46                 this.config = config;
  47                 for (Category cat: config.getCategories())
  48                         scheduler.scheduleAtFixedRate(new Refresher(cat), 2, 600, TimeUnit.SECONDS);
  49         }
  50
  51         private static SyndFeed getSyndFeed(String u) throws IllegalArgumentException, FeedException, MalformedURLException, IOException {
  52                 XmlReader r;
  53
  54                 r = new XmlReader(new URL(u));
  55
  56                 return new SyndFeedInput().build(r);
  57         }
  58
  59         private List<Article> getArticlesForUpdate(Category cat) {
  60                 List<Article> result;
  61
  62                 synchronized (articlesByCategory) {
  63                         result = articlesByCategory.get(cat);
  64                         if (result == null) {
  65                                 result = new ArrayList<>();
  66                                 articlesByCategory.put(cat, result);
  67                         }
  68                         return result;
  69                 }
  70         }
  71
  72         private boolean exists(String articleLink, List<Article> articles) {
  73                 synchronized (articles) {
  74                         for (Article a: articles)
  75                                 if (a.getLink().equals(articleLink))
  76                                         return true;
  77                 }
  78                 return false;
  79         }
  80
  81         private Instant getArticleInstant(SyndEntry entry) {
  82                 Date date;
  83
  84                 date = entry.getUpdatedDate();
  85                 if (date == null)
  86                         date = entry.getPublishedDate();
  87
  88                 if (date == null)
  89                         return Instant.now();
  90
  91                 return date.toInstant();
  92         }
  93
  94         private Article toArticle(String link, SyndEntry entry, SyndFeed feed, String lang, Instant instant) {
  95                 String desc, title, thumbnail, feedTitle, str;
  96                 List<String> entities;
  97
  98                 feedTitle = feed.getTitle();
  99                 if (feedTitle != null) {
 100                         feedTitle = feedTitle.trim();
 101                 }
 102
 103                 thumbnail = null;
 104                 for (SyndEnclosure e: entry.getEnclosures()) {
 105                         if (e.getType().startsWith("image/"))
 106                                 thumbnail = e.getUrl();
 107                         break;
 108                 }
 109
 110                 title = entry.getTitle().trim();
 111
 112                 if (entry.getDescription() != null) {
 113                         str = entry.getDescription().getValue();
 114                         desc = Jsoup.parse(str).text();
 115                 } else {
 116                         desc = null;
 117                         LOG.severe("No description for " + feedTitle + " - " + title);
 118                 }
 119
 120                 entities = new ArrayList<>();
 121                 if (lang.equals("en"))
 122                         try {
 123                                 NamedEntityRecognizer.classify(title, entities, config);
 124                                 if (desc != null)
 125                                         NamedEntityRecognizer.classify(desc, entities, config);
 126                         } catch (ClassCastException | ClassNotFoundException | IOException e1) {
 127                                 LOG.log(Level.SEVERE, "Cannot classify " + feedTitle, e1);
 128                         }
 129
 130                 return new Article(link, title, desc, thumbnail, instant, feedTitle, entities.toArray(new String[0]));
 131         }
 132
 133         private void addArticles(Category cat, SyndFeed feed) {
 134                 String feedTitle;
 135                 List<Article> articles;
 136                 Article a;
 137
 138                 feedTitle = feed.getTitle().trim();
 139
 140                 LOG.info("addArticles " + cat.getLabel() + " " + feedTitle + " number of articles: " + feed.getEntries().size());
 141
 142                 for (SyndEntry entry: feed.getEntries()) {
 143                         String link = entry.getLink().trim();
 144                         articles = getArticlesForUpdate(cat);
 145                         if (exists(link, articles)) {
 146                                 LOG.fine("addArticles " + link + " is already present");
 147                                 continue ;
 148                         }
 149
 150                         final Instant instant = getArticleInstant(entry);
 151
 152                         if (config.isObsolete(instant))
 153                                 continue ;
 154
 155                         a = ArticleStore.singleton.getArticle(link, ()->toArticle(link, entry, feed, cat.getLanguage(), instant));
 156
 157                         synchronized (articles) {
 158                                 articles.add(a);
 159
 160                                 Collections.sort(articles, new Comparator<Article>() {
 161                                         @Override
 162                                         public int compare(Article o1, Article o2) {
 163                                                 if (o1.getPublicationDate() == o2.getPublicationDate())
 164                                                         return 0;
 165                                                 if (o1.getPublicationDate() == null)
 166                                                         return 1;
 167                                                 if (o2.getPublicationDate() == null)
 168                                                         return -1;
 169                                                 return o2.getPublicationDate().compareTo(o1.getPublicationDate());
 170                                         }
 171                                 });
 172                         }
 173                 }
 174
 175                 LOG.info("addArticles done " + cat.getLabel());
 176         }
 177
 178         private void retrieveArticles(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 179                 List<Feed> feeds;
 180
 181                 feeds = config.getFeedsByCategory().get(cat);
 182
 183                 if (feeds != null)
 184                         for (Feed f: feeds)
 185                                 try {
 186                                         addArticles(cat, getSyndFeed(f.getURL()));
 187                                 } catch (Throwable e) {
 188                                         LOG.log(Level.SEVERE,
 189                                                 "retrieveArticles failure " + cat.getLabel() + " " + f.toString(),
 190                                                 e);
 191                                 }
 192                 else
 193                         LOG.severe("No feed for category " + cat);
 194         }
 195
 196         /**
 197          * Returns a copy.
 198          */
 199         public List<Article> getArticles(Category cat, String entity)
 200                         throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 201                 List<Article> articles, result;
 202
 203                 synchronized (articlesByCategory) {
 204                         articles = getArticlesForUpdate(cat);
 205                 }
 206
 207                 synchronized (articles) {
 208                         if (entity == null)
 209                                 return new ArrayList<>(articles);
 210
 211                         result = new ArrayList<>(articles.size());
 212                         for (Article a: articles)
 213                                 if (a.hasEntity(entity))
 214                                         result.add(a);
 215
 216                         return result;
 217                 }
 218         }
 219
 220         public List<EntityStat> getEntityStats(Category cat) throws IllegalArgumentException, MalformedURLException, FeedException, IOException {
 221                 List<Article> articles;
 222                 Map<String, EntityStat> entities;
 223                 final String FUNCTION_NAME = "getEntities";
 224                 EntityStat s;
 225                 List<EntityStat> stats;
 226                 Instant minInstant;
 227
 228                 LOG.entering(CLASS_NAME, FUNCTION_NAME, cat);
 229
 230                 articles = getArticles(cat, null);
 231
 232                 minInstant = Instant.now().minus(15, ChronoUnit.DAYS);
 233
 234                 entities = new HashMap<>();
 235                 for (Article a: articles)
 236                         if (a.getPublicationDate().isAfter(minInstant) && a.getEntities() != null)
 237                                 for (String e: a.getEntities()) {
 238                                         s = entities.get(e);
 239                                         if (s == null) {
 240                                                 s = new EntityStat(e);
 241                                                 entities.put(e,  s);
 242                                         }
 243                                         s.increment();
 244                                 }
 245
 246                 stats = new ArrayList<>(entities.values());
 247                 stats.sort(new Comparator<EntityStat>() {
 248
 249                         @Override
 250                         public int compare(EntityStat o1, EntityStat o2) {
 251                                 return Integer.compare(o2.getCount(), o1.getCount());
 252                         }
 253
 254                 });
 255
 256                 LOG.exiting(CLASS_NAME, FUNCTION_NAME, stats);
 257
 258                 return stats;
 259         }
 260
 261         private class Refresher implements Runnable {
 262                 private final Category category;
 263
 264                 public Refresher(Category category) {
 265                         this.category = category;
 266                 }
 267
 268                 @Override
 269                 public void run() {
 270                         LOG.info("refresher "+ category.getLabel());
 271
 272                         try {
 273                                 retrieveArticles(category);
 274                         } catch (IllegalArgumentException | FeedException | IOException e) {
 275                                 LOG.log(Level.SEVERE, "refresher failure", e);
 276                         }
 277
 278                         LOG.info("refresher "+ category.getLabel() + " done");
 279                 }
 280         }
 281 }